Factors Affecting Cost Performance of Construction in India

1.
Editorial board
Page CO2
2.

Transferring projects to their final users: The effect of planning and
preparations for commissioning on project success
Pages 257-265
Dov Dvir
3.

The effect of systemic errors on optimal project buffers
Pages 267-274
Dan Trietsch
4.

Simulation-based estimation for correlated cost elements
Pages 275-282
I.-T. Yang
5.

Factors affecting cost performance: evidence from Indian construction
projects
Pages 283-295
K.C. Iyer and K.N. J ha
6.

Cost estimation of a software product using COCOMO II.2000 model a case
study
Pages 297-307
R. Dillibabu and K. Krishnaiah
7.

Designbuild pre-qualification and tendering approach for public projects
Pages 309-320
Khaled Al-Reshaid and Nabil Kartam
8.

Evaluation of stakeholder influence in the implementation of construction
projects
Pages 321-328
Stefan Olander and Anne Landin
9.

Framework for project managers to manage construction safety
Pages 329-341
Evelyn Ai Lin Teo, Florence Yean Yng Ling and Adrian Fook Weng Chong

International Journal of Project Management
Copyright 2006 Elsevier Ltd and the International Project Management Association (IPMA). All rights reserved
Volume 23, Issue 4, Pages 257-341 (May 2005)

The International Journal of Project Management is devoted to the publication of papers which advance knowledge of
the practical and theoretical aspects of project management. The Journal aims to: provide a focus for worldwide
expertise in the required techniques, practices and areas of research; present a forum for readers to share common
experiences across the full range of industries and technologies in which project management is used; cover all areas of
project management from systems to human aspects; and link theory with practice by publishing case studies and
covering the latest important issues in special series.
Contributions Those wishing to submit full-length papers or case studies should send four copies to the editor,
Professor J R Turner, at the address below. Contributors should refer to the Notes for Authors printed in this issue
of the Journal. These are also available from the publishers.
INTERNATIONAL EDITORIAL BOARD
Professor D Arditi Professor and Chairman,
Department of Civil and Architectural
Engineering, Illinois Institute of Technology,
3201 South Dearborn Street
Chicago, IL 60616-3793, USA
E-mail: ia_arditi@vaxl.iit.edu
Professor Karlos A Artto Helsinki University
of Technology, PO Box 9500
FIN02015 HUT, Finland
E-mail: karlos.artto@hut.
Professor B Boeva University for National and
World Economic Studies, Studentslei grad
Ch. Botev, 1100 Soa, Bulgaria
E-mail: boeva@apollo.netissat.bg
Professor Sergey Bushuyev Kiev National
University of Construction and Architecture,
31 Povitrootsky pr., Kiev 04037, Ukraine
E-mail: bush@upma.freenet.kiev.ua
Dr Juan Luis Cano Department of Design and
Manufacturing Engineering, University of
Zaragoza, Maria de Luna 3, 50015 Zaragoza,
Spain
E-mail: jlcano@posta@unizar.es
Mr Giles Caupin Caupin & Associes,
3 Rue Creuse, F-77710TREUZY
LEVELAY, France
E-mail: 100661.1434@compuserve.com
Dr L. Crawford 10 Amaroo Crescent,
Mosman, NSW 2088, Australia
E-mail: lynn.crawford@uts.edu.au
D H Curling Principal Consultant,
LODAY Systems Ltd, 1786 Devlin Crescent,
Ottawa, Ontario K1H 5T6, Canada
E-mail: pm@davidcurling.info
Dr J D Frame UMT, Suite 306
1925 N. Lynn Street
Arlington, VA 22209, USA
E-mail: davidson.frame@umtweb.edu
Professor Qian Fupei PO Box 617, Northwestern
Polytechnical University, Xian, 710072,
Peoples Republic of China
E-mail: qianfp@nwpu.edu.cn
Professor Roland Gareis Projektmanagement
Austria-Institute, Wirtschaftuniversitet,
Franz Klein Gasse, 1190 Vein, Austria
E-mail: roland.gareis@wu-wein.ac.at
Professor Shlomo Globerson Faculty of
Management, Tel Aviv University, POB 39010
Ramat Aviv, Tel Aviv 69978, Israel
E-mail: globe@post-tau.ac.il
Professor Joop Halman University of Twente,
Faculty of Engineering Technology,
Department Construction Engineering
Management, PO Box 217 7500 AE Enschede
Professor Francis Hartman Faculty of
Engineering, University of Calgary,
2500 University Drive
NW, Calgary, Alberta, Canada
E-mail: fhartman@ucalgary.ca
Professor M Elizabeth C Hull School of
Computing and Mathematical Sciences,
University of Ulster, Newtownabbey,
Co Antrim, BT37 0QB, N. Ireland, UK
E-mail: mec.hull@ulst.ac.uk
Mr Adesh Jain Director in Charge,
Centre for Excellence in Project Management,
A-48, Sector V, Noida 201 301, India
E-mail: acjain@vsnl.com
Professor Jon Lereim Institutt for Teknologile-
delse, Elias Smiths vei 15, Postboks 580,
1302 Sandvika, Norway
E-mail: jon.lereim@bi.no
Professor Rolf A Lundin Jo nko ping
International Business School,
Jo nko ping University, PO Box 1026,
WE-551 11 Jo nko ping, Sweden
E-mail: rolf.a.lundin@jibs.hj.se
Professor Christophe Midler Centre de
Researche de Gestion, Ecole Polytechnique
1 rue Descartes, 75005 Paris, France
E-mail: midler@poly.polytechnique.fr
Professor Peter Morris Professor of Construction
and Project Management, The Bartlett School
of Graduate Studies, UCL The Faculty of the
Built Environment, Torrington Place Site,
Gower Street, London WC1E 6BT, UK
E-mail: pwmorris@netcomuk.com.uk
Professor Jerey Pinto School of
Business, Penn State Erie, Station Road,
Erie PA 16563, USA
E-mail: jkp4@psu.edu
Professor Nigel J Smith Department of Civil
Engineering, University of Leeds,
Leeds LS2 9JT, UK
E-mail: n.j.smith@leeds.ac.uk
Hiroshi Tanaka Project Services Division,
Yokohama World Operations Center,
3-1, Minato Mirai 2-chome, Nishi-ku,
Yokohama 220-6001, Japan
E-mail: tanaka.hiroshi@jgc.co.jp
Willi Vonrufs Project and Process
Management, Larchenstrasse 20,
CH-8903 Birmensdorf, Switzerland
E-mail: vonrufs@vonrufs.ch
Professor Terry M Williams Department
of Management Science, Graham Hills Building,
40 George Street, Glasgow, G1 1QE
E-mail: Terry@mansci.strath.ac.uk
Professor K-T Yeo Nanyang Technological
Unia versity, School of Mechanical and
Production Engineering, Nanyang Ave,
Singapore 2263
EDITOR
Professor J R Turner
Professor of Project Management
ISGI-Groupe ESC Lille
Avenue Willy Brandt
F 59777 EURALILLE
France
Postal Address: EuroProjEx
Wildwood
Manor Close
East Horsley
Surrey KT24 6SA, UK
E-mail Address: ijpm@europrojex.co.uk
INTERNATIONAL JOURNAL OF
PROJECT
MANAGEMENT
International Journal of Project Management is the ocial journal of the International Project Management Association. The
journal is available to all members of the IPMA at a special membership rate. For details of membership or enquires regarding the
association, please contact:
Ms Nancy Dol, IPMA International Secretariat, International Project Management Association, P.O. Box 1167, NL 3860 BD
Nijkerk, The Netherlands.
Members of associations aliated to IPMA should contact their National Association for subscription details
INTERNATIONAL JOURNAL OF PROJECT MANAGEMENT
NOTES FOR AUTHORS
Scope
The International Journal of Project Management is devoted to
the publication of papers which advance knowledge on prac-
tical and theoretical aspects of project management. The list of
key words at the end of this guide indicates the scope of the
journal. Papers are selected for publication based on their re-
levance, clarity, topicality, the extent to which they advance
knowledge, and their contribution to inspiring further devel-
opment and research. The journal strives to maintain a balance
between papers derived from research and from practical ex-
perience. Authors are encouraged to submit case studies de-
scribing the project environment; criteria and factors for
success; responsibilities of participants; managerial arrange-
ments; human factors; contract forms; planning and control
systems; problem areas encountered and lessons learned.
Submission of Papers
Authors are requested to submit their original manuscript and
gures plus three copies to Professor .J. R. Turner, EuroProjex,
Wildwood, Manor Close, East Horsley, Surrey, KT24 6SA,
UK. (When sending revisions. only two copies are required.)
The editorial oce does not have bulk photocopying facilities
and so cannot receive papers by e-mail onlypaper versions
must be sent by post.
Submission of a paper implies that it has not been published
previously, that is not under consideration for publication
elsewhere, and that if accepted it will not be published else-
where in the same form, in English or any other language,
without the written consent of the Publisher.
Refereeing: All papers submitted for publication will be refer-
eed on the double-blind system by two or more specialists
selected from a panel of referees. This means the author and
referees do not know each other, nor do the referees know
other referees. Thus it is important that authors names should
appear nowhere in the manuscript except on the cover page
(which can be separated from the manuscript) and in refer-
ences. When referring to their own work, authors should refer
to themselves in the third person. Any papers not adhering to
this will be returned.
Manuscript Preparation
General: Manuscripts should be 3000 to 5000 words long (6000
maximum), inclusive of gures and tables. Count each gure
and table as 300 words. Papers must be typewritten, double-line
spaced with wide margins on one side of white A4 paper. Good
quality printouts with a font size of 12 or 11 pt are required.
Number every sheet. Write in clear and concise English, using
active rather than passive voice. Authors may refer to them-
selves in the rst person, except when citing their own work.
Spelling should follow the Oxford English Dictionary. Authors
should consult a recent issue of the journal for style if possible.
An electronic copy of the paper should accompany the nal
version. The Editors reserve the right to adjust style to certain
standards of uniformity. Authors should retain a copy of their
manuscript since we cannot accept responsibility for damage or
loss of papers. Original manuscripts are discarded one month
after publication unless the Publisher is asked to return original
material after use.
Abstracts: Please supply an abstract of about 100 words out-
lining the purpose, scope and conclusions of the paper, and at
least two selected keywords. It is important that the abstract
should be very clear and understandable to those whom Eng-
lish is not their native language. The abstract should explain
why the paper is important to those who may not necessarily be
in that particular eld. A list of keywords to choose from ap-
pears at the end of this guide.
Arrangement of papers: Please arrange your paper as follows
(with each of the numbered items beginning on a new page):
1. Cover page with title, author(s), aliation(s), full ad-
dress(es), e-mail, telephone and fax numbers (this will be
removed from the copy, sent to referees).
2. Abstract page with the title repeated (but no authors
names), abstract and keywords.
3. The text, suitably divided under headings. Please do not
number headings.
4. Acknowledgements (if any).
5. References.
6. Captions to tables and illustrations (grouped on a separate
sheet).
7. Tables (each on a separate sheet, with captions).
8. Illustrations (each on a separate sheet, without captions).
Please ensure that your name appears only in item I (except when
citing a reference within a paper) and that this can be separated
from the rest of the paper. Do not import the Tables or Illus-
trations into your text. The corresponding author should be
identied with an asterisk and footnote. All other footnotes
(except table footnotes) should be identied with superscript
Arabic numerals. Trade names should have an initial capital
letter.
Units: You should use SI units, as dened by the ISO standard
or your national authorized SI standard. Where SI units do not
exist, use an internationally accepted unit. If you use any
symbol or unit that may not be generally recognized, please put
an explanatory note in the margin the rst time it is used, to
help the referees and editors.
References: All publications cited in the text should be pre-
sented in a list of references following the text of the manu-
script. In the text refer to references by a number in square
brackets on the line, e.g. Since Cooper [1]. Where you cite a
reference more than once in the text, use the same number each
time. The full reference should be given in a numerical list at
the end of the paper. References should take the following
form:
[1] Cooper DF. Chapman CB. Risk analysis for large projects:
models, methods and cases. New York: Wiley, I987.
[2] Potter M. Procurement of construction work: The clients
role: In: U J. Orams AM, editors. Proceedings of 7th Annual
Conference on Construction Law and Management, Kings
College, London, UK, 1995. p. 169194.
[3] Norwood SR, Manseld NR. Joint venture issues concerning
European and Asian construction markets of the 1990s. Inter-
national Journal of Project Management 1999;17(2):8993.
Please ensure that references are complete, i.e. that they in-
clude, where relevant, authors name, article or book title,
volume and issue number, publisher and location, date and
page reference.
Illustrations: All illustrations should be provided in camera-
ready form, suitable for reproduction (which may include re-
duction) without retouching. Photographs, charts and dia-
grams are all to be referred to as Figure(s) and should be
numbered consecutively in the order to which they are referred.
They should accompany the manuscript, but should not be
included within the text. All illustrations should be clearly
marked on the back in pencil with the gure number and the
authors name. All gures are to have a caption. Captions
should be supplied on a separate sheet and should be in low-
ercase letters with an initial capital for the rst word only.
Line drawings: Good quality printouts on while paper pro-
duced in black ink are required. All lettering, graph lines and
points on graphs should be suciently large and bold to permit
reproduction when the diagram has been reduced to a size
suitable for inclusion in the journal. Dye-line prints or photo-
copies are not suitable for reproduction. Do not use any type of
shading on computer-generated illustrations.
Graphs: The minimum amount of descriptive text should be
used on graphs (label curves, points etc. with single-letter
symbols). Descriptive matter should be placed in the gure
caption. Scale grids should not be used in graphs, unless re-
quired for actual measurements. Graph axes should be labelled
with variables written out in full, along the length of the axes,
with the unit in parentheses (e.g. Time(s)). A table is usually
more satisfactory for recording data.
Photographs: Original photographs must be supplied as they
are to be reproduced (e.g. black and white or colour) plus two
photocopies. If necessary, a scale should be marked on the
photograph. Please note that photocopies of photographs are
not suitable for reproduction.
Colour: Where colour gures are required, the author will be
charged accordingly (further details of costs are available from
Author Services at Elsevier). In cases where colour is paid
for, authors will receive an additional one hundred free o-
prints.
Tables: Tables should be numbered consecutively in Arabic
numerals and given a suitable caption at the top of the table.
Type each table on a separate sheet. Footnotes to tables should
be typed below the table and should be referred to by super-
script lowercase letters. No vertical rules should be used. Tables
should not duplicate results presented elsewhere in the manu-
script (e.g. in graphs).
Electronic submission
Authors should submit an electronic copy of their paper with the
nal version of the manuscript. This may be sent on disk or by e-
mail, but only with the nal accepted version of the paper. The
electronic copy must match the hardcopy exactly and should
be accompanied by two paper copies of the manuscript.
Always keep a backup copy of the electronic le for refer-
ence and safety. Guidelines on electronic submission are sent
to the authors with the request to revise the paper. Full
details of electronic submission and formats can be obtained
from http://authors.elsevier.com or from Author Services at
Elsevier.
Proofs
Proofs will be sent to the author (rst-named author if no
corresponding author is identied on multi-authored papers)
by PDF wherever possible and should be returned within 48
hours of receipt, preferably by e-mail. Corrections should be
restricted to typesetting errors; any other amendments made
may be charged to the author. Any queries should be answered
in full. Elsevier will do everything possible to get your article
corrected and published as quickly as possible. Therefore, it is
important to ensure that all of your corrections are returned to
us in one all-inclusive e-mail or fax. Subsequent additional
corrections will not be possible, so please ensure that your rst
communication is complete. Should you choose to mail your
corrections, please return them to: Log-in Department.
Elsevier, Stover Court, Bampfylde Street, Exeter. Devon EX1
2AH, UK.
Oprints
Fifty oprints will be supplied free of charge to the corre-
sponding author. Additional oprints can be ordered at a
specially reduced rate using the order form sent to the corre-
sponding author after the manuscript has been accepted. Or-
ders for reprints (produced after publication of an article) will
incur a 50%, surcharge.
Copyright
All authors must sign the Transfer of Copyright agreement
before the article can be published. This transfer agreement
enables Elsevier Ltd and the International Project Management
Association to protect the copyrighted material of an author,
without the author relinquishing his/her proprietary rights. The
copyright transfer covers the exclusive rights to reproduce and
distribute the article, including reprints, photographic re-
productions, microlm or any other reproductions of a similar
nature, and translations. It also includes the right to adapt the
article for use in conjunction with computer systems and pro-
grams, including reproduction of publication in machine-
readable form and incorporation in retrieval systems. Authors
are responsible for obtaining from the copyright holder per-
mission to reproduce any material for which copyright already
exists.
Checklist
Have you told readers, at the outset, what they might gain
by reading your paper?
Have you made the aim of your work clear?
Have you explained the signicance of your contribution?
Have you set your work in the appropriate context with
sucient background, and all relevant references?
Have you addressed the question of practicality and use-
fulness?
Have you identied future developments that may result
from your work?
Have you structured your paper in a clear and logical
fashion?
Author Services
For enquiries relating to the submission of articles (including
electronic submission where available) please visit the Author
Gateway from Elsevier at http://authors.elsevier.com. The
Author Gateway also provides the facility to track accepted
articles and set up e-mail alerts to inform you of when an ar-
ticles status has changed, as well as detailed artwork guide-
lines, copyright information, frequently asked questions and
more.
Keywords
Please choose at least two key words from the following lists, as
appropriate. This will assist the editor in choosing referees, as
well as helping with cataloguing.
General:
Implementing Strategy; Managing Programmes; Managing
Projects; Success and Strategy; Processes, Procedures; Systems.
Project Oce; Audits, Health Checks; Systems Approach.
External Context:
PEST: Legal; Environmental; Value, Benet, Finance.
Implementation:
Functionality. Value; Conguration; Scope of Work; Organi-
zation Resources; Quality; Cost; Time; Risk; Safety and
Health.
Life cycle:
Integration-Life Cycle; Start-up; Proposal and Feasibility;
Design and Appraisal; Implementation; Progress; Commis-
sioning and Close-out.
Commercial:
Value and Benet; Finance; Cash Flow Management; Taxa-
tion; Insurance.
Contractual: Organization Design; Partnerships; Alliances;
Procurement; Bidding; Contract Administration; Materials,
Purchasing & Supply; Commercial Law; Claims; International
Projects.
People:
Management Structure; Teams; Individuals; Managing and
Leading; Stakeholders; Competence; Culture; Ethics; Change.
General Management:
Human resource management: Marketing; Operations; In-
formation technology; Finance and accounting; Strategy;
Technology, Innovation.
Transferring projects to their nal users: The eect of planning
and preparations for commissioning on project success
Dov Dvir
School of Management, Ben Gurion University, P.O.B. 653, Beer Sheva 84105, Israel
Received 20 April 2004; received in revised form 22 June 2004; accepted 1 December 2004
Abstract
This paper examines the relationship between planning and preparing the project for transfer to its nal users and project success.
Four planning and preparation aspects are considered (development of operational & maintenance requirements, customer partic-
ipation in the development process, developers preparations for turning over the project to its nal users, and nal user preparations
for introduction into operational use), along with three measures of project success (project eciency, customer benets, and overall
success). The study is based on data from 110 defense projects performed in Israel and includes regression and correlation analysis
between the two sets of variables. The ndings suggest that customer participation in the development process and nal user prep-
arations have the highest impact on project success. Customer participation in the development process is highly correlated with
project eciency (0.45), while nal user preparations are highly correlated with customer benets (0.46).
2005 Elsevier Ltd and IPMA. All rights reserved.
Keywords: Project planning; Project commission & close out; Project success
1. Introduction
There are four fundamentally dierent ways to close
out a project: extinction, addition, integration, and star-
vation [1]. Termination by extinction means the project
has been successful and achieved its goals: the new prod-
uct has been developed and handed over to the client; or
the building has been completed and accepted by the
purchaser. Projects terminated by extinction may have
been successful or unsuccessful. A project may be termi-
nated by institutionalizing it as a formal part of the
organization (addition) or by distributing the personnel,
equipment and functions among the existing elements of
the parent organization (integration). Projects which are
unsuccessful or obsolete, may be terminated by starva-
tion, or in other words, by cutting out the funds for its
completion. Starvation is usually used when manage-
ment is reluctant to admit that the project actually
failed.
Regardless of a successful project is completed by
inclusion, integration, or extinction; a plan must be
developed to terminate it. The process of project termi-
nation is not an easy task. It is to be planned, budgeted
and scheduled like any other phase of the project life cy-
cle. Sometimes a special termination manager, whose
primary responsibility is to eectively and eciently
complete the termination process, is appointed. The du-
ties of a termination manager may include the following:
Ensure the project is complete, ensure delivery and client
acceptance, prepare a nal report, redistribute person-
nel, materials, equipment, and any other resources, as-
sign responsibility for product support, if necessary [1].
Although the use of a termination manager for ensur-
ing that the project is complete and to deliver its out-
come (if successful) to its customers, is advocated by
several authors, and though project termination
0263-7863/$30.00 2005 Elsevier Ltd and IPMA. All rights reserved.
doi:10.1016/j.ijproman.2004.12.003
E-mail address: dvird@som.bgu.ac.il.
www.elsevier.com/locate/ijproman
International Journal of Project Management 23 (2005) 257265
PROJECT
MANAGEMENT
constitutes a signicant part in the total project, it is of-
ten overlooked by project managers [2].
This paper examines the relationship between plan-
ning and preparing the project termination and com-
mission and project success. Our objective is to
analyze the relationship between the amount of eort
invested in planning and preparing the project for
transfer to its nal users and the degree of success
achieved, as seen from dierent points of view. The
analysis is based on data collected from 110 defense
R&D projects performed in Israel and includes four
planning and preparing for transfer aspects (develop-
ment of operational & maintenance requirements, par-
ticipation of the customer in the development process,
developers preparations for turning over the project to
its nal user, and nal user preparations for receipt of
the project and starting its operational use), along with
three measures of project success (project eciency,
customer benets, and overall success). The paper is
organized as follows: we begin with a review of the
existing literature. Based on the review we propose four
hypotheses on the contribution of planning and prepar-
ing the project for transfer to its nal users to project
success. A description of the research methodology is
presented in the next section followed by presentation
of the data structure and the reliability of the various
constructs. The next section contains the analysis of
the correlations between planning and preparations
variables and success variables, and the regression re-
sults between the three success measures and the plan-
ning and preparing variables. We conclude with a
discussion of the ndings and their implications for
the practice of project management.
2. Theoretical background
The research body on project termination is rela-
tively small in comparison to other research areas of
project management such as project planning, control,
success measurement, and risk assessment. Buell [3] in
an early article claims that the main reason for so little
information on the subject is simply because it is hard
to spell out specic guidelines for termination of
projects.
Most research on project termination focused on
reasons for premature termination and not on the
introduction of the outcomes of successful projects into
use. Although, the decision to terminate a project may
be in certain situations more important than the deci-
sion to go on with the project, there is almost a unan-
imous agreement [1] that the termination stage of the
project rarely has much impact on technical success
or failure of the project. It has though, a great deal
to do with residual attitudes toward the project
the taste left in the mouth of the client, senior man-
agement, and the project team, which is important for
future projects, but of course have no impact on the
current one.
Among the studies on project termination we can
nd for example a study by Dean [4], who provides,
based on a small-scale survey, the frequencies of fac-
tors reported as reasons for termination of R&D pro-
jects. Balachandra and Raelin [5,6] performed a
discriminant analysis of variables aecting R&D pro-
jects termination. De et al. [7,8] did a detailed quanti-
tative work on taking abandonment decisions from a
nancial point of view at dierent contexts. Shafer
and Mantel [9] developed a decision support system
(DSS) for project termination. The DSS is able to ana-
lyze the sensitivity of various parameters of project ter-
mination, but the requirement for an extensive
database on projects of dierent types, limits its use
in practice. Archibald [10] prepared a check-list for
project termination. Stallworthy and Kharbanda [11]
categorized the problems involved in project termina-
tion into emotional and intellectual problems. Several
other studies on the same issue are those of Pinto
and Mantel [12], Green et al. [13], Broockho [14],
Black [15], and Chi et al. [16].
Another stream of research closely related to the re-
search on project premature termination is the research
on project critical success factors (CSF). The list of crit-
ical success factors is often used as a yard stick for
assessing the chances of a project to end successfully
when encountering problems. Pinto and Slevins work
[17,18] and other lists of critical success factors devel-
oped over the years, can be used for that purpose. When
several CSFs do not exist in a project, management may
consider terminating it in order to cut the potential
loses.
Only few researchers see project commission, when
the projects outcome is handed over to its customers
for use, as an integral part of the project life-cycle. That
is probably the reason for the lack of research on that
issue. The importance of the transfer phase to the suc-
cess of projects (not only the residual attitudes toward
the project), is indirectly evident from some of the stud-
ies on critical success factors of projects which have
identied the act of selling the project to its nal users
as one of the critical success factors [19,20]. Kleinsch-
midt [21], who studied the dierences in project manage-
ment practices between Europe and North America,
noticed that in Europe project managers more actively
encourage customer involvement in the project execu-
tion than in the US. Customer involvement is clearly
one of the most important ingredients that contribute
to an ecient and smooth transfer of the project out-
come to its users.
Hadjikhanis [22] perception that every project is an
episode in project marketing is one of a few exceptions.
The goal of marketing is dened as repetitive selling to
258 D. Dvir / International Journal of Project Management 23 (2005) 257265
the same buyer. Following Hyden [23] who dened pro-
jects life to begin when one rm draws up a contract
with another rm, and it is dissolved at the end of the
operation phase when the transaction is completed,
Hadjikhani focused his study on the management of
the relationship left after project completion and the
development and marketing activities after project sell-
ing. His hypothesis was that every project leaves sedi-
ment, and accordingly studied cases focused on the
phases before negotiation and after project completion.
This view is also shared by Faulkner and Anderson [24]
who claim that a project cannot be regarded as isolated
from former projects; projects are connected to each
other somehow.
In summary, there are many studies on project termi-
nation, but only few of them deal (implicitly) with the
impact of the termination activities on the project suc-
cess. The act of selling the project to its future users
and customer involvement in the project execution were
already identied as factors contributing to project suc-
cess. It is obvious that the main reason for customer
involvement is to increase the chances that the projects
outcome will meet the customer requirements and de-
sires. But why sell the project to the customer who al-
ready paid for it. The answer is probably that there are
other factors that help the customer to accept and start
using the new product. Furthermore, in our studies of
projects we have seen other activities that were meant
to improve the orderly transfer of projects to their nal
users and not only to establish successful and lasting
relations with the customer. Active involvement in the
development process and preparations made by the con-
tractor for turning over the project to its users are exam-
ples for such activities.
Some of these activities are aimed at improving the
eciency of the project execution, but most of them
are done to ensure that the customer will get the product
he needs and will start using it as soon as possible. Based
on the literature review and our own observations, we
propose the following hypotheses:
1. User involvement in the development of operational
& maintenance requirements is positively contribut-
ing to the overall success of the project and the cus-
tomer benets from its outcomes.
2. The involvement during the project execution of an
escorting team representing the customer, positively
contributes to the eciency of the project and to
the customer benets from its outcomes.
3. The developers support for turning over the project
to its nal user positively contributes to the customer
benets from the project.
4. The nal user preparations for receipt of the project
and starting its operational use positively contribute
to the customer benets from the project and to its
overall success.
3. Research methodology
Correlation analysis is often used when it is necessary
to relate a set of variables to other sets of variables.
When one set or both contain a large number of vari-
ables, it is dicult to interpret the results and to nd
causal relationships among the variables. To simplify
the analysis, factor analysis is very often employed to
represent the large data set by fewer, easy-to-interpret
factors. The current study follows that methodology.
First, we employ factor analysis to reduce the large
number of planning variables pertaining to the hand-
over of the project to its nal users. The resultant factors
are correlated with the three main success dimensions
(which were found in an earlier study to be more impor-
tant than the other dimensions [25]) and the results are
analyzed and discussed. Regression analysis is used to
enhance our understanding of the relationship between
the planning and preparing variables and success
dimensions.
3.1. Data
Data on 110 defense projects performed in Israel over
the last two decades were gathered using structured
questionnaires [26]. The questionnaires were lled out
by at least three key personnel related to the project
and representing the various stakeholders (the customer,
project management and procurement organization).
The projects in the sample were performed by a vari-
ety of rms in the areas of electronics, computers, aero-
space, mechanics and others. The respondent
population included many types of defense projects:
new weapon systems, communication, command and
control systems, electronic warfare equipment, and sup-
port equipment development projects. The sample in-
cludes almost all defense development projects
performed under the hospices of the Israeli Ministry of
Defense at that period. Since the processes and proce-
dures used in the Israeli defense industry are similar to
those used in the US, the sample is probably a good rep-
resentation of that population too, nevertheless, it might
not be a good representation of defense projects per-
formed in other countries.
The questions solicited subjective evaluations on a
seven-point scale. For example, the level of involve-
ment of the team escorting the development and inte-
gration of the project was determined by asking the
respondent the following question: According to your
assessment, how active was the escorting team in devel-
opment, integration and testing activities during the
project execution? The answer was given on the scale:
1 (not active at all) to 7 (extremely active). The ques-
tionnaires were administered in face to face sessions
by specially trained interviewers, all of whom had been
previously involved with this type of projects in various
D. Dvir / International Journal of Project Management 23 (2005) 257265 259
capacities. Most of the interviewers were retired ocers
from the Israeli Defense Forces with technical back-
ground and the others were MBA students in the man-
agement of technology track. For each project there
were three respondents: the project manager (or a se-
nior representative from the project oce); a represen-
tative from the customer or end-users community; and
a representative from the government sponsoring orga-
nization. Although an eort was made to locate the
most informed interviewees for each project, it did
not always succeed. In some cases, instead of the pro-
ject manager, only a team member could be located
and interviewed, in other cases the contracting ocer
was replaced with a new ocer whose knowledge on
the project was only second hand. To compensate for
the less reliable answers, the interviewer, who became
well acquainted with the project history, completed a
separate questionnaire integrating the three sets of re-
sponses, giving a lower weight to the less reliable re-
sponses. The analysis presented here is based on the
integrative questionnaire.
Out of the data collected by this questionnaire, rele-
vant data to project success and planning and prepara-
tion activities for the transfer of the project to its nal
users were used in this study to examine the relationship
between the development of the project operational and
logistic requirements, the eort invested in escorting the
development, the developer support for transferring the
project to its users and the user preparations for receiv-
ing the product and starting using it, and project
success.
3.2. Measures
3.2.1. Success criteria
Project success was measured along three criteria
(two constructs measuring success from two dierent
points of view and an overall success measure) that were
validated in previous research by Lipovetsky et al. [25].
1. Meeting planning goals (project eciency).
2. Customer benets (success from the customers point
of view).
3. Overall success (an integrative measure of project
success).
Table A.1 in Appendix A describes the responses to
the questionnaire items that addressed the two dimen-
sions of success used in this study.
3.2.2. Overall success measure
In addition to the two sets of success measures de-
scribed above, the questionnaire included an item
reecting the overall success of the project. Overall suc-
cess was also measured on a 1 to 7 scale, where 1 repre-
sents a complete failure and 7 represents outstanding
success. There were 108 responses, ranging from 1 to
7, with an average of 4.85 and a SD of 1.54.
3.2.3. Planning the Transfer of the project to its nal users
Twenty-two variables were used for estimating the
investment in planning and preparing for transfer of
the project to its nal user. Factor analysis with Vari-
max rotation on all items was carried out in order to val-
idate previous grouping of the variables into four factors
and examine their internal consistency. The results of
the factor analysis closely resemble the original grouping
of variables in the questionnaire and include only vari-
ables with loadings of 0.55 or higher (Table A.2 in the
Appendix A). The independence of the four factors is
demonstrated by the low loadings of the variables com-
prising each factor on the other factors. Cronbachs
coecient a was also calculated for each group. The
four resultant factors are listed below. A short title for
each dimension and Cronbachs coecients are in
parenthesis.
1. Development of operational & maintenance require-
ments (O&M requirements, a = 0.76).
2. Escorting the development process (escorting,
a = 0.82).
3. Developers support for turning over the project to its
nal user (developer support, a = 0.90).
4. Final user planning and preparing for introduction
into operational use (user preparations, a = 0.95).
While the names of factors 1, 3, and 4 are self-
explanatory, the term escorting the development
process, needs some explanation. In many projects,
customer participation is a common practice; usually
representatives of the nal user are taking part in the
project denition phase and in the nal testing. Here,
in addition to representatives of the nal users, techni-
cal experts knowledgeable in the specic technological
areas related to the project were an integral part of
the contractors development team. The main purpose
of building such a team is to gain in-depth knowledge
of the nal product to enable maintaining and even
improving the product by the user organization with-
out any external help.
The list of items comprising each factor and their
descriptive statistics are shown in Tables A.3A.6 in
Appendix A.
3.3. Data analysis and results
The central part of the data analysis consists of exam-
ining the correlations between the four composite mea-
sures of planning and preparing for transfer of the
project to its nal users with the three measures of pro-
ject success. The factors scores were computed as the
average scores of all measures comprising a specic fac-
tor. The results of the correlation analysis appear in Ta-
ble 1.
There are 21 correlation coecients, and it is possible
that some will appear to be statistically signicant due to
the compounded eect of Type I error. Consequently,
we have adjusted the critical signicance level to the
rather conservative value of 0.001. This means that the
compounded Type I error, i.e., the probability that
one of the 21 correlations will appear to be statistically
signicant while in fact it is not, is about 0.021, well be-
low the commonly applied threshold of 0.05.
Several interesting results emerge from the correla-
tions presented in Table 1. First, there is a high correla-
tion between escorting and O&M requirements. There is
also a high correlation between user preparations and
developer support. But, there is no correlation at all be-
tween the developer support and preparation of O&M
requirements and no signicant correlation with escort-
ing. On the other hand there is a correlation (although
not signicant at the 0.001 level) between user prepara-
tions and O&M requirements and escorting. Although
the four dimensions are almost independent (as dis-
cussed earlier), the correlations between some of the
dimensions is not surprising. For example, the developer
support is a complementary activity to the user prepara-
tions; it is not likely that the user (probably via the
escorting team) will make preparations for introduction
into use without the developers support. These activities
are dierent but performed with the same goal in mind.
Each of the three success measures is highly and sig-
nicantly correlated only with one of the planning and
preparations factors. The project eciency dimension
is correlated with escorting (0.45), customer benets
dimension is correlated with user preparations (0.46)
and overall success is correlated with O&M require-
ments (0.33). Furthermore, project eciency is also cor-
related, though to a lesser extent with the developer
support and user preparations (0.30 and 0.38, respec-
tively). Customer benets dimension and overall success
are positively, but not signicantly, correlated with all
other planning and preparation factors.
Finally, all three success measures (project eciency;
customer benets; and overall project success) are highly
inter-correlated, implying that projects perceived to be
successful are considered successful by all their major
stakeholders.
Table 2 presents the results of a step-wise regression
analysis between the three success dimensions and the
Table 1
Correlations between average planning and preparing for transfer to nal user scores and average success scores
Escorting Developer support User preparation Project eciency Customer benets Overall success
O&M requirements Corr. 0.37 0.08 0.26 0.22 0.22 0.33
Sig. 0.000 0.531 0.053 0.093 0.031 0.001
N 102 64 58 101 92 101
Escorting Corr. 0.19 0.30 0.45 0.27 0.28
Sig. 0.127 0.022 0.001 0.009 0.005
N 64 58 101 93 101
Developer support Corr. 0.52 0.30 0.23 0.17
Sig. 0.000 0.024 0.070 0.168
N 57 65 62 65
User preparations Corr. 0.38 0.46 0.28
Sig. 0.003 0.000 0.032
N 59 59 59
Project eciency Corr. 0.62 0.57
Sig. 0.000 0.000
N 94 105
Customer benets Corr. 0.71
Sig. 0.000
N 95
Bold font represents p 6 0.001.
Table 2
Multiple step-wise regression results (N = 56)
Success dimension Intercept point O&M requirements Escorting Developer support User preparations p Adjusted R
2
Project eciency 2.138 0.372 0.269 0.0003 0.239
Customer benets 3.714 0.447 0.0006 0.185
Overall success 4.033 0.272 0.0424 0.057
Bold font represents p < 0.05.
planning and preparations factors. Only signicant be-
tas are presented (p < 0.05).
User preparations factor is the only factor that comes
out signicant in all three regression analyses. The
escorting factor is also signicant in the regression anal-
ysis of the planning variables with project eciency. The
regression analysis of the overall success with the plan-
ning and preparation variables has very little explana-
tory power while the other two are explaining each
about 20% of the variability of project eciency and
the customer benets from the project.
Referring now to the hypothesized relationships be-
tween the planning and preparation variables and pro-
ject success, we can see that only hypothesis 4 is fully
supported by the study results while hypotheses 2, 3
and 4 were only partially supported.
User preparations are positively and signicantly cor-
related with all three success dimensions. This result is
also supported by the regression analysis (Table 2).
User involvement in development of O&M require-
ments is positively and signicantly correlated with the
overall success of the project as hypothesized and posi-
tively correlated but to a signicance level of only
0.031 with customers benets.
Escorting the development process by a dedicated
team positively and signicantly contributes to project
eciency as hypothesized while the contribution to the
customer benets and overall success, although positive,
is only signicant to the level of 0.009 and 0.005, respec-
tively. The positive relationship between escorting and
project eciency is also supported by the regression
analysis.
Finally, developer support is positively but not signif-
icantly correlated with project success. Counter to our
hypothesis, the highest correlation is with project e-
ciency (0.30) and not with customer benets or overall
success.
4. Discussion and conclusions
Planning is considered a central element of modern
project management. Commonly accepted professional
standards, such as the PMI Guide to the Project Man-
agement Body of Knowledge, recommend investing in
project management processes and procedures that
support planning in order to reduce uncertainty and
increase the likelihood of project success. Nevertheless,
planning of the termination phase, especially planning
for commissioning, has not received proper attention.
As we have seen in the literature review, most studies
on project termination focus on premature termination
and even those who see projects as somehow con-
nected to each other [22,24] focus on the management
of the relationship with the customers left after project
completion or on the development and marketing
activities after project selling and their eect on project
success.
Our study focuses on the planning and preparing the
introduction of the project into use. It shows that the ef-
fort invested in these activities directly aects project
success both from the eciency point of view and from
the customer benets perspective.
From the correlations table we can see that all plan-
ning and preparation eorts positively aect project suc-
cess, the escorting team and the nal user preparations
have the greatest impact. Project eciency is signi-
cantly aected by the involvement of an escorting team
and by the nal user preparations for introduction of the
project into use, while customer benets are mainly the
result of the nal user preparations. These ndings are
in line with Dvir et al. [27] conclusions that studied
the eect of planning at large on project success and rec-
ommended that end-user involvement should start at the
rst stage of the project execution and continue until its
successful end. Kleinschmidts [21] observation that in
Europe customers are more actively involved in the pro-
ject execution than in the US, represents probably an
intuitive understanding of the crucial role of customers
involvement in project success.
The current study goes one step further and helps to
understand specically how to improve the chances for a
successful commission. By establishing a team that will
escort the development and a team that will plan and
prepare the project commission, the nal user can di-
rectly improve the chances for success, his eorts have
a greater inuence on project success than the activities
done by the developer himself.
Furthermore, these nal user activities have a much
larger eect than the up-front formulation and deni-
tion of operational and maintenance requirements.
The main conclusion of this study is that projects per-
formed under contract for a specic customer, either an
external customer or an internal one, should devote con-
siderable eorts for planning and preparing in advance
the hand-over of the project to its nal users. Customer
involvement in all phases of the project can highly con-
tribute to the project success, especially to its ecient
execution.
The term we use in this paper defense projects is
somehow misleading and limiting the applicability of
the study results. Defense projects are usually associated
with weapon systems which their eectiveness is mainly
determined by their functional performance and to a les-
ser extent by the way they were introduced into service.
However, defense projects also include command and
control systems, electronic warfare systems, communica-
tions systems and logistic and support devices. These
types of systems are complicated, require continuous
maintenance, and it takes time to learn how to use them
eciently, and therefore they can benet from proper
planning and preparing their introduction into service.
Members of the contracting oce are ideal candidates
for the escorting team due to their technical skills. They
can contribute to the project denition, participate in
the development, integration and commissioning and as-
sist in future improvements of the product.
There are many other types of projects, not necessar-
ily defense projects that can benet from adopting this
approach. Projects performed within an organization
(i.e., production improvement projects, improving logis-
tic procedures, and IT projects) or government, munici-
pal and state contracted projects are examples.
This study is of an exploratory nature; the limited sam-
ple size, the homogeneity of projects types and the focus
onIsraeli defense projects, may limit its applicability. Fur-
ther research should be done in other countries and in dif-
ferent industries to study the termination and hand over
phase of projects in order to develop better ways for intro-
ducing projects into service and ensuring their nal users
satisfaction, whichis the ultimate proof of project success.
Appendix A
Descriptive statistics for the Project Success Items and
Factor loadings and descriptive statistics of the 22 Plan-
ning and Preparations variables (see Tables A.1A.6).
Table A.1
Descriptive statistics for the project success items
Success
dimensions
Success measures N Min. Max. Mean SD
Project
eciency
Meeting functional
requirements as
dened
103 1 7 5.82 1.23
Meeting
technological
specications
101 1 7 5.69 1.31
Meeting schedule 103 1 7 3.89 1.78
Meeting budget
goals
102 1 7 4.22 1.74
Meeting
procurement goals
82 1 7 4.62 2.30
Customer
benets
Satisfying customer
operational need
93 1 7 5.56 1.66
Project end-product
is in use
90 1 7 4.83 2.45
Systems delivered to
end user on time
83 1 7 4.24 2.16
System has
signicant usable life
expectancy
86 1 7 5.24 1.99
Performance level
superior to previous
release
75 1 7 6.08 1.47
End user capabilities
signicantly
improved
74 1 7 4.96 2.01
End user satised
from project end-
product
75 1 7 4.79 2.03
Table A.2
Factor loadings of the 22 planning and preparations variables
Variable Factor 1 Factor 2 Factor 3 Factor 4
Criteria for operational
eectiveness dened
0.24 0.37 0.62 0.22
User representatives
involved in
requirement denition
0.06 0.04 0.83 0.00
Combat mode of
operation dened
0.34 0.12 0.72 0.19
Logistic support
requirements dened
0.02 0.22 0.67 0.39
Human engineering
specications
0.13 0.31 0.59 0.17
Representative of the
nal user
active in the project
escorting team
0.25 0.55 0.05 0.08
Escorting team actively
involved in the
development process
0.09 0.79 0.03 0.05
Key escorting personnel
stayed for
the whole duration of
the project
0.11 0.79 0.19 0.06
Escorting team capable
to maintain
and improve the
operational system
0.11 0.70 0.32 0.01
Escorting team
committed to project
during operation
0.11 0.76 0.11 0.02
Prociency level of
escorting team
0.19 0.79 0.18 0.15
Developer prepared
detailed training
program
0.53 0.01 0.12 0.73
Quality of training by
the developer
0.51 0.03 0.01 0.74
Simulators were used for
operational training
0.01 0.23 0.13 0.57
Quality of training
documentation
0.29 0.28 0.34 0.67
Developers teams
supported
introduction
to operation
0.33 0.06 0.18 0.80
Developers teams
available in case of
problems at all times
0.45 0.15 0.04 0.61
Existence of professional
team to escort
introduction to
operational use
0.87 0.05 0.12 0.08
Systematic monitoring
of the introduction to
operation process
0.91 0.23 0.04 0.15
Adaptation of the
introduction process
due to lessons learned
0.93 0.06 0.02 0.19
Evaluation of the
introduction process
after completion
0.84 0.16 0.11 0.37
Formal introduction to
operation program
prepared by user
0.84 0.12 0.10 0.17
References
[1] Meredith J, Mantel Jr SJ. Project management: a managerial
approach. New York: Wiley; 2000. [p. 5413].
[2] De PK. Project termination practices in Indian industry: a
statistical review. Int J Proj Manage 2001;19:11926.
[3] Buell CK. When to terminate a research and development project.
Res Manage 1967(July).
[4] Dean BV. Evaluating, selecting and controlling R&D pro-
jects. New York: American Management Association; 1968.
[5] Balachandra R, Raelin AJ. How to decide when to abandon a
project. Res Manage 1980(July).
[6] Raelin JA, Balachandra R. R&D project termination in high-tech
industries. IEEE Trans Eng Manage 1985(February).
[7] De PK, Acharya D, Sahu KC. Capital investment with aban-
donment options and uncertainty in project life using multi-period
CAPM. Eur J Oper Res 1983;13:3618.
[8] De PK. Project evaluation with uncertain life and abandonment
options. Ind Eng J 1996;25:812.
[9] Shafer S, Mantel Jr SJ. A decision support system for the project
termination decision: a spreadsheet approach. Graduate Center
for the Management of Advanced Technology and Innovation,
Paper No. MATI-88W-001, College of Business Administration,
University of Cincinnati; 1988.
[10] Archibald RD. Managing high technology programs and pro-
jects. New York: Wiley; 1976.
[11] Stallworthy EA, Kharbanda OP. Total project management: from
concept to completion. Aldershot, England: Gower Press; 1983.
[12] Pinto JK, Mantel SJ. The causes of project failure. IEEE Trans
Eng Manage 1990;EM-37(4):26976.
[13] Green SG, Walsh MA, Dehler GE. Red ags at dawn: predicting
R&D project termination at start-up. Res Technol Manage
1993;1012(MayJune).
[14] Broockho K. R&D project termination decisions by discrimi-
nant analysis an international comparison. IEEE Trans Eng
Manage 1994;41:24554.
[15] Blake K. Causes of project failures: a survey of professional
engineers. PM Network 1996(November).
Table A.3
Descriptive statistics for the development of operational & maintenance requirements (O&M requirements) items
Measures N Min. Max. Mean SD
Criteria for operational eectiveness dened 99 1 7 4.75 1.77
User representatives involved in requirement denition 100 1 7 5.80 1.42
Combat mode of operation dened 95 1 7 4.38 1.82
Logistic support requirements dened 87 1 7 3.99 1.85
Human engineering specications 87 1 7 4.34 1.89
Table A.4
Descriptive statistics for the escorting the development process (escorting) items
Representative of the nal user active in the project escorting team 102 1 7 5.29 1.78
Escorting team actively involved in the development process 103 1 7 5.40 1.64
Key escorting personnel stayed for the whole duration of the project 101 1 7 4.86 1.67
Escorting team capable to maintain and improve the operational system 77 1 7 3.99 2.00
Escorting team committed to project during operation 77 1 7 4.88 1.76
Prociency level of escorting team 100 1 7 5.6 1.21
Table A.5
Descriptive statistics for the developers preparations for turning over the project to its nal user (developer support) items
Developer prepared detailed training program 64 1 7 5.23 1.80
Quality of training by the developer 62 1 7 5.03 1.72
Simulators were used for operational training 61 1 7 2.87 2.01
Quality of training documentation 60 1 7 4.43 1.78
Developers teams supported introduction to operation 60 1 7 4.72 1.83
Developers teams available in case of problems at all times 61 1 7 5.70 1.45
Table A.6
Descriptive statistics for the nal user planning for receipt of the project and starting its operational use (user preparations) items
Formal introduction to operation program prepared by user 58 1 7 5.12 1.80
Existence of professional team to escort introduction to operational use 59 1 7 4.95 1.80
Systematic monitoring of the introduction to operation process 59 1 7 5.08 1.70
Adaptation of the introduction process due to lessons learned 57 1 7 5.04 1.51
Evaluation of the introduction process after completion 55 1 7 4.60 1.77
[16] Chi T, Liu J, Chen H. Optimal stopping rule for a project with
uncertain completion time and partial salvage-ability. IEEE Trans
Eng Manage 1997(February).
[17] Pinto JK, Slevin DP. Project success: denitions and measurement
techniques. Proj Manage J 1988;19(3):6773.
[18] Pinto JK, Slevin DP. The causes of project failures. IEEE Trans
Eng Manage 1990;EM37(4):26976.
[19] Pinto JK, Slevin DP. Critical factors in successful project
implementation. IEEE Trans Eng Manage 1987;EM34:227.
[20] Tishler A, Dvir D, Shenhar A, Lipovetsky S. Identifying critical
success factors in defense development projects: a multivariate
analysis. Technol Forecast Social Change 1996;51(2):15172.
[21] Kleinschmidt EJ. A comparative analysis of new product
programmes: European versus North American companies. J
Prod Innovation Manage 1995;12(3):2545.
[22] Hadjikhani A. Project marketing and the management of
discontinuity. Int Bus Rev 1996;5(3):31936.
[23] Hyden EW. Technology transfer to East Europe. New
York: Praeger; 1976.
[24] Faulkner R, Andersson A. Short-term projects and emergent
careers: evidence from Hollywood. Am J Sci 1987;92(4):
879909.
[25] Lipovetsky S, Tishler A, Dvir D, Shenhar A. The relative
importance of defense projects success dimensions. R&D Manage
1997;27(2):97106.
[26] Shenhar AJ, Dvir D, Levy O. Mapping the dimensions of project
success. Proj Manage J 1997;28(2):513.
[27] Dvir D, Raz Z, Shenahar AJ. An empirical analysis of the
relationship between project planning and project success. Int J
Proj Manage 2003;21:8995.
The eect of systemic errors on optimal project buers
Dan Trietsch
*
ISOM Department, University of Auckland, Old Choral Hall, 7 Symonds Street, Auckland City 1001, New Zealand
Received 7 September 2004; received in revised form 22 October 2004; accepted 3 December 2004
Abstract
Existing mathematical models for setting buers for time or cost in project management assume that project activities are sta-
tistically independent. This leads to a highly counterintuitive and damaging conclusion that project buers should become relatively
negligible for projects with long chains of activities. We present a model that considers the statistical dependence between activities
caused by estimation bias. We show that if relatively high service levels are desired, this imposes a positive lower bound on the buer
as a data-based fraction of the estimated project duration or budget. We also introduce a new approach for collecting data and
estimating the parameters necessary to implement the model. This approach places a smaller burden on decision makers than tra-
ditional PERT: they provide single point estimates for means, while variance elements and bias correction are computed electron-
ically using historical data.
Keywords: Time; Cost; Operations
1. Introduction
Recent developments in the practice of project sched-
uling focus attention on the correct specication of
safety buers. Time buers are used to protect the sche-
dule, and, similarly, budget contingencies are used for
cost control. For example, to prevent wasting earliness,
a project buer is placed at the end of the project sche-
dule. Similarly, where chains of activities merge with
the critical path, feeding buers may be inserted. Note
that to specify such buers is mathematically equivalent
to deciding when to start each chain of activities. Yet the
size of these buers is usually specied as an arbitrary
fraction of the estimated chain duration. The purpose
of this paper is to provide a rst step towards optimizing
such buers based on plausible theory and relevant data
analysis.
In this paper we study the most basic case: the opti-
mal buering of projects with n (not necessarily indepen-
dent) activities in series, and without intermittent idling.
Thus there is only one chain of activities and it is neces-
sarily also the critical path. Assuming that either the due
date or the start time is a decision variable, the project
buer is the dierence between the due date and the ex-
pected completion time. Similar analysis is directly
applicable to the determination of a cost buer as well,
by the simple substitution of the due date by a project
budget. The main cost of a typical project is the sum
of the individual activity costs, although it also includes
overhead charges and holding costs that depend on the
project duration. Therefore, at least approximately, we
can say that cost is additive and not sensitive to the pro-
ject network structure. Similarly, due to the simplicity of
the serial structure, if we assume that earliness is uti-
lized, then the expected duration of the project is equal
to the sum of the expected durations of the activities.
Our result for the serial structure can serve to develop
insights for more realistic project networks, where such
doi:10.1016/j.ijproman.2004.12.004
*
Tel.: +649 373 7599x87381; fax: +649 373 7430.
E-mail address: d.trietsch@auckland.ac.nz.
PROJECT
MANAGEMENT
chains are often embedded. In such a case, the (serial)
project buer becomes a feeding buer for non-critical
chains feeding other project activities; e.g., feeding the
critical path. Such chains are sub-projects with a serial
structure. A useful approach for extending the results
we obtain here to more complex project networks is
by simulation, but this is beyond our scope.
Henceforth, we will discuss time buers, keeping in
mind that the results also apply to contingency budget-
ing. For this, we need to know the cumulative distribu-
tion function of the project completion time, which, in
turn, is a function of the individual activities distribu-
tions. In Malcolm et al. [1], the PERT developers dis-
cussed the need for stochastic estimation of activity
durations. Nominally, they proposed tting a beta dis-
tribution. Practically, they required a triplet of estimates
for each activity {min, mode, max} and then arbi-
trarily set standard deviation = (max min)/6 and
mean = (min + 4 mode + max)/6. Assuming indepen-
dence, they noted that the sum of a serial chain of activ-
ities is approximately normal by the central limit
theorem. Although they recommended it for tractability,
the team recognized that the use of the most likely crit-
ical path is optimistic: it ignores the statistical depen-
dence between paths that share activities. They already
had a computationally intensive remedy: Clark [2] pro-
vided approximate but unbiased calculations for the
project duration distribution. Finally, they highlighted
a major concern about systemic bias which is at the
heart of this paper and we will discuss it later. Armed
with 20/20 hindsight, one may criticize invoking the beta
distribution, considering that it was not really utilized,
but this is harmless. More seriously, it is dicult, per-
haps impossible, to obtain reliable triplets from process
owners. Anecdotal evidence to this eect, related to
work place politics, was given by Woolsey [3]. To this
we must add systemic human error as studied by Tver-
sky and Kahneman [4]; e.g., they observed that 98%
condence intervals estimated by experts tend to fail in
about one third of the cases (i.e., experts tend to grossly
underestimate the true variance).
Britney [5] presented optimal buers for stochastic
project activities on a one-by-one basis. Planned idling
occurs between activities unless the preceding buer is
exceeded. He did not, however, optimize the project as
a whole. Also, he did not consider the possibility of
combining serial buers. The combined buer approach
was used by Yano [6], who obtained the optimal single
buer for a project (or supply chain) composed of inde-
pendent serial activities. Yano [7] inserted individual
buers between the serial activities. Ronen and Trietsch
[8], Kumar [9], and Chu et al. [10], independently, opti-
mized the ordering times of n parallel project (or supply
chain) items with stochastic lead times, where the latest
one determines the completion time; i.e., they showed
how to coordinate n parallel feeding buers.
This idea using planned project- and feeding buers
to achieve improved reliability and avoid [implicit] pro-
ject delay penalties had recently been popularized un-
der the title Critical Chain, and many practitioners
nd it useful. Reviews and discussions of this modern
development abound; e.g., Herroelen and Leus [11],
Leach [12], Raz et al. [13], Trietsch [14]. The basic
assumption is that planned idling should not occur with-
in the critical path or within any chain of activities.
Nonetheless, it is incorporated where such chains merge
(yielding feeding buers). The size of protective buers is
determined as an arbitrary fraction of the expected
chain leading to the buer (e.g., 50%). Leach [12] sug-
gested the maximum of a buer based on the traditional
independence assumption and the arbitrary fraction.
Leach [15] listed 11 reasons why projects are typically
longer than expected. He then proposed a larger buer
based on the sum of the two elements mentioned before.
Although Clarks approximation can handle depen-
dent activities, none of the existing mathematical models
for buer setting accounts for systemic error. We intro-
duce an elementary model for this purpose and show
that it yields dependency between activities. We show
that if relatively high service levels are desired, this im-
poses a positive lower bound on the project buer as a
data-based fraction of the expected duration. In con-
trast, the independence assumption leads to a highly
counterintuitive and damaging conclusion that project
buers should become relatively negligible for projects
with long chains of activities. We also recommend an
approach for collecting data and estimating the param-
eters necessary to implement our model. This approach
places a smaller burden on users than traditional PERT:
single point estimates are required instead of triplets.
Variance and bias estimates can then be computed by
a decision support system that uses historical data,
and we show how.
The remainder of the paper is organized as follows.
After introducing the main notation, Section 2 presents
the [existing] basic buer optimization model subject to
a given project distribution. Section 3 introduces our
model with the bias-induced correlation. Section 4 ad-
dresses the estimation of the necessary model inputs.
Section 5 provides numerical illustrations. Section 6 is
the conclusion.
2. Notation and basics
B the bias of time or cost estimates of project
activities (a random variable)
b E(B), where E() is the expected value function
V
b
r
2
b
V(B), where V() is the variance function
C the cost to postpone the due date (per time
unit)
268 D. Trietsch / International Journal of Project Management 23 (2005) 267274
D the nominal project duration/due date (a deci-
sion variable)
P time unit cost during delay, including penalty
(where P > C)
n
j
the number of activities in project j
(j = 1, . . ., J). (Here and elsewhere, j is used
exclusively to index projects, and may be sup-
pressed)
SL service level, the probability of meeting or beat-
ing the due date
Y
ij
the duration of activity i in project j
(i = 1, . . ., n
j
; j = 1, . . ., J) a random variable
with the realization y
ij
Y
j
the set {Y
ij
} for all i in project j
Y
j
P
Y
ij
, the true project duration
F
j
(t) the cumulative distribution function of Y
j
l
ij
E(Y
ij
)
r
ij
the standard deviation of Y
ij
r
j
the standard deviation of the completion time
of project j
X
ij
a random variable that models Y
ij
(but not nec-
essarily correctly)
X
j
the set {X
ij
} for all i in project j
e
ij
the nominal estimate of l
ij
, assumed propor-
tional to E(X
ij
)
We assume that we have to determine a project due
date, D, or that one is given with enough slack to make
possible a delayed start. The time from the project start
to the due date is the planned duration. For conve-
nience, we will treat the start as time zero, so the
planned duration is equal to D. We also assume that
there is an economic cost per time unit, C, which pro-
vides an incentive to reduce the planned duration; e.g.,
the customer is more likely to award us the contract if
we quote an early (but reliable) due date. If the customer
is not likely to pay before the due date when the project
is early, C should also include any xed charge the pro-
ject incurs while on the books. Any delay beyond D in-
curs a cost (including penalty) of P(Y D)
+
, where
Y =
P
Y
i
and h
+
= max{0, h}. P > C, or it would be
optimal to set D = 0 and pay P as long as necessary.
The objective is to minimize the expected total cost
(TC),
ETC CD EPY D
(By dening P
0
= P C and subtracting CY from both
sides of the equation a value which is not a function
of D we can show that minimizing this objective func-
tion is equivalent to minimizing
ECD Y
P
0
Y D
:
Thus both earliness and tardiness have economic losses
associated with them.)
Denote the optimal probability of meeting or beating
the due date, i.e., the service level, by SL* (we will use
asterisks to denote optimality in general). For continu-
ous distributions, a straightforward application of the
newsboy model shows that SL* = (P C)/P. Thus,
D* = arg{SL* = F(D*) = (P C)/P}, where F(t) is the
project duration distribution. For discrete due dates,
where it is possible to hit the due date exactly and thus
be neither early nor late, we must set D such that the
probability of delay will not exceed C/P and the proba-
bility of earliness will not exceed (P C)/P. Either way,
we need F(t). The technical focus of this paper is obtain-
ing a correct duration distribution (or expenditure distri-
bution) based on the estimates we have for each activity,
but subject to systemic error.
3. Modeling positive dependence due to systemic error
It is well known that project activities are sometimes
statistically dependent. Nonetheless, with one exception,
the literature on setting buers assumes statistical inde-
pendence. This includes academic papers and practi-
tioner books. But if we increase the number of
activities along the critical path, the independence
assumption leads to so-called optimal project buers
that, as a fraction of the mean, approach zero asymptot-
ically. Because this is a highly counterintuitive conclu-
sion, the practical approach has always been to set the
buer, or the contingency, as a fraction of the antici-
pated duration; e.g., see OBrien [16]. In addition to
leading to such a highly counterintuitive conclusion,
and to ying in the face of experience, the independence
assumption should be challenged on theoretical grounds
as well. Leach [15], after citing empirical evidence dem-
onstrating that the independence assumption is not valid
in practice, identies positive bias as the culprit. He then
provides 11 causes of positive bias mostly related to
large projects with multiple paths. Our focus here is
on a single cause of positive or negative bias that applies
even to serial projects and that, arguably, explains the
bulk of the problem. The crux is that even when activi-
ties are independent by nature, their estimated durations
are often subject to the same estimation error. Thus,
when the same optimist estimates many activities, they
will all tend to be underestimated. Equivalently, pres-
sure by management may inuence people to give too
short estimates that cannot be met later. Similarly, bad
weather during the project or the loss of a key employee
may impede several activities. A booming economy may
increase queueing time for more than one activity, etc.
The same or similar causes also apply in the reverse
direction; e.g., pessimists will tend to overestimate dura-
tions (and severe punishment for missing due dates in
the past will turn optimists to pessimists). The result is
a systemic bias across a project. But the magnitude of
this bias is a random variable and even its orientation
is not known in advance. This introduces a strong
D. Trietsch / International Journal of Project Management 23 (2005) 267274 269
dependence as far as deviations from plan are concerned.
But, operationally, the only variation that concerns us is
relative to plan!
Notably, the need to account for estimation bias by
some calibrating program has been on the agenda from
the earliest PERT days. Malcolm et al. [1] stated that the
problem had been raised by many. They then suggested
a program of comparing estimates with actual perfor-
mance over a period of time to permit calibration of
the estimators. Alas there is no evidence that such cal-
ibration was ever implemented on a wide scale, if at all.
Instead, subsequent papers about bias focused mostly
on the optimistic bias due to ignoring near-critical
parallel paths that may become critical in reality; e.g.,
Klingel [17], Schonberger [18]. Thus, the clear recom-
mendation that the PERT team expressed in [1] was
ignored, while an issue that they considered more minor
and for which [2] had already provided an approxi-
mate remedy was highlighted.
Furthermore, [1] stated objective with respect to bias
was to calibrate the estimates to make them accurate.
But there is also a statistical dependence issue that
arises. For example, suppose we suspect that there is a
random bias error of 25%, and we nd that the rst se-
ven activities took 85% of their combined estimated
time, then we would probably consider it likely that
the next activities will also tend to consume around
85% of their estimate in other words, we form an infor-
mal estimate of 85% for the necessary calibration. This
means that we perceive a correlation between the rst
few activities and the ones that are yet to follow. This,
by denition, is a case of statistical dependence (because
independent variables are not correlated). In practice,
we need to know not only the average bias but also its
impact on the covariance of activities. Our purpose here
is to address both the need for calibration and the mag-
nitude of the correlation that is involved, so we can draw
conclusions for the optimal combined buer (including
elements for bias correction and for safety).
Let the true activity times compose the random vec-
tor Y = {Y
i
}. If decision makers would have the distri-
bution of Y, there would be no need for this paper. In
reality, however, decision-makers never know Y, so they
must use some estimate that acts as a model of Y.
Accordingly, let X = {X
i
} be a vector of the decision-
makers models of the actual activity durations, {Y
i
}.
So X
i
is a random variable that represents another ran-
dom variable. Estimates are based on X, and not directly
on Y. Systemic error arises because X is not a perfect
model of Y. Mathematically, perhaps the simplest possi-
ble model for systemic error involves the introduction of
an additional independent random variable, B, that
multiplies X to obtain Y. In this paper, we limit our-
selves to this basic model. Other potential approaches
exist, however, and further research may be justied to
identify the best one. We will use b and V
b
r
2
b
to de-
note the mean and variance of B. We assume that e
i
,
the nominal (single point) estimate of Y
i
, is proportional
to E(X
i
). For example, if a particular provider of esti-
mates includes a hidden 100% buer in her estimates,
then e
i
= 2E(X
i
). Nonetheless, for simplicity (and with-
out loss of generality), we will set e
i
= E(X
i
). As long
as the assumption that e
i
is proportional to E(X
i
) holds,
there always exists a B that corrects any deviation intro-
duced by this simplication.
In broad terms, the randomness of X
i
relates to
chance events that are specic to Y
i
(as perceived),
e.g., problems with raw materials or power supply. X
also includes some (but not all the) eects of random-
ness in estimation, since estimates are the result of pro-
cesses that are not deterministic. Even with the same
data, dierent people at dierent times typically come
up with dierent estimates. For example, they may for-
get or neglect dierent aspects of the job. Random esti-
mation errors are operationally equivalent to random
deviation from plan. Because perceived activity-specic
chance events and some estimation errors are con-
founded with each other, they must be represented by
the same random variable, X
i
. In contrast, B models ef-
fects that are common to all activities, such as pressure
to produce attractive estimates quickly (leading to opti-
mistic bias and omissions), personal safety buers,
weather, economic conditions, etc. Note that B is impor-
tant even if b = 1; i.e., if we take steps to remove bias on
average, as suggested by [1], and thus achieve accuracy,
B would still capture important information about sys-
temic imprecision. Specically, those estimation errors
that are not confounded with activity-specic chance
events.
To present the most basic model, we assume that the
elements of X are independent of each other. This
assumption is often realistic, by which we mean that
decision makers typically estimate the elements of Y
(by the elements of X) as if they are independent and
therefore the elements of X are indeed independent
after all, X is a model only. (Modeling dependence into
X may be useful for some purposes, but requires further
research.) Because B and X are independent of each
other, l
i
= b e
i
, but the multiplication by the same real-
ization of B introduces [positive] dependence between
the elements of Y even if the elements of X are indepen-
dent. Specically, r
2
i
b
2
V X
i
V X
i
V
b
V
b
e
2
i
,
and COV(Y
i
, Y
k
) = V
b
e
i
e
k
; "i 6 k. We can separate
r
2
i
to two parts, b
2
V(X
i
) + V(X
i
) V
b
and V
b
e
2
i
. The
former equals E(B
2
) V(X
i
) and the latter is a special
case of V
b
e
i
e
k
. Thus the covariance matrix of Y is
the sum of a diagonal matrix with elements V(X
i
) E(B
2
)
and a full symmetric matrix with elements V
b
e
i
e
k
;
"i, k (i.e., the vector product {r
b
e
i
}{r
b
e
i
}
T
). To cor-
rect for the average bias we add a bias correction of
(b 1)
P
e
i
to the nominal makespan estimate
P
e
i
. Thus
we obtain a relative bias correction of b 1. If B = 1, we
obtain the classic model with independence. Similarly, if
V(B) = 0 but b 6 1, then after the bias correction we
again obtain the classic model. Therefore, our model
generalizes the classic approach. Finally, monitoring a
project over time during its execution always involves
estimating the bias that applies to it, either explicitly
or implicitly.
Bias correction is the rst response to bias, but the
standard deviation of the project completion time is vital
for rational determination of the safety buer. Since the
elements of X are independent, we have, r
2
=
E(B
2
)
P
"i
V(X
i
) + (r
b

P
"i
e
i
)
2
. To see this, let V de-
note the covariance matrix of Y, and let 1 be a column
vector in R
n
with elements of 1, then the result is
obtained directly by the matrix product 1
T
V1. To study
the eect of the standard deviation on the optimal
buer further, let q
1
(e) = E(B
2
)
P
"i
V(X
i
), and let
q
2
(e) = (r
b

P
"i
e
i
)
2
, where e={e
i
}, then
maxf
q
1
e
p
;
q
2
e
p
g 6

q
1
e q
2
e
p
6

q
1
e
p

q
2
e
p
:
The central element in the inequality is r.
p
q
2
is propor-
tional to the nominal makespan estimate (before the
bias correction),
P
e
i
. So there exists a fraction of this
estimate, namely r
b
, that acts as a lower bound on r.
Therefore, if we wish to specify a safety buer against
random variation of kr for some k > 0, this safety buer
is bounded from below by k r
b

P
"i
e
i
, which consti-
tutes a fraction of k r
b
of the nominal makespan esti-
mate. Furthermore, when n ! 1 then q
1
/q
2
! 0 so the
same bound serves as the approximate optimal safety
buer. Recall that Leach [12] and Leach [15] suggest rel-
ative buers that are associated with Max{
p
q
1
,
p
q
2
}
and
p
q
1
+
p
q
2
, respectively, so for a positive k these
values provide a lower- and an upper bound for the cor-
rect result. (There is no theoretical reason to limit our-
selves to k > 0, so we do not limit our analysis to this
case. Nonetheless, most project managers are uncom-
fortable with negative buers and the low service levels
they entail.)
4. Estimating model parameters
A tempting approach is to estimate X
i
by the classical
3-point method of PERT, thus yielding e
i
and V(X
i
). It
would then only remain to estimate B. However, there
are major diculties with this approach (as discussed
in Section 1), even without the new requirement to dis-
tinguish consciously between systemic bias and individ-
ual activity variation causes. Therefore we propose to
limit the information elicited from process owners to
single point activity estimates, e
i
, and obtain all the
other necessary estimates from historical data with the
help of a computerized decision support system (DSS).
An ideal model-based DSS takes simple inputs and con-
verts them to simple-to-use outputs. The conversion it-
self, however, has to be programmable and ecient
but not necessarily simple. With this in mind, we need
a programmable model to estimate b, r
b
, and V(X
i
),
and then compute r
2
j
. This requires a two-stage econo-
metric model, however, because V(X
i
) cannot be esti-
mated until we estimate b.
The historical data we need consists of pairs of activ-
ity estimates and realizations. Even though our formal
model assumes a serial structure, the network structure
of the historical projects is immaterial. We treat activity
estimates as explaining variables and realizations as
dependent variables. Assume we have J > 1 projects in
the history, and in this section we use the double index,
i, j where i = 1, 2, . . ., n
j
and j = 1, 2, . . ., J. Here i de-
notes a specic activity and j a specic project with n
j
activities. Optionally, we may elect to treat some sub-
projects as projects in their own right for this purpose.
This is reasonable sometimes, e.g., if the subprojects
were managed (and thus also estimated) relatively inde-
pendently or involved distinct sets of resources. For pro-
ject j our data consists of n
j
pairs (y
ij
, e
ij
), where y
ij
is the
realization and e
ij
the original activity estimate. One
possible estimator of the mean systemic error of project
j is given by
^
b
j

P
nj
i1
y
ij
P
nj
i1
e
ij
; 8j
leading to
^
b
P
j
j1
n
j
^
b
j
P
j
j1
n
j
; ^ r
2
b

P
j
j1
n
j
^
b
j

^
b
2
P
j
j1
n
j
J
:
This completes the rst stage (and leaves
P
n
j
J de-
grees of freedom for the second stage).
Our decision to only employ single point activity esti-
mates implies that V(X
ij
) must be modeled as a function
of the activity estimate e
ij
itself. The parameters of this
function can only be estimated from historical data.
Suppose we start with a quadratic approximation of
the form V X
ij
a
0
a
1
e
ij
a
2
e
2
ij
. We rst note
that, because V(X
ij
) P0, a
k
P0 is required. Further-
more, in the limit, as e
ij
! 0 we expect to obtain
r
2
ij
! 0 as well, so a
0
= 0 is indicated. Accordingly, we
obtain the model V X
ij
a
1
e
ij
a
2
e
2
ij
. Let CV
ij
denote
the coecient of variation of X
ij
, and recall that
E(X
ij
) = e
ij
, then CV
ij
P
p
a
2
("i, j), with equality
i a
1
= 0. So if CV
ij
is approximately constant (for all
i, j), our model will t with a
2
dominating and a
1
may
be insignicant. But if long activities often include sev-
eral independent sub-activities, a
1
will be signicant.
(To see this consider that if a
2
= 0 and we combine any
number of activities with the same positive a
1
to one
composite activity, its variance will be a
1
times its com-
bined mean. But this is exactly what we would expect if
these activities were independent.) Let q
ij
denote the
residual of x
ij
, then
q
ij

y
ij
b
j
e
ij
; ^ q
ij

y
ij
^
b
j
e
ij
and note that E(q
ij
) = 0. We can now regress ^ q
2
ij
by e
ij
and e
2
ij
to obtain estimates for a
1
and a
2
. Since both must
be non-negative, if either estimate is negative then to
minimize the mean squared error subject to our con-
straints (a
i
P0) we must set the oender to zero and re-
gress for the other value.
Over time, as more projects enter the data base, the
estimates of b, r
b
, a
1
, and a
2
should be updated regu-
larly. We need an adaptive, self-calibrating, approach
because the very use of a model such as ours is likely
to inuence the systemic error distribution over time.
This can be based on exponential smoothing. One
advantage of including such an adaptation mechanism
is that it is possible to start the system with relatively
inaccurate data and over time it will calibrate itself.
Once in place, the data collection burden this system
places on participants is very low. The estimates they are
required to provide are limited to the bare minimum
single point activity estimates. As for data to be col-
lected during ongoing projects, since the estimates are
already in the data base, it is enough to record the actual
performance of each activity. One question remains:
how to obtain initial values for such a scheme. This
should not be too dicult for most project organiza-
tions, since the data required is very elementary. How-
ever, if such data is not available, we propose to use
generic data from the relevant industry, or even resort
to guesstimates. This is permissible because the system
will calibrate itself eventually.
Finally, it may be reasonable to assume that B is log-
normal (i.e., that it reects the multiplicative eects of
many small independent biases) while re
i
tends to be
approximately normal. The product of a lognormal ran-
dom variable and a normal one is neither normal nor log-
normal, but to an extent that depends on the variance
elements its shape is in between the symmetric normal
and the skewed lognormal. Thus we may expect a skewed
distribution, but not very skewed if r
b
/b is small. Leach
[15] cites practical experience that deviations from plan
are indeed skewed, as the model predicts. Rough analysis
of data from his Fig. 2 [15, p. 37] suggests that time esti-
mates in a particular company had b 1.8 and
r
b
0.55. Improvements reduced these to b 1.06 and
r
b
0.3. This analysis should be treated with caution,
however, because it is based on second-hand data that
was originally reported for another purpose.
5. Numerical illustrations
For simplicity, in spite of the recommendation to use
a lognormal distribution for B, in the following numer-
ical examples we use an additive normal approximation
for the bias eect. Here, the dierence is not signicant
(because B does not possess a very high coecient of
variation).
5.1. Example 1
Suppose that by analyzing past data we obtain esti-
mates of 1.25 and 0.15 for b and r
b
(if we would choose
to model B by a lognormal distribution its parameters
would be 0.2160 and 0.1196). This applies to any
new project. For a particular project, suppose that our
current estimate for a chain of activities is 28 weeks.
The variance contribution due to r
b
is (0.15 28)
2
=
4.2
2
= 17.64. Therefore, the standard deviation of the
chain is now bounded from below by 4.2 (even when
we ignore all other variances). The bias correction re-
quired is (1.25 1)28, i.e., we should add 7 weeks to
the estimate.
Now suppose that
P
V(X
i
) = 19.785. After multiply-
ing by E(B
2
) = 1.585, we obtain an updated contribution
of 31.36. Taking dependence into account implies add-
ing 4.2
2
to this value, yielding a total variance of about
49 (7
2
). If we require a project service level of, say, 95%,
then, assuming a normal distribution, we need a safety
buer of 1.645r, or 11.52, in addition to the bias correc-
tion. So our due date should be set to 28 + 7 + 11.52 =
46.52 (of course, in practice such a result would
be rounded). The bias correction plus the safety time
add up to 18.52, which is a large increase. But it is
justied because our history suggests optimism and
sizable dependence, and also because 95% is a high
service level.
5.2. Example 2
For the same B as in Example 1. Suppose X com-
prises n iid activities, each with a normal distribution
with e
i
= 4, V(X
i
) = 0.794
2
(so E(B
2
) V(X
i
) = 1.585
0.794
2
= 1 for all i). Let P = 4C. Determine the optimal
project due date for n = 1, n = 5, and n = 25. Also, deter-
mine the asymptotic behavior of the relative optimal to-
tal buer dened as b 1 + k r/
P
e
i
, i.e., the relative
bias correction combined with the relative safety buer
as a function of n.
5.3. Solution
P = 4C leads to SL* = 0.75. Using the normal approx-
imation this implies a safety buer of 0.6745r. In all cases,
we need bias correction elements of n (0.25 4 = 1); i.e.,
the bias corrections are 1, 5, and 25, respectively. For
n = 1, our variance is 1 + 4
2
0.15
2
= 1.36 = 1.166
2
lead-
ing toa total buer of 1.787 (compare to0.6745 that might
be specied without the bias correction). This is 44.7% of
the estimated mean. The due date is 5.787. For n = 5, the
variance is 5 + 20
2
0.15
2
= 14, leading to a due date of
27.524, and the total buer is 37.6% of the estimated
mean. n = 25 leads to a due date of 135.66, and the total
buer is 35.7% of the estimated mean. Asymptotically,
we need a bias correction of 25% and an additional
15 0.6745 = 10.1% of the estimated mean for safety,
thus imposing a lower bound of 35.1% on the total size
of the buer as a fraction of the estimated mean. So most
of the protectionfor n = 25 is due toexpansionanddepen-
dence, while a sizable part of the protection for n = 1 is for
processing time variation. Fig. 1 depicts the optimal due
date and the total buer for n = 1 to n = 10. We see that
when expressed in percents the buer approaches the limit
asymptotically. We also see that the optimal buer as a
function of n follows an approximately straight line, but
it has a positive intercept. Thus, short chains require rela-
tively more protection.
6. Conclusion
The simplest non-trivial project structure is a serial
chainof activities bueredby a single project buer. With-
out correct analysis of such chains it is impossible to ana-
lyze more realistic structures. For project buers to
provide adequate protection, we must take into account
positive dependence between activities due to systemic
estimation errors and causes that are shared by many
activities. We developed a model for this which yields pro-
ject buers that tend to grow approximately linearly with
the expected estimated duration i.e., they do not become
negligible as the independence assumption implies. We
also presented a plausible estimation model that only re-
quires single point estimates andderives bias andvariance
elements from historical data, which is updated as new
projects are completed. When necessary, this process
can initialized by using generic industry data.
In the analysis of general project networks, the rst
step is often to reduce every embedded simple chain
by convolution, as if independence applies. Our model
suggests that this is inadequate. The next stage of this re-
search should be the study of such networks. One way to
n - number of activities
Buffer (%) Minimal buffer (%) Optimal due date
Corrected mean Estimate (days) Buffer (days)
0
10
20
30
40
50
60
1 2 3 4 5 6 7 8 9 10
P
e
r
c
e
n
t
s

o
r

d
a
y
s
Fig. 1. The optimal protective time cushion for n = 1 to n = 10 in Example 2.
do this is by simulation, but incorporating the insights
we provided.
References
[1] Malcolm DG, Rosebloom JH, Clark CE, Fazar W. Application
of a technique for a research and development program evalu-
ation. Oper Res 1959;7(5):64669.
[2] Clark CE. The greatest of a nite set of random variables. Oper
Res 1961;9:14562.
[3] Woolsey RE. The fth column: the PERT that never was or data
collection as an optimizer. Interfaces 1992;22(3):1124.
[4] Tversky A, Kahneman D. Judgement under uncertainty: heuris-
tics and biases. Science 1974;185:112431.
[5] Britney RR. Bayesian point estimation and the PERT scheduling
of stochastic activities. Manage Sci 1976;22(9):93848.
[6] Yano CA. Setting planned leadtimes in serial production systems
with tardiness costs. Manage Sci 1987;33(1):95106.
[7] Yano CA. Planned leadtimes for serial production systems. IIE
Trans 1987;19(3):3007.
[8] Ronen B, Trietsch D. A decision support system for purchasing
management of large projects. Oper Res 1988;36(6):88290.
[9] Kumar A. Component inventory costs in an assembly problem
with uncertain supplier lead-times. IIE Trans 1989;21(2):11221.
[10] Chu C, Proth J-M, Xie X. Supply management in assembly
systems. Nav Res Log 1993;40:93349.
[11] Herroelen W, Leus R. On the merits and pitfalls of critical chain
scheduling. J Oper Manage 2001;19:55977.
[12] Leach LP. Critical chain project management. Artech House;
2000.
[13] Raz T, Barnes R, Dvir D. A critical look at Critical Chain project
management. Project Manage J 2003;34(4):2432.
[14] Trietsch D. Why a critical path by any other name would smell
less sweet: towards a holistic approach to PERT/CPM. Project
Manage J [in press].
[15] Leach L. Schedule and cost buer sizing: How to account for the
bias between project performance and your model. Project
Manage J 2003:3447.
[16] OBrien JJ. CPM in construction management: Scheduling by the
critical path method. McGraw-Hill; 1965.
[17] Klingel AR. Bias in PERT project completion time calculations
for a real network. Manage Sci (Series B) 1966;13(4):
B-194201.
[18] Schonberger RJ. Why projects are always late: a rationale based
on manual simulation of a PERT/CPM network. Interfaces
1981;11(5):6670.
Simulation-based estimation for correlated cost elements
I.-T. Yang
Department of Civil Engineering, Tamkang University, No. 151 Ying-chuan Road, Tamsui, Taipei 251, Taiwan
Received 12 August 2004; received in revised form 10 November 2004; accepted 1 December 2004
Abstract
In this paper, we present a general method to incorporate correlations between cost elements in the process of cost estimation.
The proposed method rst checks the feasibility of the correlation (Pearson or Spearman) matrix, adjusts it if necessary, then uses
the correlations to generate correlated multivariate random vectors, which are employed to model possible outcomes of the cost
elements. The method is applied to a full data set of 216 British oce buildings to illustrate its practical use. The application result
indicates that the impact of correlations is signicant and may cause serious problems if neglected. The result is also used to validate
that the proposed method can capture the correlations with relatively small deviations.
Keywords: Cost; Estimation; Risk; Simulation; Project management
1. Introduction
Cost estimation is a process of collecting, analyzing,
and summarizing data in order to prepare an educated
projection of the anticipated cost of a project [26]. This
process begins inthe early stages of the project andrepeats
frequently during the entire life cycle. The reliability of
cost estimation is important to ensure the success of the
project since it serves as the foundationfor making critical
decisions on nancial investments and borrowings.
The prices of all the resources (material, equipment,
and labor) are exposed to certain levels of uncertainty,
particularly when the project life cycle is lengthy [25].
To manage the inherent uncertainty, Monte-Carlo sim-
ulation methods have been widely applied for various
types of projects, such as [3,11,15,31]. In Monte-Carlo
simulation, a mathematical model is constructed based
on pre-specied probability distributions, which de-
scribes the possible outcomes of major cost elements
(e.g., substructure, exterior walls, and electrical nish-
ing) involved in a project, and run to see what the over-
all project cost will be for each simulation replication.
After a certain number of replications, the collected
samples are used to derive the output distribution of
the overall project cost.
An enhancement of ordinary simulation methods has
been directed to consider statistical correlations (depen-
dencies) between cost elements. The correlation repre-
sents the co-movement of two cost elements; when one
is more expensive, the other tends to cost more as well
(or cost less for a negative correlation). Arguments and
evidences for the existence of correlations and their pro-
found impact on simulation results have been presented
in the literature [10,23,33]. To treat the correlations, var-
ious approaches have been proposed, such as [29,30,34].
The goal of this paper is to present a simulation-
based method to incorporate correlations between cost
elements with more modeling capabilities. Specically,
the present method is unique in handling the following
requirements:
1. To allow the distributions (i.e., marginal distribu-
tions) of individual cost elements to be of dierent
types. Namely, some of them may only be expressed
doi:10.1016/j.ijproman.2004.12.002
E-mail address: ityang@mail.tku.edu.tw
PROJECT
MANAGEMENT
with discrete and nite options whereas others can be
expressed as continuous functions. In addition, those
continuous distributions may come from dierent
families (e.g., some are lognormal while some are
beta).
2. To provide an automatic procedure to check the
feasibility (a mathematical denition will be given
later) of a correlation matrix and adjust it if
infeasible.
This paper is organized as follows. In the following
two sections, we introduce two dierent approaches
(historical and subjective) to obtain necessary input
for the analysis: the marginal distributions of individ-
ual cost elements and their correlations. The charac-
teristics of the two approaches lead to the need for
a more general method. In the section followed, we
present the proposed method in two stages and pro-
vide computational guidelines. The proposed method
is then applied to a modied British data set com-
posed of 216 oce buildings to illustrate its practical
use. The results are used to validate the proposed
method. Closing remarks are presented in the last
section.
2. Required input: distributions and correlations
There are two sets of input data required to perform a
simulation-based cost analysis considering correlations.
The rst set describes marginal distributions of individ-
ual cost elements and the second is a correlation matrix
consisting of the correlation coecients between pairs of
cost elements. Both sets of data can be estimated in two
ways: (1) by summary statistics on historical data, or (2)
by subjective judgments. In what follows we give brief
introduction to these two approaches and elucidate
how their characteristics give rise to a more general sim-
ulation method.
2.1. Choice of distribution
When historical data is used to describe the marginal
distributions, it involves an attempt to t theoretical dis-
tributions to the data and verify goodness-of-t statisti-
cally. While a complete review of goodness-of-t tests
and their theoretical backgrounds is given in [18], the t-
ting process can be done very eciently by commercial
software packages (such as Palisade BestFit, Arena,
ProModel, and Crystal Ball). A typical result is a list
of several good distributions and their associated
parameters, based on which the estimator can select
the most proper one. Previous studies suggested that
the lognormal distribution ts historical cost data better
than other well-known distributions, such as normal or
beta [30,33].
Despite its theoretical maturity, using historical data
to forecast possible outcomes has some pitfalls. First,
actual values may lie outside the range of historical re-
cords due to new technology, equipment, and material.
Second, historical data may not adequately represent
the true underlying population because of sampling er-
ror [12]. Last, the prices of resources may not always
be repeatable, thus, the historical approach may be
fallacious [23].
In the absence of reliable historical data, the sec-
ond best alternative is for a cost estimator to rely
on his/her experiences to subjectively specify the mar-
ginal distributions. In the context of probabilistic esti-
mation, it is usual to assume the underlying
distribution is a beta distribution whose parameters
are specied by three point estimates: the minimum,
maximum, and most likely values [4]. Some recent
controversy has been whether the beta distribution
should be replaced with the triangular distribution
since the former does not have clear-cutting bounds
[16] and requires four parameters, which do not have
a one-to-one correspondence with the three estimates
[32].
Both approaches mentioned above (historical and
subjective) may involve the following practical concerns.
First, they may actually be mixed in practice. A cost esti-
mator may not have historical data of all the cost ele-
ments if some works are usually outsourced or
subcontracted. Thus, he/she can obtain summary statis-
tics only on those data on hand and has to rely upon
subjective estimation for the remaining cost elements.
Another possibility occurs when the estimator has rea-
sons to believe some of the cost elements in a new pro-
ject have their own bounds (minimum and maximum)
and thereby cannot be represented by historical data.
For these particular elements, the subjective approach
is more appropriate. Second, it has been frequently
encountered that the price options of some cost elements
are collections of discrete outcomes in lieu of continuous
functions [2].
The practical concerns above give rise to the need for
a more general simulation model, which should be able
to treat all dierent types of distributions in one frame-
work (some are discrete while some are continuous;
some are lognormal while some are beta). This is the
rst goal of this paper.
2.2. Specication of correlations: product-moment and
rank
Two dierent measures have been used to reect the
degree of relation between cost elements. The rst one
is an ordinary product-moment (Pearson) correlation
coecient and the second is a rank (Spearman) correla-
tion coecient. The product-moment correlation can be
calculated as follows:
276 I.-T. Yang / International Journal of Project Management 23 (2005) 275282
q
xy

n
P
n
i1
X
i
Y
i

P
n
i1
X
i
P
n
i1
Y
i
n
P
n
i1
X
2
i

P
n
i1
X
i
2
n
P
n
i1
Y
2
i

P
n
i1
Y
i
q ;
1
where X and Y denote two populations of variables,
each of which contains n observations.
The fundamental assumptions behind the product-
moment correlation are that the relationship between
two variables is linear and both populations are nor-
mally distributed. When these two assumptions cannot
be met, a more attractive alternative is to use the rank
correlation.
The rank correlation is essentially the product-mo-
ment correlation coecient between the ranks in each
set of the two variables. To compute the rank correla-
tion coecient, one should sort each set of variables
based on their values, replace the value by the rank,
and compute the product-moment correlation coe-
cient to the rank. A useful equation is
r
xy
1
6
P
n
i1
d
2
i
nn
2
1
; 2
where d is the dierences between ranks of correspond-
ing X and Y.
Both coecients can range between +1 and 1. A
correlation coecient of +1 signies a perfect positive
relationship, while 1 shows a perfect negative rela-
tionship. The smallest correlation is 0. Because correla-
tion coecients are dened between two cost elements,
they are traditionally stored in a two-dimensional cor-
relation matrix, which is always symmetric with unit
diagonal.
Similar to the marginal distributions, the correla-
tion matrix may be a result of statistic analysis on
historical data or of subjective judgment. The histori-
cal approach requires a test on the signicance of the
correlation coecients. A null hypothesis shall be
tested against an alternative hypothesis to determine
if a cost element is, not by coincidence, independent
from others.
The historical approach has been used to nd the
product-moment correlations in an Australian data
set drawn from 8 oce-building projects [20], and in
a British data set drawn from 216 oce-building pro-
jects [33]. The statistical approach has also been ap-
plied to calculate the rank correlations in the
samples of 131 American projects [29]. The empirical
results of these studies suggest the existence of
correlations.
On the other hand, subjective judgment also nds its
application in specifying the correlations between cost
elements qualitatively. To this respect, researchers can
subjectively choose two groups of correlations to assess
strong, moderate, and weak relations: {0.8,0.45,0.15}
[28] and {0.85,0.55,0.25} [7].
When a correlation matrix is used in simulation, an
important requirement is to ensure its feasibility, which
restricts the matrix to be positive semidenite regardless
of its type (product-moment or rank) or the way it is
estimated (historical or subjective) [19]. Being positive
semidenite means the eigenvalues of the correlation
matrix must be non-negative. In a mathematical form,
the requirement can be written as
E
T
ME P0; 3
where M is the correlation matrix, E, the eigenvector,
and E
T
, transpose of E.
In the literature, it has frequently occurred that the
correlation matrix is not positive denite as indicated
by Ranasinghe [24]. This is particularly an issue when
the number of dimensions increases because the possibil-
ity of having an infeasible correlation matrix will grow
rapidly as the dimension increases [17].
Two procedures were proposed to adjust an infeasible
correlation matrix. Tourans approach was to reduce all
the correlations slightly (say 0.01) and repeat until the
correlation matrix becomes feasible [28]. This approach
overlooks the possibility of increasing some correlations
while reducing others. Ranasinghe [24] developed a
computer program to iteratively calculate and list the
bounds of each correlation to make the matrix positive
semidenite. The program then asks the estimator to
change the original values and wait until the program
re-checks the feasibility and new bounds. This process
continues until reaching the feasibility. This approach,
however, may be time consuming due to its iterative
nature.
Our discussion leads to the second goal of this study:
to have an automatic procedure to check the feasibility
of the correlation matrix and adjust it if necessary. To
attain eciency, the adjustment should be performed
only once and in a prior fashion.
3. Proposed method
The proposed method takes two sets of input: mar-
ginal distributions of the cost elements (measured in unit
cost, for example /m
2
) and a correlation matrix be-
tween these elements. The method is composed of two
stages, which will be explained below.
3.1. Setup stage
The proposed method starts with a check on the fea-
sibility of the original correlation matrix. If it is already
positive semidenite, one can immediately begin the sim-
ulation steps described in the following section; other-
wise, we adopt the eigenvalue correction method from
Ghosh and Henderson [13] to approximate the infeasible
correlation matrix into a feasible one.
I.-T. Yang / International Journal of Project Management 23 (2005) 275282 277
The setup stage consists of the following steps:
1. Decompose the correlation matrix M into a diagonal
vector D of the eigenvalues and a full matrix V whose
columns are the corresponding eigenvectors so that
MV VD: 4
2. Locate the negative eigenvalues and change them to a
tiny positive number e to yield a new diagonal vector
D.
3. Adjust the correlation matrix M by
M VDV
T
: 5
4. Take the diagonal elements of M and store their
inverses as the diagonals in a full matrix E
E
ij

1=
M
ij
p
8 i j;
0; otherwise:
( )
6
5. Normalize the diagonal elements to ensure unit diag-
onals (fundamental requirement for a correlation
matrix)
M EME; 7
where Mis the new(approximated) correlation matrix.
A preliminary test is performed to ensure the eec-
tiveness of the setup stage. We use a documented
19 19 correlation matrix (Table 2 in [7]) as an exam-
ple. This correlation matrix has been proved to be
non-positive-semidenite [24]. Our tasks here are to
apply the proposed steps and to check if the dier-
ences between the approximated values and the origi-
nal specications are small enough. The dierences are
quantied in two metrics: L
ave
(average) and L
max
(maximum)
L
ave

P
i>j
j MM j
nn1
2
; 8
where M is the specied correlation matrix and M is the
approximated one; n is the number of cost elements
L
max
max
i>j
jMMj: 9
After performing the correction steps, L
ave
is
4.8954 10
18
and L
max
is 1.7208 10
15
. This shows
empirically that the setup stage can, de facto, adjust
the infeasible correlation matrix to a feasible one with
ignorable changes.
The setup stage is to treat possible infeasibility, which
may result from either erroneous input or inconsistent
estimation. In other words, the setup stage would not
be of any good if the correlation coecients are incor-
rect or inconsistent. Thus, careful review is critical to en-
sure the correlation coecients can reect the true
behavior of the correlation relationships.
3.2. Simulation stage
The fundamental concept of the simulation stage is
to generate a vector of correlated normal variates,
transform them into uniform variates by the aid of
the cumulative normal probability function, and then
map the variates into their individual marginal distri-
butions by the inverse transform method [18]. The gen-
erated random variates are used to model the cost
elements with the desired correlation structure. The
procedure described here incorporates ideas from a
new correlated multivariate generation technique (Nor-
mal To Anything, NORTA) [6]. In what follows, we
enumerate all the steps and provide computational
guidelines:
1. Apply the Cholesky decomposition to the correla-
tion matrix, so that M = CC
T
, where C represents
the Cholesky triangular.
2. Generate an IID (independent and identically dis-
tributed) unit scaled uniform random vector,
Y = (Y
1
,Y
2
, . . . ,Y
n
), where n is the number of cost
elements.
3. Translate Y into a standard-normal random vector
P = (P
1
,P
2
, . . . ,P
n
).
4. Transform P into a correlated standard-normal
random vector Z = (Z
1
,Z
2
, . . . ,Z
n
).
5. Compute for U
i
= U(Z
i
) for i = 1,2, . . . n, where
U(.) denotes the standard normal cumulative dis-
tribution function (CDF).
6. Compute for X
i
F
1
i
U
i
for i = 1,2, . . . n, where
F
1
i
U
i
represents the inverse of the ith marginal
CDF.
7. Return X
i
as the estimate for cost element i.
8. Compute the total project unit cost by summing
up all the cost elements.
9. Repeat Steps 28 for each simulation replication,
j = 1, 2, . . . ,m.
10. Return summary statistics on all simulation
replications.
For Step 1, there exist several ecient algorithms to
perform the Cholesky decomposition (see [14,22]). The
generation of a uniform random vector in Step 2 is a
standard feature supported by almost all the popular
computer languages (such as C++, Java, Visual Basic,
FORTRAN). The transformation in Step 3 can be
approximated by the following equation [27]:
P
i
Y
0:135
i
1 Y
i
0:135
=0:1975: 10
The transformation in Step 4 is
Z
i

X
i
j1
c
ij
P
j
for c
ij
2 C: 11
Step 5 involves the following integral:
U
i

Z
Zi
1
1
2p
p e
x
2
=2
dx; 12
which can be approximated by [1]
where:
t = 1/(1 + 0.2316419 jZ
i
j);
b
1
= 0.319381530;
b
2
= 0.356563782;
b
3
= 1.781477937;
b
4
= 1.821255978;
b
5
= 1.330274429.
The marginal CDF in Step 6 can be of any type of
distribution as long as their inverses can be calculated
either directly or via approximation. This is why the
proposed method is able to treat dierent kinds of distri-
butions simultaneously. It is much easier when the mar-
ginal distribution has a closed-form inverse (such as
uniform or triangular). Otherwise, one has to rely on
numerical approximation algorithms to nd the inverses
of the commonly-used distributions, such as beta [9],
gamma [5], and normal distributions [21].
By using X
i
as the estimate of the ith cost element,
Step 8 is simply
E
j

X
n
i1
X
i
; 14
where E
j
is the jth observation of the unit cost of the
project. To make the addition meaningful, the estimates
of dierent cost elements should be converted into the
same unit of measure, such as /m
2
.
4. Practical application
The proposed method is applied to the British data
set described in [33] to demonstrate its practical use.
The data set is drawn from 216 oce buildings built be-
tween 1980 and 1994 and consists of 8 major cost ele-
ments. The data set has been standardized based on
the times and locations the buildings were built.
All the cost elements and their marginal distributions
are shown in Table 1. The value of each cost element is
expressed as /m
2
. Here a cost element represents a rel-
atively large work package, which may consist of several
tasks. For example, superstructure involves form-
work, steelwork, and concrete pouring. This level of
granularity is suitable for higher level estimation. More-
over, the measure of /m
2
can be changed to reect the
usual unit for progress measurement, if the proposed
method is applied to other construction projects. For in-
stance, a reasonable measure of cost elements for a high-
way project may be /m while that for a residential
community project may be /house.
In the example, we consider three families of distribu-
tions, i.e., lognormal, beta, and discrete. The lognormal
distributions are used because they t the data better as
argued by Wall [33]. The use of the other two is based on
a pragmatic situation when a cost estimator prefers not
using historical data but rather using a discrete distribu-
tion to describe possible outcomes of tting and fur-
nishings, and beta distributions (three points) to
estimate the distributions of services and external
works. These arrangements have been justied in previ-
ous sections.
Table 2 shows the rank correlation coecients be-
tween the cost elements of the full data set. Before
applying the proposed method, the rank correlation
coecients are reviewed and adjusted to verify (1) if they
can reect the actual behavior of the correlations and (2)
if they, derived from past data, are suitable for the cur-
rent project. This process is based on practical judg-
ments and can complement pure mathematic analysis.
In this example, the rank correlation coecients be-
tween external works and other cost elements are ad-
justed to be zero.
A simulation experiment is designed to implement the
proposed method and to evaluate the impact of correla-
tions between cost elements. In the experiment, every
simulation replication leads to a sample of the project
U
i

1=
2p
p
expZ
2
i
=2 b
1
t b
2
t
2
b
3
t
3
b
4
t
4
b
5
t
5
if Z
i
< 0;
1 1=
2p
p
expZ
2
i
=2 b
1
t b
2
t
2
b
3
t
3
b
4
t
4
b
5
t
5
if Z
i
P0;
( )
13
Table 1
Descriptive estimates for cost elements (distributions and parameters)
Cost elements Descriptive estimate (/m
2
)
Substructure Lognormal (47.2,30.9)
a
Superstructure Lognormal (263.6,82.4)
a
Internal nishes Lognormal (63.2,24.4)
a
Fittings and furnishings Discrete (7,0.2; 8,0.5; 9,0.2; 10,0.1)
b
Services Beta (150,180,220)
c
External works Beta (70,85,120)
c
Preliminaries Lognormal (76.4,47.3)
a
Contingencies Lognormal (21.2,13.2)
a
a
Lognormal (mean, standard deviation); the lognormal distributions
are estimated by the historical approach based on 216 buildings.
b
Discrete (outcome, probability); the discrete distribution is sub-
jectively specied.
c
Beta (minimum, mode, maximum); the three parameters are sub-
jectively specied.
cost by simply summing up cost elements drawn from
individual distribution. The output statistics can then
be used to assess the behavior of the true project cost.
Before a full-scale simulation, validation runs are con-
ducted to ensure the code is correct and the random
variables have the specied distributions.
After 1000 simulation replications, Table 3 lists the
descriptive statistics for the unit cost of the project. To
assess the impact of correlations, we compare two sce-
narios: including and excluding correlations. Fig. 1 is a
box-and-whisker plot which is used to visually compare
the distributions of the two scenarios. The rst observa-
tion is that both distributions are skewed to the right be-
cause the mean (shown as the cross) is larger than the
median.
The second observation is that the scenario of
including correlations has a much longer tail to the
right than that of excluding correlation. This indicates
the former has a larger variability (uncertainty) than the
latter. This conclusion is unsurprising because the for-
mer has a much greater standard deviation that the lat-
ter (149.55 versus 108.92, a 37% dierence).
Consequently, the 95% condence interval of the former
is much wider than that of the latter.
Fig. 2 plots the CDFs of both scenarios. A practical
use of the chart is to estimate the unit cost of the pro-
ject with a certain probability. Taking correlations into
consideration, the unit cost with a 0.90 probability is
958.50 /m
2
, which would be profoundly underesti-
mated as 903.52 /m
2
if the correlations are neglected.
The dierence of 54.98 /m
2
is greater than the cost of
substructure (with a mean of 47.2 /m
2
in Table 1).
In other words, by neglecting the correlations, the er-
ror can be as serious as doing the substructure for
free.
Table 2
Rank correlation coecients between cost elements
a
Substructure Superstructure Internal nishes Fittings and
furnishings
Services External
works
Preliminaries Contingencies
Substructure 1.00
Superstructure 0.33 1.00
Internal nishes 0.26 0.52 1.00
Fittings and furnishings 0.10 0.26 0.28 1.00
Services 0.28 0.57 0.64 0.33 1.00
External works 0.00
b
0.00
b
0.00
b
0.00
b
0.00
b
1.00
Preliminaries 0.35 0.37 0.44 0.18 0.39 0.00
b
1.00
Contingencies 0.23 0.28 0.34 0.21 0.29 0.00
b
0.36 1.00
a
Correlations above 0.10 signicant at 95% condence.
b
Subjective correlations.
Table 3
Statistics of two scenarios: including and excluding correlation
(in /m
2
)
Statistics Excluding
correlations
Including
correlations
Mean 759.21 756.88
SD 108.92 149.55
Minimum 514.50 470.50
Q1 (25% percentile) 680.99 647.07
Q2 (Median) 759.21 756.88
Q3 (75% percentile) 823.71 843.35
Maximum 1147.20 1393.30
95% CI lower bound 590.30 522.00
95% CI upper bound 1024.00 1091.00
Estimate with 0.9 probability 903.52 958.50
Fig. 1. Box-and-whisker plot for comparison between two scenarios: including and excluding correlations.
5. Validation
In this section, we validate whether the proposed
method can capture the desired correlation matrix.
Note, however, the proposed method is an approxima-
tion because of the following reasons. First, it is as-
sumed that the correlation between X
i
and X
j
in Step 7
(denoted by M
X
) is close to the correlation between Z
i
and Z
j
in Step 4 (denoted by M
Z
). Theoretically, to nd
a proper M
Z
that leads to the desired M
X
requires solv-
ing n(n 1)/2 non-linear equations but the computation
can be cumbersome [8]. Second, Steps 3, 5, and 6 require
numerical approximation techniques.
Since the proposed method is an approximation, it is
necessary to check the aggregated dierence between the
original specied correlation matrix and the generated
one on the aforementioned metrics: L
ave
and L
max
de-
ned by Eqs. (8) and (9). For this particular application,
L
ave
is 0.018 and L
max
is 0.051. Moreover, the standard
deviation of the dierences is 0.015. Thus, the condence
limits for the mean of the dierence is estimated to be
0.018 0.0059 at a two-tailed signicance level of 0.05
with 27 degrees of freedom. The empirical numbers give
us condence that the proposed method, despite being
an approximation, can model the desired rank correla-
tions with relatively small deviations and thereby can
help assess the true impact of correlations on cost
estimation.
6. Closing remarks
In this paper, we propose a simulation-based method
to estimate project cost while considering correlations
between cost elements. The proposed method is more
general than previous approaches because (1) it can treat
dierent types of marginal distributions (discrete or con-
tinuous, dierent families distributions) for cost ele-
ments in one framework and (2) it can automatically
adjust an infeasible correlation matrix into a close and
feasible one very eciently.
This paper introduces two sets of input data (mar-
ginal distributions of the cost elements and their correla-
tions) and elaborates on how they are specied
(historical and subjective approaches). The historical
and subjective approaches possess their own characteris-
tics and should both be allowed in the process of cost
estimation. This leads to the development of the pro-
posed method.
The modeling capabilities of the proposed method
are empirically validated by an application to a modied
British data set consisted of 216 oce buildings. With
the modeling capabilities, the proposed method helps
cost estimators assess the true impact of correlations be-
tween cost elements on the project unit cost. The impact
has been shown signicant and should be considered
with caution.
With the recognition that the proposed method is an
approximation, users are suggested to perform the vali-
dation analysis as we demonstrated in this paper to en-
sure the deviation between the sampled and specied
correlation matrices is within an acceptable range. It is
also important to note that the proposed method relies
heavily on the input of correlation data, which require
not only mathematical computations but also practical
judgments.
Acknowledgments
The author is thankful to Professor Shane Henderson
and Dr. Soumyadip Ghosh at Cornell University and
Professor Bruce Schmeiser at Purdue University for
their helpful communications and discussions.
References
[1] Abramowitz M, Stegun IA. Handbook of mathematical functions
with formulas, graphs and mathematical tables 1993:932.
[2] Attoh-Okine NO. Probabilistic analysis of factors aecting
highway construction costs: a belief network approach. Can J
Civil Eng 2002;29(3):36974.
[3] Badri MA, Mortagy A, Davis D, Davis D. Eective analysis and
planning of R&D stages: a simulation approach. Int J Proj Manag
1997;15(6):3518.
[4] Benjamin JR, Cornell CA. Probability, statistics, and decision for
civil engineers. New York: McGraw-Hill; 1970.
[5] Best DJ, Roberts DE. The percentage points of the v
2
distribu-
tion. Appl Stat 1975;24:3858.
[6] Cario MC, Nelson BL. Modeling and generating random vectors
with arbitrary marginal distributions and correlation matrix.
Fig. 2. Comparison on cumulative distribution functions of two
scenarios: including and excluding correlations.
Technical Report, Department of Industrial Engineering and
Management Sciences, Northwestern University; 1997.
[7] Chau KW. Monte-Carlo simulation of construction costs using
subjective data. Construct Manag Econ 1995;13:36983.
[8] Chen H. Initialization for NORTA: generation of random vectors
with specied marginals and correlations. INFORMS J Comput
2001;13:31231.
[9] Cran GW, Martin KJ, Thomas GE. A remark on algorithm
AS63: the incomplete beta integral, AS64: inverse of the incom-
plete beta function ratio. Appl Stat 1977;26:1114.
[10] Diekmann JE. Probabilistic estimating: mathematics and
applications. J Construct Eng Manag ASCE 1983;109(3):
297308.
[11] Elkjaer M. Stochastic budget estimation. Int J Proj Manag
2000;18(2):13947.
[12] Evans JR, Olson DL. Introduction to simulation and risk
analysis. 2nd ed. Englewood Clis, NJ: Prentice-Hall; 2002.
[13] Ghosh S, Henderson SG. Behavior of the NORTA method for
correlated random vector generation as the dimension increases.
ACM Transact Model Comp Simul 2003;13(3):27694.
[14] Golub GH, Van Loan CF. Matrix computations. 3rd ed. Balti-
more, MD: John Hopkins University Press; 1996.
[15] Hayes RW, Perry JG, Thompson PA, Willmer G. Risk
management in engineering constriction, implication for project
managers. Project management group. Thomas Telford Ltd;
1986.
[16] Johnson D. The triangular distribution as a proxy for the beta
distribution in risk analysis. J Roy Stat Soc (Ser D): Statistician
1997;46(3):38798.
[17] Kurowicka D, Cooke RM. Conditional, partial and rank corre-
lation for the elliptical copula; dependence modeling in uncer-
tainty analysis. Technical Report, Delft University of Technology,
Netherlands; 2001.
[18] Law A, Kelton WD. Simulation modeling and analysis. 3rd
ed. New York: McGraw-Hill; 1999.
[19] Lurie PM, Goldberg MS. An approximate method for sampling
correlated random variables from partially specied distributions.
Manag Sci 1998;44:20318.
[20] Newton S. Methods of analyzing risk exposure in the cost
estimates of high quality oces. Construct Manag Econ
1992;10:43149.
[21] Odeh RE, Evans JO. The percentage points of the normal
distribution. Appl Stat 1974;23:967.
[22] Press WH, Teukolsky SA, Vetterling WT, Flannery BP. Numer-
ical recipes in C: the art of scientic computing. 2nd ed. Cam-
bridge: Cambridge University Press; 1992.
[23] Raftery J. Risk analysis in project management. E&FN Spon
1994.
[24] Ranasinghe M. Impact of correlation and induced correlation on
the estimation of project cost of buildings. Construct Manag Econ
2000;18:395406.
[25] Russell AD, Ranasinghe M. Analytical approach for economic
risk quantication of large engineering projects. Construct Manag
Econ 1992;10:277301.
[26] Schaufelberger JE, Holm L. Management of construction pro-
jects. Precedence Hall; 2002.
[27] Schmeiser BW. Some myths and common errors in simulation
experiments. In: Proceedings of the winter simulation conference,
2001. p. 3946.
[28] Touran A. Probabilistic cost estimation with subjective correla-
tions. J Construct Eng Manag ASCE 1993;119(1):5871.
[29] Touran A, Suphot L. Rank correlation in simulating construction
costs. J Construct Eng Manag ASCE 1997;123(3):297301.
[30] Touran A, Wiser EP. Monte-Carlo technique with correlated
random variables. J Construct Eng Manag ASCE
1992;118(2):25872.
[31] Tummala VMR, Burchett JF. Applying a risk management
process (RMP) to manage cost risk for an EHV transmission line
project. Int J Proj Manag 1999;17(4):22335.
[32] Van Dorp JR, Kotz S. A novel extension of the triangular
distribution and its parameter estimation. J Roy Stat Soc (Ser D):
Statistician 2002;51(1):6379.
[33] Wall DM. Distributions and correlations in Monte-Carlo simu-
lation. Construct Manage Econ 1997;15:24158.
[34] Wang WC. Simulation-facilitated model for assessing cost corre-
lations. J Comp Aided Civil Infrastruct Eng 2002;17(5):36880.
Factors aecting cost performance: evidence
from Indian construction projects
K.C. Iyer
a,
*
, K.N. Jha
b
a
Department of Management Studies, Indian Institute of Technology, Hauz Khas, New Delhi 110016, India
b
Department of Civil Engineering, Indian Institute of Technology, Kanpur, Uttar Pradesh 208016, India
Received 2 December 2003; received in revised form 30 March 2004; accepted 8 October 2004
Abstract
This paper presents the ndings of a questionnaire survey conducted on the factors aecting cost performance of Indian construc-
tion projects. Factor analysis of the response on the 55 success and failure attributes identied through literature review and per-
sonal interview extracted seven factors. Critical success factors obtained by the analyses are: project managers competence; top
management support; project managers coordinating and leadership skill; monitoring and feedback by the participants; coordination
among project participants; and owners competence and favourable climatic condition. However factors adversely aecting the cost
performances of projects, are: conict among project participants; ignorance and lack of knowledge; presence of poor project specic
attributes and non existence of cooperation; hostile socio economic and climatic condition; reluctance in timely decision; aggressive com-
petition at tender stage; and short bid preparation time. Further analysis indicates coordination among project participants as the
most signicant of all the factors having maximum positive inuence on cost performance.
Keywords: Cost; Managing projects; Success; Coordination; Factor analysis
1. Introduction
Measuring the performance of any construction pro-
ject in terms of success or failure though looks simple, is
in fact a very complex process. Modern construction
projects even moderate in size are generally multidisci-
plinary in nature and they involve participation of
designers, contractors, subcontractors, specialists, con-
struction managers, and consultants. With the increas-
ing size of the project, number of participants in the
project also increases. The objectives or goals of all par-
ticipants need not be same even in a given project. Hence
to dene the success or failure of a project without spec-
ifying the participant and without specifying the criteria
for judging the performance holds no meaning. Success
for one participant may be a failure for the other partic-
ipant depending on the perspective with which each one
is looking at the outcome. Incidentally, past researchers
have employed dierent criteria, such as compliance to
schedule, cost and quality to judge the project perfor-
mance. However, in the present paper only the factors
aecting cost compliance criteria have been discussed.
Researchers in the past have identied various causes
or reasons (called attributes in this paper) for project
success. Their works are area specic or project specic
and are mostly from the developed countries and based
on researchers experience on completed projects. Fur-
ther these researchers have identied the critical attri-
butes that are responsible for success of the projects
and they advise that these attributes should be carefully
handled and if possible be further exploited to achieve
greater success. There are also certain other attributes
doi:10.1016/j.ijproman.2004.10.003
*
Corresponding author. Tel.: +91 11 2659 1209; fax: +91 11 2686
2620.
E-mail address: kciyer@dms.iitd.ac.in (K.C. Iyer).
PROJECT
MANAGEMENT
responsible for failure of a project. These attributes, if
not handled properly may be detrimental for the success
of the project. This study tries to identify all such success
and failure attributes for the construction projects in a
developing country like India. Thus the objectives set
for the study are given below.
2. Objectives and methodology
The objectives of the study set are as follows:
To identify the relative importance of success and
failure attributes in Indian construction projects as
perceived by the professionals including owners and
contractors of the construction industry.
To understand the latent properties of these success
and failure attributes.
The study is however restricted to construction stage
of projects and the study required a huge amount of
documented data on completed projects. Due to non-
availability of documented data of completed projects
for study in India, a questionnaire survey approach is
considered to establish the impact of various attributes
on project performance. Attributes aecting project suc-
cess and failure are listed out through literature survey
and interviews with select professionals from construc-
tion industry. Pilot survey is then undertaken and neces-
sary modications in the questionnaire are carried out.
Questions are framed to ascertain the impact of these
attributes individually on project performance evaluat-
ing parameters such as completion schedule, project
cost, project quality and project disputes.
3. Literature review
Project success is repeatable and it is possible to nd
certain success attributes has been the genesis of many re-
search studies in this area since 1960s. These studies range
from theoretical work based on experience of researcher
on one end to structured research work on the other
end. Some of the past researchers have adopted question-
naire survey approach for data collection and employed
mathematical tools like AHP, Neural networks and sta-
tistical techniques like factor analysis and multivariate
regression, etc., for analysis and drawing conclusions.
But these researchers seem to use their own performance
criteria for measuring success and not much commonality
appears among various performance criteria adopted by
these researchers. The criteria included the conventional
ones like schedule, cost, quality to recent criteria like
perceived performance, client satisfaction, etc. Summa-
ries of important conclusions of previous studies [113]
are given in Table 1 to provide an overview.
4. Identication of attributes aecting project
performance
Literature review reveals a number of critical attri-
butes for dierent types of projects viz., construction
projects, research and development projects, defence
projects, etc. It can be seen from Table 1 that most of
these studies have focused on specic success measuring
parameters, and the critical success factors derived are
either applicable to a particular industry or contract
type. It can be also observed from Table 1 that almost
all the studies have been carried out in the developed
countries context and their ndings may not be relevant
to the developing countries. However, taking lead from
above studies and through personal interviews with con-
struction professionals of India, 55 project success/pro-
ject failure attributes were identied. Though the list
of 55 attributes may not be called exhaustive due to
the vast magnitude and fragmented nature of construc-
tion industry, the list covered attributes pertaining to a
large variety of construction projects. A survey was then
framed to get respondents views on impact of these
attributes on project outcome (success or failure). The
performance evaluation parameters considered to mea-
sure the impacts of these attributes were restricted to
schedule, cost, quality and occurrence of dispute
to limit the size of the questionnaire. However impact
of these attributes on cost parameter is only discussed
in this paper and only the relevant portion of the ques-
tionnaire is given in Appendix A. A ve-point scale was
used to measure the attributes inuence on cost perfor-
mance. In this scale, 1 represents adversely aecting the
cost of the project, 2 represents signicantly aecting
the cost, 3 represents marginally aecting the cost, 4
represents no eect on the cost, and 5 represents helps
in saving in the cost of the project.
A total of 450 questionnaires were mailed to top In-
dian construction industry professionals covering about
50 top and medium size organizations, selected randomly
from across the country. A total of 112 completed re-
sponses were received giving a response rate of 25%.
The responses are stored and analysed using SPSS soft-
ware. The respondents included the owners, contractors
and consultants. Since the consultants were employed by
the owners to take care of the owners interests, and that
the consultants responses were not signicantly dierent
from that of owners, they were merged with owners re-
sponses. The next step was to rank the attributes in the
order of their criticality. From the ve point scale used
in the questionnaire, the mean scores of responses (l)
for dierent project attributes are interpreted in the man-
ner as given in Table 2. Depending upon the mean scores
of responses for various attributes, the attributes were
then segregated in three groups: the rst group (with
l P4.5) that showed positive contribution; the second
group (with 4.5 < l < 3.5) which was neutral and passive
284 K.C. Iyer, K.N. Jha / International Journal of Project Management 23 (2005) 283295
Table 1
Summary of studies in the eld of project success/failure factors
Author Summary of the past works
Rubin and Seeling [1] Authors assess the impact of project managers experience on the projects success/failure. They use technical
performance of the project as a measure of success and conclude the following
(a) Project managers previous experience has minimal impact on the projects performance
(b) Size of the previously managed project does aect the managers performance
Avots [2] This is a theoretical study to understand the reasons for project failure. The author concludes that choice of wrong
project manager, the unplanned project termination, and unsupportive top management are the main reasons of
failure
Sayles and Chandler [3] Authors conclude the following factors as critical for success of a project
Project managers competence, scheduling, control systems and responsibilities, monitoring and feedback and
continuing involvement in the project
Martin [4] Author concludes the following factors as critical for success of a project
Dene goals, select project organizational philosophy, general management support, organize and delegate authority,
select project team, allocate sucient resources, provide for control and information mechanics, require planning and
review
Baker et al. [5] The authors suggest that instead of time, cost and performance as the project success criteria, perceived performance
should be used as the success criteria. They observe the following success factors
Clear goals, goal commitment of project team, on site project manager, adequate funding to completion, adequate
project team capability, accurate initial cost estimates, minimum start-up diculties, planning and control techniques,
task (vs. social orientation), absence of bureaucracy
Cleland and King [6] Authors identify the following success factors
Project summary, operational concept, top management support, nancial support, logistic requirements, facility
support, market intelligence (who is the client), project schedule, executive development and training, manpower and
organization, acquisition, information and communication channels and project review
Locke [7] Author identies the following success factors
Make project commitments known, project authority from the top, appoint competent project manager, set up
control mechanisms (schedules, etc.), progress meetings
Hughes [8] The author identies that the projects fail because of improper basic managerial principles, such as the improper focus
of the management system, by rewarding the wrong actions, and the lack of communication of goals
Morris and Hough [9] Authors identify the following success factors through a study of eight large and complex projects having great
potential economic impact but poorly managed and generally failed
Project objectives, technical uncertainty innovation, politics, community involvement, schedule duration urgency,
nancial contract legal problems and implementation problems
Schultz et al. [10] Authors classify critical success factors in two groups as given below and conclude that these groups aect project
performances at dierent phases of implementation
(a) The Strategic Group consisting of factors like project mission, top management support and project scheduling
(b) Tactical group consisting of factors like client consultation and personnel selection and training
Pinto and Slevin [11] Continuing the previous work authors evaluate the relative importance of tactical group and strategic group of factors
over the project life cycle. They conclude that when external success measures are employed, planning factors
dominate tactical factors throughout the project life cycle
Chua et al. [12] Budget performance is given the primary importance in the study. Through an application of neural network
approach authors identify the eight important project management attributes associated with achieving successful
budget performance: (1) number of organisational levels between the project manager and craft workers; (2) amount
of detailed design completed ate the start of construction; (3) number of control meetings during the construction
phase; (4) number of budget updates; (5) implementation of a constructability program; (6) team turnover; (7) amount
of money expended on controlling the project; (8) the project managers technical experience
They also claim that their model can be used as a predictive tool to forecast budget performance of a construction
project
Chan et al. [13] They identify a set of project success factors for design and build (D&B) projects and examine the relative importance
of these factors on project outcome. Using factor analysis from the response of 53 participants on 31 variables they
extracted six project success factors. These are project team commitment, contractors competencies, risk and liability
assessment, clients competencies, end-users needs, and constraints imposed by end-users. Further Project team
commitment, clients competencies, and contractors competencies were found to be important to bring successful
project outcome from multiple regression ndings
K.C. Iyer, K.N. Jha / International Journal of Project Management 23 (2005) 283295 285
having no signicant impact on the project outcome and
the third group (with l 6 3.5) indicating negative im-
pact. A total of 30 attributes emerged in rst group
(l P4.5), only 3 attributes in group 2 (4.5 < l < 3.5),
and remaining 23 attributes fell in group3 (l 6 3.5). So
it was decided to drop the project attributes of the second
group (with 4.5 < l < 3.5).
Chan and Kumaraswamy [14] are of the opinion that
the mean and standard deviation of each individual
attribute is not suitable statistics to assess the overall
rankings because they do not reect any relationship be-
tween them and accordingly they have used the relative
importance index (RII) method to determine the rela-
tive ranking of the attributes. The RII is evaluated using
the following expression:
Relative importance indexRII
P
w
AxN
; 1
where w is the weight given to each attribute by the
respondents and ranges from 1 to 5, A is the highest
weight (i.e., 5 in this case), and N is the total number
of respondents.
The attributes of the rst group (with l P4.5) were
arranged on their descending order of RII values and
ranked. The highest RII indicates the most critical suc-
cess attributes with rank 1 and the next indicating the
next most critical success attribute with rank 2 and so
on. On the other hand, attributes of the third group
(with l 6 3.5) were arranged in the ascending order of
the RII and ranked. The lowest RII indicates the most
critical failure attribute with rank 1, the next indicating
the next most critical failure factor 2 and so on. Separate
lists of the success attributes and the failure attributes
are given along with their RII and ranks in Tables 3
and 4, respectively.
5. Success attributes
Using the RII, the rank orders of attributes were ob-
tained for all responses as well as separately for owner
and contractor responses. The high ranked attributes
(top 5 in Table 3) that are observed in three categories
of responses: all reponse, owner, and contractor are dis-
cussed below. The all response column (columns 2 and 3
in Table 3) shows the most important factor to be mon-
itoring and feedback by PM with RII = 0.958, followed
by coordinating ability and rapport of PM with top man-
agement, monitoring and feedback by project team mem-
bers positive attitude, and technical capabilities of PM as
indicated by their decreasing RIIs. Here the PM (pro-
ject manager) refers to contractors project manager.
The owners view (columns 4 and 5 in Table 3) the
coordination ability and rapport of PM with top man-
agement as the most important attribute (RII = 0.979)
followed by the monitoring and feedback by team mem-
bers and monitoring and feedback by PM. Also owners
realise the importance of their own timely decision either
directly or through their engineer as fourth most suc-
cessful attribute.
The most important attribute in contractors view
(columns 6 and 7 in Table 3) is found to be the project
managers technical capability (RII = 0.957) followed
by his authority to take nancial decision and select key
team members, and positive attitude of PM and partici-
pants and eective monitoring and feedback.
The importance of the attribute monitoring and feed-
back can also be gauged from the fact that almost all the
surveyed companies have their MIS (Management
Information System) in place.The companies do insist
on getting the feedback from site management at regular
interval.
The reason of attribute, coordinating ability and rap-
port of PM with top management being given the impor-
tance is due to the fact that most of the times timely help
from top management in getting the resources or getting
critical decision can have far reaching implications on
cost performance. This is especially true for short dura-
tion projects where each day of delay can have large det-
rimental eect on cost.
The emergence of positive attitude of PM and project
participants among top ve attributes substantiates the
widely held view that many dicult things can be made
simple and workable with a positive attitude. Subse-
quent to analysis of responses during validation stage
of research ndings, it has been observed that the sur-
veyed companies organise regular training programmes
to generate a sense of belongingness and create positive
attitude in their project managers. Project managers
technical ability getting high rank (in top 5 in all re-
sponses and in contractors response) implies the realisa-
tion of high importance of this activity by organisation.
High technical skill of project manager indicated by
PMs academic record is ensured during the selection
process itself in most of the surveyed companies.
Table 2
Interpretation of various ranges of mean values of responses pertaining to project attributes
Project performance evaluation parameter Range of mean values
Group 1 Group 2 Group 3
l P4.5 4.5 < l < 3.5 l 6 3.5
Cost Helps in saving Neither positively contributing nor adversely aecting Adversely aecting
6. Failure attributes
Table 4 lists the group 3 attributes having (l 6 3.5),
i.e., attributes causing adverse eect on cost perfor-
mance of project. It can be seen from the scale of ques-
tionnaire, lesser the value of mean value of responses on
any attribute, more severe will it be. Hence the rank or-
der is decided based on increasing (ascending) values of
RIIs of attributes. The table actually lists the failure
attributes. It can be seen from the table that there is a
clear consensus between the two groups representing
contractors and owners in the ranking of most critical
failure attribute. They all have rated Poor human re-
source management and labour strike as the most criti-
cal failure attribute. The next top ranked failure
attributes are negative attitude of PM and project partic-
ipants (0.329), inadequate project formulation in the
beginning (0.333).
In addition the contractors have also ranked vested
interest of client representative in not getting the project
completed in time as the most critical attribute. The
interviewees narrated a number of projects that faced se-
vere cost and schedule overrun due to this reason. The
owners also tend to accept this fact by placing this attri-
bute among top ve critical failure attributes.
Some of the other failure attributes ranking high in
the list are conicts between PM and top management,
the project ailments like ignorance and lack of knowledge,
and indecisiveness.
Subsequent to the nding of ranks of all success and
failure attributes, their ranks under owner and contrac-
tor respondents are compared and Spearmans rank cor-
relation coecient R
s
is found. It is observed that while
there is a strong agreement between owner and contrac-
tor for failure attributes (R
s
= 0.910), they seem to dier
in success attributes (R
s
= 0.271). It means that while
there is an agreement between the two sets of respon-
dents about the causes of the failure, there is dierence
in opinion towards the reasons for the successful cost
performance of the project.
7. Critical success and failure factors
In the present study, factor analysis is performed sep-
arately on group of 30 success attributes and 23 failure
Table 3
Critical success attributes of projects (l P4.5)
Project attributes All response Owner Contractor
RII Rank RII Rank RII Rank
1 Eective monitoring and feedback by PM 0.958 1 0.969 3 0.943 4
2 Coordinating ability and rapport of PM with top management 0.958 2 0.979 1 0.929 7
3 Eective monitoring and feedback by the project team members 0.955 3 0.974 2 0.929 8
4 Positive attitude of PM, and project participants 0.947 4 0.950 6 0.943 3
5 Project managers technical capability 0.938 5 0.925 11 0.957 1
6 Understanding operational diculties by the owner engineer
thereby taking appropriate decisions
0.934 6 0.928 10 0.943 4
7 Timely decision by the owner or his engineer (reluctance or otherwise) 0.931 7 0.964 4 0.886 23
8 Selection of PM with proven track record at an early stage by top management 0.925 8 0.938 7 0.907 19
9 Authority to take day to day decisions by the PMs team at site 0.924 9 0.925 11 0.923 10
10 Scope and nature of work well dened in the tender 0.922 10 0.925 11 0.919 12
11 Monitoring and feedback by top management 0.921 11 0.958 5 0.871 26
12 Understanding of responsibilities by various project participants 0.921 12 0.930 8 0.907 16
13 Leadership quality of PM 0.917 13 0.911 16 0.926 9
14 Top managements enthusiastic support to the project manager (PM) and project team at site 0.916 14 0.918 15 0.914 13
15 Coordinating ability and rapport of PM with his team members and sub-contractor 0.916 15 0.923 14 0.907 16
16 Project managers authority to take nancial decision, selecting key team members, etc. 0.910 16 0.882 26 0.950 2
17 Commitment of all parties to the project 0.906 17 0.895 20 0.921 11
18 Coordinating ability and rapport of PM with owner representatives 0.903 18 0.900 17 0.907 16
19 Coordinating ability and rapport of PM with other contractors at site 0.903 19 0.880 27 0.936 6
20 Top managements backing up the plans and identify critical activities 0.902 20 0.900 17 0.904 20
21 Regular budget update 0.899 21 0.887 22 0.914 13
22 Delegating authority to project manager by top management 0.897 22 0.884 24 0.914 13
23 Training the human resources in the skill demanded by the project 0.885 23 0.885 23 0.886 22
24 Ability to delegate authority to various members of his team by PM 0.884 24 0.882 25 0.886 24
25 Construction control meetings 0.882 25 0.875 29 0.893 21
26 Favourable political & economic environment 0.882 26 0.930 8 0.808 30
27 Favourable climatic condition at the site 0.881 27 0.897 19 0.859 27
28 Availability of resources (funds, machinery, material, etc.)
as planned throughout the project duration
0.879 28 0.880 27 0.879 25
29 Monitoring and feedback by client 0.870 29 0.892 21 0.837 29
30 Developing & maintaining a short and informal line of communication among project team 0.853 30 0.850 30 0.857 28
attributes. Factor analysis is a powerful method of sta-
tistical analysis that aims at providing greater insight
of relationship among numerous correlated, but seem-
ingly unrelated, variables in terms of a relatively few
underlying factor variate [15,16]. Many researchers from
other areas including politics, sociology, economics, hu-
manmachine systems, accident research, taxonomy,
biology, medicine, geology, and construction manage-
ment have also applied this technique [13,1719]. Factor
analysis is performed for responses of all respondents as
well as separately for owner and contractor responses.
While the all response set has given some meaningful
interpretations, the variables emerging in various factors
under isolated response sets of contractor and owner
have been found out to be jumbled and not leading to
any meaningful interpretation. Hence the analysis and
subsequent discussion is restricted to factor analysis of
all responses. Since the factors extracted using principle
component analysis are orthogonal and contain a large
number of overlapping attributes across various factors
it is not amenable to understand. So oblique rotation
using varimax rotation is employed. Altogether seven
success factors are extracted for each group of success
attributes and failure attributes. In the case of success
attributes group, these factors explain a total of about
75% of the variance whereas for the failure attributes
group, factors extracted explain about 70% of the vari-
ance. Details of the critical success factors and critical
failure factors are presented in Tables 5 and 6,
respectively.
7.1. Critical success factors
The description of success factors is given in the fol-
lowing sections.
7.1.1. Project managers competence
This has four aspects associated with it. The rst is
the inherent traits that the project manager has such
as his technical capability, leadership quality, and his
positive attitude. The second aspect is the empowerment
of his team through delegation of authority to take day
to day decisions, making his team understand their
responsibilities and generating a sense of commitment
in them, developing and maintaining a short and infor-
mal line of communication among his team, and train-
ing the human resources in the skill demanded by the
project. The third aspect is to get empowered himself
though demanding authority to take nancial decision,
and selecting key team members, etc., and getting the re-
quired resources (funds, machinery, material, etc.) as
planned throughout the project duration from his higher
ups. It is not enough to possess the skills mentioned
above unless the Project manager exerts himself for
the project by getting involved in the project through
regular budget update and taking active part in con-
struction control meetings. All these four characteristics:
inherent or personal traits; empowering team; getting
empowered; and getting involved with the project are
typical characteristics of a competent project manger.
Several attributes emerging in this factor (Table 5) also
Table 4
Critical failure attributes of projects (l 6 3.5)
Project attributes All response Owner Contractor
RII Rank RII Rank RII Rank
1 Poor human resource management and labour strike 0.309 1 0.295 1 0.329 1
2 Negative attitude of PM, and project participants 0.341 2 0.350 3 0.329 3
3 Inadequate project formulation in the beginning 0.346 3 0.355 4 0.333 4
4 Vested interest of client representative in not getting the project completed in time 0.350 4 0.365 5 0.329 1
5 Conicts between PM and top management 0.355 5 0.345 2 0.370 8
6 Mismatch in capabilities of client and architect 0.367 6 0.374 6 0.357 7
7 Conicts between PM and other outside agency such as owner,
sub-contractor or other contractors
0.368 7 0.380 8 0.350 5
8 Reluctance in timely decision by PM 0.376 8 0.390 9 0.356 6
9 Lack of understanding of operating procedure by the PM 0.391 9 0.379 7 0.407 13
10 Conicts among team members 0.391 10 0.390 10 0.393 11
11 Ignorance of appropriate planning tools and techniques by PM 0.397 11 0.395 11 0.400 12
12 Holding key decisions in abeyance 0.400 12 0.411 13 0.385 10
13 Reluctance in timely decision by top management 0.406 13 0.431 14 0.371 9
14 Harsh climatic condition at the site 0.415 14 0.395 12 0.444 16
15 Hostile political & economic environment 0.430 15 0.445 16 0.408 14
16 Tendency to pass on the blame to others 0.441 16 0.445 16 0.436 15
17 Hostile social environment 0.455 17 0.442 15 0.471 18
18 Project completion date specied but not yet planned by the owner 0.478 18 0.495 18 0.454 17
19 Uniqueness of the project activities requiring high technical know-how 0.559 19 0.589 20 0.515 19
20 Urgency emphasized by the owner while issuing tender 0.588 20 0.575 19 0.608 20
21 Size and value of the project being large 0.672 21 0.630 21 0.733 22
22 Aggressive competition at tender stage 0.724 22 0.774 23 0.652 21
23 Presence of crisis management skill of PM 0.771 23 0.730 22 0.829 23
address these characteristics only hence the name of the
factor.
7.1.2. Top management support
The top management means both the contractors
and owners top management here. From the owners
side the attributes under this factor are: understanding
operational diculties by the owner engineer thereby
taking appropriate decisions, and timely decision by
the owner or his engineer. From the contractors side
the attributes are: top managements enthusiastic sup-
port to the project manager (PM) and project team at
site, top managements backing up the plans and iden-
tify critical activities, delegating authority to project
manager by top management, selection of PM with pro-
ven track record at an early stage by top management,
Table 5
Factor prole of project success attributes (critical success factors)
Details of factor and the attributes Factor loading Variance explained
Project managers competence
Authority to take day to day decisions by the PMs team at site 0.800 22.151%
Construction control meetings 0.795
Regular budget update 0.795
Availability of resources (funds, machinery, material, etc.) as planned throughout the project duration 0.776
Project managers authority to take nancial decision, selecting key team members, etc. 0.749
Understanding of responsibilities by various project participants 0.740
Project managers technical capability 0.701
Commitment of all parties to the project 0.665
Developing and maintaining a short and informal line of communication among project team 0.644
Training the human resources in the skill demanded by the project 0.632
Leadership quality of PM 0.574
Positive attitude of PM, and project participants 0.408
Top managements support
Understanding operational diculties by the owner engineer thereby taking appropriate decisions 0.786 11.410%
Top managements enthusiastic support to the project manager (PM) and project team at site 0.751
Top managements backing up the plans and identify critical activities 0.666
Delegating authority to project manager by top management 0.592
Selection of PM with proven track record at an early stage by top management 0.500
Timely decision by the owner or his engineer (reluctance or otherwise) 0.424
Developing and maintaining a short and informal line of communication among project team 0.407
Project managers coordinating and leadership skill
a
Coordinating ability and rapport of PM with other contractors at site 0.880 10.279%
Coordinating ability and rapport of PM with owner representatives 0.780
Training the human resources in the skill demanded by the project 0.515
Leadership quality of PM 0.505
Project managers authority to take nancial decision, selecting key team members, etc. 0.453
Monitoring and feedback
Monitoring and feedback by top management 0.761 9.284%
Timely decision by the owner or his engineer (reluctance or otherwise) 0.752
Selection of PM with proven track record at an early stage by top management 0.686
Favourable political & economic environment 0.672
Monitoring and feedback by client 0.576
Coordination between project participants
Coordinating ability and rapport of PM with top management 0.851 8.257%
Coordinating ability and rapport of PM with his team members and sub-contractor 0.678
Eective monitoring and feedback by PM 0.451
Committed project participants
Ability to delegate authority to various members of his team by PM 0.639 8.128%
Positive attitude of PM, and project participants 0.607
Eective monitoring and feedback by the project team members 0.572
Eective monitoring and feedback by PM 0.471
Commitment of all parties to the project 0.432
Owners competence and favourable climatic condition
Favourable climatic condition at the site 0.832 6.162%
Monitoring and feedback by client 0.655
Scope and nature of work well dened in the tender 0.580
a
Taken with the rst factor for subsequent discussion.
developing and maintaining a short and informal line of
communication among project team. These attributes
are seen emerged in the second factor (Table 5) and
hence the name.
7.1.3. Project managers coordinating and leadership skill
The third factor shares three common attributes
with the rst factor out of ve attributes emerged.
Although the rst two attributes (Table 5) predomi-
nantly describe the coordinating and leadership skill
of Project manager, they cannot be taken as dierent
from the rst factor which covers the overall aspects
of project managers traits or competence. Hence this
factor along with the rst factor can be said to explain
the common latent property and together explain
33.43% variance (22.151% from factor 1 and 10.279%
from factor 3).
7.1.4. Monitoring and feedback
The attributes under this factor mainly focus on mon-
itoring and feedback by the project participants. The rst
and last attributes under this factor directly mention this
whereas second and third attribute under this factor
indicate action by a watchful owner and top manage-
ment, which is possible only when these participants
are monitoring the projects. As can be seen from the sec-
tion on literature review (Table 1), monitoring and feed-
back is given importance by most of the studies for the
successful outcome of a project and identied as key fac-
tor responsible for success of many projects. This factor
explains a variance of 9.28%.
7.1.5. Coordination between project participants
The high loading attributes here are coordinating
ability and rapport of PM with top management, team
Table 6
Factor prole of project failure attributes (critical failure factors)
Details of factor and the attributes Factor loading Variance explained
Conict among project participants
Poor human resource management and labour Strike 0.779 19.951%
Mismatch in capabilities of client and architect 0.752
Negative attitude of PM, and project participants 0.720
Vested interest of client representative in not getting the project completed in time 0.694
Project completion date specied but not yet planned by the owner 0.639
Conicts among team members 0.617
Conicts between PM and top management 0.595
Conicts between PM and other outside agency such as owner, sub-contractor or other contractors 0.541
Lack of understanding of operating procedure by the PM 0.486
Tendency to pass on the blame to others 0.470
Reluctance in timely decision by top management 0.443
Ignorance & lack of knowledge
Ignorance of appropriate planning tools and techniques by PM 0.839 10.786%
Reluctance in timely decision by PM 0.746
Lack of understanding of operating procedure by the PM 0.650
Conicts among team members 0.402
Presence of poor project specic attributes and non existence of cooperation
Inadequate project formulation in the beginning 0.786 9.551%
Conicts between PM and other outside agency such as owner, sub-contractor or other contractors 0.604
Tendency to pass on the blame to others 0.581
Conicts between PM and top management 0.35
Holding key decisions in abeyance 0.445
Uniqueness of the project activities requiring high technical know-how 0.428
Hostile socio economic and climatic condition
Hostile political & economic environment 0.894 8.251%
Hostile social environment 0.591
Harsh climatic condition at the site 0.440
Reluctance in timely decision
Reluctance in timely decision by top management 0.704 8.1765%
Size and value of the project being large 0.631
Presence of crisis management skill of PM 0.630
Aggressive competition at tender stage
Aggressive competition at tender stage 0.824 6.672%
Harsh climatic condition at the site 0.420
Holding key decisions in abeyance 0.540
Short bid preparation time
Urgency emphasized by the owner while issuing tender 0.783 5.650%
members, and subcontractors. Eective monitoring and
feedback by PM also makes it in this factor. These attri-
butes mainly point toward the interaction or personal
rapport with the participants. Often personal rapport
with dierent project participant can save lot of cost.
Individual rapport can lead to each other providing
helping hand and thereby sometimes reducing the neces-
sary paper work or action and paper work, going simul-
taneously thereby reducing considerable time and hence
cost nally.
7.1.6. Committed project participants
The positive attitude as well as monitoring and feed-
back by project manager and the participants show the
commitment. Project manager also shows his commit-
ment by delegating the authority to other members of
the team and not sticking with all the powers granted
to him by top management. This way PM ensures
commitment.
7.1.7. Owners competence and favourable climatic
condition
This factor explains the lowest variance (6.16%)
among all the factors through two broad aspects: own-
ers competence and climatic conditions. A competent
owner would have his scope of work well outlined and
presented to the contractor and he would closely moni-
tor his project regarding its progress, budget, quality,
and other aspects. Providing favourable climatic condi-
tion is beyond the control of owner or contractor until
one is given the choice to select his project location.
As this factor explains two diverse aspects, the name
of this factor has been kept as owners competence
and favourable climatic condition. Owners competence
has been recognised as the most important factor for
Design and Build (D&B) Projects [13].
7.2. Critical failure factors
The seven factors generated from the studies are de-
scribed below.
7.2.1. Conict among project participants
As can be seen from Table 6 this factor explains
19.95% of variance, the highest of all factors and this
contains eleven attributes with high factor loadings
(P0.4). The attributes under this factor mainly explain
either the dierence of opinion or lack of coherence in
some way barring one or two attributes. The top man-
agement must devise suitable means to avoid conict
among participants.
7.2.2. Ignorance and lack of knowledge of project
manager
This factor accounts for 10.78% of variances ex-
plained. The attributes having high loading in this factor
are: ignorance of appropriate planning tools and tech-
niques by PM; reluctance in timely decision by PM; lack
of understanding of operating procedure by the PM;
and conicts among team members. This factor is in line
with one of the established facts in literature and the
ndings of this study given in previous section. While
a competent PM becomes responsible for success of
the project, ignorance and lack of knowledge of PM
can cause failure as seen from this factor. Contracting
organisations are well advised not to compromise on
the competence of project manager. The top manage-
ment can devise means to supplement the knowledge
needs of project participants by providing training at
regular intervals.
7.2.3. Presence of poor project specic attributes and non
existence of cooperation
This factor points to two broad categories of attri-
butes. The rst category has project specic attributes
viz. Inadequate project formulation in the beginning;
uniqueness of the project activities requiring high techni-
cal know-how; holding key decisions in abeyance and
the other category has attributes related to non existence
of cooperation among project participants in the form of
conicts and passing blame. Accordingly the name of
the factor has been suggested. The factor accounts for
9.55% variance explanation.
7.2.4. Hostile socio economic and climatic condition
This factor aects the cost performance adversely in
the form of frequent stoppage of work, labour unrest,
and reduced productivity. Respondents have narrated
many projects like the famous Enron power project (In-
dia) and other projects where political views against the
project have led either to inate the schedule/cost man-
ifold or has led to shelving of the project itself. Respon-
dents also mention the current status of Tehri dam
project and Sardar Sarovar Project (India), which have
also suered on account of opposition by a section of
people resulting into severe cost and time overrun. These
two cases are internationally known for hostile socio
economic conditions.
7.2.5. Reluctance in timely decision
The name of this factor is evident, as it has been di-
rectly taken from the only attribute under this factor
that has positive factor loading. No meanings could be
assigned to parameters with negative factor loading
nor could this factor be clubbed with any other failure
factors explained in this section. This is an important
and evident factor, which are generally talked by profes-
sionals based on their experience.
7.2.6. Aggressive competition at tender stage
Although aggressive competition at the tender stage
should enhance the chances for improving the cost
performance of the project however it is not true in prac-
tice. Since most of the times such projects land up in dis-
putes arising out of petty things and claims/
counterclaims extend the duration of the project result-
ing into large cost overrun. This is why probably respon-
dents rate this as failure factor rather than a success
factor.
7.2.7. Short bid preparation time
The project duration generally includes the time from
conception/approval stage till execution and handing
over. In order to gain time for execution or unforeseen
events, owners or their representatives tend to squeeze
the bid preparation time itself. In an attempt to get
the job, contractors are unable to force the owners to
provide a reasonable time to quote for the project. Ide-
ally reasonable time should be allowed for proper site
investigation, and collection of relevant details required
for estimation purpose, etc. The short bid preparation
time leads to a number of errors/omissions on contrac-
tors part which they try to settle later through claims.
This raises disputes and nally the project lands up with
schedule and cost overruns. This factor with one attri-
bute alone explains a variance of 5.65%.
8. Most critical factor
The factor analysis discussed in the previous sections
has only grouped the like attributes in various factors
depending upon the level of correlation among them.
It does not indicate as to which factor is most inuential
in the success or failure of the projects. Accordingly the
next objective had been to identify the most important
among these factors. As explained earlier due to non-
availability of documented data for study, a need to take
up second stage questionnaire was felt. This question-
naire was aimed to get responses on the inuence of var-
ious factors on the actual outcome of the project. Thus
using the above success and failure factors as inuencing
variables, questions were developed to know their actual
impact on 5, . . . ,0, . . . ,+5 (11 point scale, with 5 indi-
cating most negative eect and +5 indicating most posi-
tive eect) on each of the project performance criteria
such as schedule, cost, quality, and no-dispute. The
respondents are however asked to select only one project
of their choice where they have in-depth knowledge of
the particular project and base all their responses on
that project. This project is named as Choice project
of the respondent. The respondents are also asked to
rate the actual achievement of project performance in
terms of schedule, cost, quality, and no-dispute in the
choice project in a scale of 110 where 1 represents
highly failure with minimum score and 10 represents
most successful with maximum score. Since the scope
of this paper is restricted only to discussion on cost per-
formance of a project that is inuenced by various attri-
butes, ndings on other project performances are not
discussed.
A total of 90 responses are received as against 300
questionnaires sent to randomly selected professionals
from top contracting organisations of the country. The
mean score of cost performance is found to be 8.011,
which indicates that respondents have chosen only those
projects that have fared well on cost account. The rea-
son respondents are given the choice of self-selection
of projects is due to lack of availability of the docu-
mented data on performance of Indian construction
projects. It is also recognized that there could be bias
in the self-selection of the choice project by the respon-
dents as many of them have chosen to share their suc-
cessful experiences only. Lim and Ling [20] point out
the leniency bias in self-selection while Somers and Birn-
baum [21] assert that self-selection is generally accurate.
To nd the most important factor among the critical
success and failure factors stepwise regression technique
has been used with the factors as causal variables and
the cost performance as the dependent variable. The
regression results are summarized in Table 7. The regres-
sion analysis results are generally used helps to develop
predicting or forecasting models provided the R
2
value is
reasonably high and the model helps in assessing the or-
der of importance of each of the criterion variables. In
the present case since the R
2
value is not too high, the
relationship obtained from the regression equations
are not being used to develop a predicting model but re-
stricted to ascertain the most important factor through
the b coecient. It can be seen from the Table 7 that
coordination between project participants is the most
important factor that has maximum inuence in success-
ful cost performance. The research concludes that
although other factors obtained from the study are also
important, the major contribution comes from coordi-
nation. Similar conclusion has also been drawn by
Rad [22] on the case study of construction of a nuclear
power plant. It can be seen that with proper coordina-
tion among project participants many avenues of cost
savings are opened up. Respondents, in the open-ended
question of the questionnaire also point out that lack of
coordination among participants results in the duplicity
of work and resources are also underutilized resulting
into wasteful expenditures.
While the technical skills and some human skills can
be improved by designing the suitable course content at
undergraduate and post graduate levels, training to im-
prove other skills like proper interaction and coordina-
tion may be dicult to impart in the college level
when one does not have adequate exposure of project
work. Hence such training should be taken up at work-
ing level. Having regular meetings among the project
participants and developing a rapport among themselves
can also improve these skills. With the increasing project
size and increase in number of project participants this
particular factor becomes all the more important.
9. Summary and conclusions
Questionnaire survey on an extensive project attri-
butes that aect the cost performance of project has re-
vealed the important success and failure attributes. The
important success attributes are: eective monitoring
and feedback by the project manager and project team
members; coordinating ability and rapport of PM with
top management; positive attitude of PM, and project
participants; and Project managers technical capability.
The important failure attributes are poor human resource
management and labour strike; negative attitude of PM
and project participants; inadequate project formulation
in the beginning; vested interest of client representative in
not getting the project completed in time; and conicts
between PM and top management. There is high agree-
ment in the ranking of critical failure attributes by the
two groups of respondents viz. contractors and owners.
Factor analysis of responses on the project attributes
has extracted critical success and failure factors. Project
managers competence; top management support; pro-
ject managers coordinating and leadership skill; top
management and owner involvement in the project;
interaction between project participants, monitoring
and feedback by project participants; owners compe-
tence and favourable climatic condition are the seven
critical success factors. Similarly critical failure factors
extracted are: conict among project participants; igno-
rance and lack of knowledge; indecisiveness; hostile so-
cio economic and climatic condition; reluctance in
timely decision; aggressive competition at tender stage;
short bid preparation time.
Identication of critical success as well as failure
attributes and factors by this study also leads one to rea-
lise that it is not sucient to only maximise the results of
the critical success factors but it is also necessary to min-
imise the negative impact of failure factors.
The most important factor among all success and
failure factors turns out to be coordination among pro-
ject participants which is obtained from the results of
regression analysis. Skills like proper interaction and
coordination need to be taken up through suitable train-
ing programmes of project participants. Having regular
meetings among the project participants and developing
a rapport among themselves can also improve these
skills. With the increasing project size and increase in
number of project participants this particular factor be-
comes all the more important.
As can be seen the success factors obtained from the
present analyses are consistent with the ndings in the
context of developed countries, a summary of which is
presented in Table 1. Thus the study concludes that
the critical success factor remains the same irrespective
of geographical boundaries.
Table 7
Summary of stepwise regression results
Unstandardized
coecients
Standardized coecients t a R
2
B Standard error b
Constant 6.372 0.380 16.773 0.000 0.210
Coordination between project participants 0.463 0.100 0.459 4.616 0.000
Appendix A. Questionnaire on project attributes
Listed below are some of the attributes responsible for advantages/hindrances to project success; please indicate the
eects of these attributes on various project success evaluation criteria given alongside the attributes
S. No. Project success attributes Eect on project cost
1 Size and value of the project being large 1 2 3 4 5
2 Scope and nature of work well dened in the tender 1 2 3 4 5
3 Aggressive competition at tender stage 1 2 3 4 5
4 Urgency emphasized by the owner while issuing tender 1 2 3 4 5
5 Inadequate project formulation in the beginning 1 2 3 4 5
6 Uniqueness of the project activities requiring high technical know-how 1 2 3 4 5
7 Favourable political & economic environment 1 2 3 4 5
8 Hostile political & economic environment 1 2 3 4 5
9 Hostile social environment 1 2 3 4 5
10 Favourable social environment 1 2 3 4 5
11 Harsh climatic condition at the site 1 2 3 4 5
(continued on next page)
Appendix A (continued)
S. No. Project success attributes Eect on project cost
12 Favourable climatic condition at the site 1 2 3 4 5
13 Project completion date specied but not yet planned by the owner 1 2 3 4 5
14 Monitoring and feedback by client 1 2 3 4
15 Timely decision by the owner or his engineer (reluctance or otherwise) 1 2 3 4 5
16 Understanding operational diculties by the owner engineer thereby taking
appropriate decisions
1 2 3 4 5
17 Top managements enthusiastic support to the project manager (PM) and project
team at site
1 2 3 4 5
18 Top managements backing up the plans and identify critical activities 1 2 3 4 5
19 Selection of PM with proven track record at an early stage by top management 1 2 3 4 5
20 Delegating authority to project manager by top management 1 2 3 4 5
21 Monitoring and feedback by top management 1 2 3 4 5
22 Reluctance in timely decision by top management 1 2 3 4 5
23 Eective monitoring and feedback by PM 1 2 3 4 5
24 Eective monitoring and feedback by the project team members 1 2 3 4 5
25 Lack of understanding of operating procedure by the PM 1 2 3 4 5
26 Ignorance of appropriate planning tools and techniques by PM 1 2 3 4 5
27 Reluctance in timely decision by PM 1 2 3 4 5
28 Ability to delegate authority to various members of his team by PM 1 2 3 4 5
29 Coordinating ability and rapport of PM with his team members and sub-contractor 1 2 3 4 5
30 Coordinating ability and rapport of PM with top management 1 2 3 4 5
31 Coordinating ability and rapport of PM with owner representatives 1 2 3 4 5
32 Coordinating ability and rapport of PM with other contractors at site 1 2 3 4 5
33 Leadership quality of PM 1 2 3 4 5
34 Project managers authority to take nancial decision, selecting key team members,
etc.
1 2 3 4 5
35 Project managers technical capability 1 2 3 4 5
36 Construction control meetings 1 2 3 4 5
37 Regular budget update 1 2 3 4 5
38 Commitment of all parties to the project 1 2 3 4 5
39 Understanding of responsibilities by various project participants 1 2 3 4 5
40 Authority to take day to day decisions by the PMs team at site 1 2 3 4 5
41 Conicts among team members 1 2 3 4 5
42 Conicts between PM and top management 1 2 3 4 5
43 Conicts between PM and other outside agency such as owner, sub-contractor or
other contractors
1 2 3 4 5
44 Tendency to pass on the blame to others 1 2 3 4 5
45 Availability of resources (funds, machinery, material, etc.) as planned throughout
the project duration
1 2 3 4 5
46 Developing and maintaining a short and informal line of communication among
project team
1 2 3 4 5
47 Poor human resource management and labour strike 1 2 3 4 5
48 Presence of crisis management skill of PM 1 2 3 4 5
49 Vested interest of client representative in not getting the project completed in time 1 2 3 4 5
50 Training the human resources in the skill demanded by the project 1 2 3 4 5
51 Mismatch in capabilities of client and architect 1 2 3 4 5
52 The capability of project participants to market the end product to the intended
users
1 2 3 4 5
53 Positive attitude of PM, and project participants 1 2 3 4 5
54 Negative attitude of PM, and project participants 1 2 3 4 5
55 Holding key decisions in abeyance 1 2 3 4 5
1, Adversely aect; 2, signicantly aect; 3, marginally aect; 4, no eect; 5, helps in saving.
References
[1] Rubin IM, Seeling W. Experience as a factor in the selection and
performance of project managers. IEEE Trans Eng Manage
1967;14(3):1314.
[2] Avots I. Why does project management fail?. California Manage
Rev 1969;12(1):7782.
[3] Sayles LR, Chandler MK. Managing large systems. New
York: Harper and Row; 1971.
[4] Martin CC. Project management. New York: Amaco; 1976.
[5] Baker BN, Murphy DC, Fisher D. Factors aecting project
success. Project management handbook. New York: Van Nos-
trand Reinhold; 1983. p. 66985.
[6] Cleland DI, King WR. Systems analysis and project manage-
ment. New York: McGraw-Hill; 1983.
[7] Locke D. Project management. NewYork: St. Martins Press; 1976.
[8] Hughes MW. Why projects fail: the eects of ignoring the
obvious. Ind Eng 1986;18:148.
[9] Morris PW, Hough GH. The anatomy of major projects. New
York: Wiley; 1987.
[10] Schultz RL, Slevin DP, Pinto JK. Strategy and tactics in a process
model of project implementation. Interfaces 1987;17(3):3446.
[11] Pinto JK, Slevin DP. Critical success factors in R&D projects. Res
Technol Manage 1989(JanuaryFebruary):315.
[12] Chua DKH, Kog YC, Loh PK. Critical success factors for dierent
project objectives. J Construct Eng Manage 1999;125(3):14250.
[13] Chan APC, Ho DCK, Tam CM. Design and build project success
factors: multivariate analysis. J Construct Eng Manage
2001;127(2):93100.
[14] Chan DWM, Kumaraswamy MM. A comparative study of causes
of time overruns in Hong Kong construction projects. Int J
Project Manage 1997;15(1):5563.
[15] Overall JE, Klett CJ. Applied multivariate analysis. New
York: McGraw-Hill Book Company, Inc; 1972.
[16] Dillon WR. Goldstein multivariate analysis. New York: Wiley;
1984.
[17] Child D. The essentials of factor analysis. London: Cassell
Educational Limited; 1990.
[18] Iyer KC. Identication and evaluation of dispute-prone clauses in
Indian construction contracts. A Doctoral Thesis, Indian Institute
of Technology, Madras; 1996.
[19] Trost SM, Oberlender GD. Predicting accuracy of early cost
estimates using factor analysis and multivariate regression. J
Construct Eng Manage 2003;129(2):198204.
[20] Lim EH, Ling FYY. Model for predicting Clients contribution to
project success. Eng Construct Architect Manage 2002;9(5/6):
38895.
[21] Somer MJ, Birnbaum D. Assessing self-appraisal of job perfor-
mance as an evaluation device: are the poor results a function of
method or methodology. Human Resource 1991;44:108191.
[22] Rad PF. Delays in construction of nuclear power plant. J Energ
Div, Proc ASCE 1979;105(EY/1):3346.
Cost estimation of a software product using
COCOMO II.2000 model a case study
R. Dillibabu
*
, K. Krishnaiah
Department of Industrial Engineering, College of Engineering, Anna University, Guindy, Chennai 600025, Tamilnadu, India
Received 25 May 2004; received in revised form 25 June 2004; accepted 10 November 2004
Abstract
This paper discusses the estimation cost in terms of eort spent on a software product (project). COCOMO II.2000 Model has
been employed for estimating the eort of an embedded system project. This study has been made in a software services company,
which is involved in software development for an embedded system, client-server and Internet environment. The embedded systems
group is involved in developing software for major car manufacturers. This study is based on a sample of ten projects, of which eight
are development projects and two are porting projects. The actual eort on the projects has been collected from the metrics database
of the company. The Lines of code for the various projects have been enumerated using Code Count tool to achieve the logical
source lines of code for each project. The standard questionnaire has been used to collect the required data to arrive at the various
scale factors and eort multipliers. The calibration of the COCOMO (Constructive Cost Model) has been done through Natural
Log approach and curve tting approach. Statistical tools like MS-EXCEL 2000, SPSSv10, Curve Expert 1.3 and data t have been
used for this purpose. The study shows that the curve tting approach yields better estimates of the model parameters. The calibra-
tion of COCOMO Model helps the company estimate the eort that is to be spent on the software development projects.
Keywords: Embedded systems; Calibration; Scale factors and eort multipliers
1. Introduction
According to Gartner Report [20], one of the main
challenges the Indian software industry is facing, espe-
cially in the aftermath of sept 11th is the intense down-
ward pressure on pricing and oshore billing rates for
low-end work (e.g., maintenance and coding) have
dropped from $80 to $40 per hour to $12 to $20
for smaller vendors. This has forced the companies to
look into their practice of estimating the cost for soft-
ware development. Until now, most companies have re-
lied on experience and Price-to-win strategies for
getting past competitors to win projects. With the emer-
gence of concepts like CMM (Capability Maturity
Model) and other initiatives, the software companies
adopted formal techniques like Work Breakdown Struc-
ture (WBS) based cost estimation and parametric esti-
mation. One can never rely completely on experience-
based estimation in the software industry because of
the rapidly changing technology, which renders the
experience-based estimates ineective. Further, price-
to-win strategy is not very favorable for any company
in the long run. Hence, there is a need to come up with
a suitable cost model to account for the eort spent on
developing software, right from requirements specica-
tion to delivery and maintenance.
1.1. Software cost estimation and project management
As software industry is very competitive, it is indis-
pensable, to establish the market with the right pricing.
doi:10.1016/j.ijproman.2004.11.003
*
Corresponding author. Tel.: +91 44 22203187.
E-mail address: prdillibabu@redimail.com (R. Dillibabu).
PROJECT
MANAGEMENT
This makes software cost estimation as one of the very
important issues in the software development process.
For a software project, estimation comprises size, eort,
and schedule estimate. Size estimate is the measure of
the size of the nal work product. It gives a measure
of the ground to be covered in order to achieve the
end result. Eort estimate is the eort, in person-
months, to produce the work product. Schedule estimate
is the calendar time it would take to accomplish the ef-
fort, taking into account the resource constraints of an
organization and the extent of parallelism that can be
derived. Eventually, both the customer and the organi-
sation are interested in the cost of the project in its dol-
lar value. Cost estimate [18] in dollar value is given by
Cost estimate No: of person months
Cost per person-months of effort$:
The size, eort and schedule estimates get translated
into cost estimates. The purpose of going through the
size, eort and schedule estimates is to analyse factors
that contribute to the overall software development cost
in the nal dollar value of the estimate.
1.2. Software cost estimation in embedded system
This paper discusses the software cost estimation in
terms of eort spent for the embedded systems group
of a company. Initially the existing system for software
cost estimation has been studied and the following
observations are made:
(i) The projects in the embedded system group are of
short duration.
(ii) The projects are up to 4K in size.
(iii) More number of completed projects is available for
the study.
(iv) The waterfall model has been used in these projects.
(v) The present eort (cost) estimation procedure lacks
scientic approach.
Based on these observations, the following objectives
are listed:
1. To study the COCOMO model and its increments.
2. To come up with suitable estimates for the scale driv-
ers and eort multipliers for the COCMO II.2000
model using data from industries.
3. To t COCOMO II.2000 Model for practical data in
a software company.
4. To evaluate the model using Contes criteria [5].
2. Literature
From literature it was found that the following pa-
pers are more relevant for software cost estimation.
Software cost estimation by Danfeng Hang [6] and
Estimating software cost by Lawrence H Putnam
and Ann Fitzsimmons [17]. The authors have reviewed
the literature and presented the major software cost esti-
mation models with their strengths and weaknesses. The
abstract from these two papers has been presented in
Table 1.
The following conclusions can be drawn from the re-
view of literature:
(i) None of the alternatives is better than the others in
all aspects.
(ii) The Parkinson and Price-to-win methods are not
widely accepted and do not produce satisfactory
cost estimates.
(iii) The strength and weakness of other techniques are
complementary (particularly the algorithmic model
versus expert judgment and top-down versus bot-
tom-up).
Since most of the researchers have used the algorith-
mic model, it has been selected for applying in this
study. Algorithmic models are the most formal amongst
Table 1
Abstract from review papers on software cost estimation models
Models and authors Strengths Weakness
1. Algorithmic Model, Putnam [23] Objective, repeatable, analogy formula,
ecient, good for sensitivity analysis &
objectively calibrated to experience
Subjective inputs, calibrated to past, not the future &
assessment of exceptional circumstances
2. Expert Judgment Delphi Techniques [12] Assessment of representativeness,
interactions, exceptional circumstances
No better than participants, biases & incomplete recall
3. Analogy Norden, Peter [16] Based on representative experience Representativeness of experience
4. Parkinson, Parkinson [17] Correlates to some experience Reinforces poor practice
5. Price-to-win, Boehm [4] Often gets the contract Generally produces large overruns
6. Top-down, Albrecht and Ganey [1] System level focus and ecient Less detailed basis. Fewer stables
7. Bottom-up, Albrecht and Ganey [1] More detailed basis and more stable May overlook system level cost & requires more eort
8. Dynamic Modeling, Putnam [23] More detailed basis More time and data required for validating &
developing a model
298 R. Dillibabu, K. Krishnaiah / International Journal of Project Management 23 (2005) 297307
the various techniques. Fenton and Peeger [8] describes
two types of algorithmic model:
(a) Cost Models: These models provide direct estimates
of eort. Most cost models are based on empirical
data reecting factors that contribute to the overall
cost. Often, they have a primary cost factor such as
size and a number of secondary adjustment factors
or cost drivers.
(b) Constraint Models: Constraint models demonstrate
the relationship over time between two or more
parameters of eort, duration or stang levels.
The Rayleigh curve [14] is used as a constraint in
several commercial proprietary models like Put-
nams SLIM [17] and RCA Price-S [13]. Sucient
material is not available for evaluating these
models.
2.1. Function points
(i) Function point [7] counts do not depend on the
source languages used. It can be obtained early in
the development cycle.
(ii) Function Point counts are oriented toward the cus-
tomers view rather than the producers view, this
emphasis the focus on value received, rather than
on the particular design that is employed.
(iii) Function point counts are not equally applicable to
all kinds of software. Although eective in business
environments, they have not enjoyed widespread
success in embedded systems or heavily computa-
tional applications.
3. Current estimation process
The Case Company has an elaborate metrics system
to capture data from the projects executed. This is
summarized in the form of Project Managers (PM)
workbook and Software Quality Analysts (SQA)
Workbook. The PM workbook concentrates on captur-
ing the various metrics for various milestones of the
projects, like project start date, end date, modules
included, estimated eort, actual eort, LOC, etc. The
details available in this report enable the reader to visu-
alize the project. The SQA Workbook contains details
regarding the quality related activities like number of
defects, severity of defect, defect injection, defect detec-
tion, etc.
The cost (eort) is estimated using combination of
function points and WBS. The process of estimation be-
gins once the Request For Purpose (RFP) and the scope
change are nalised. The estimation process uses two
models for the purpose of size estimation:
(i) System Software Sizing (SSS).
(ii) Business Application Software Sizing (BASS).
Depending on the nature of the project, i.e., a sys-
tems project or a business application project, the rel-
evant model is chosen for the purpose of estimation.
The function points are counted based on the stan-
dardized template for SSS/BASS. The factors that
inuence the general system characteristics are consid-
ered to arrive at adjusted function points. This forms
the basis for arriving at the suitable eort estimate.
The process adopted is as proposed by the Interna-
tional Function Point Users Group (IFPUG [10])
for BASS and is a variant of the IFPUG process
for SSS. The WBS for the project is designed simul-
taneously once the RFPs and scope change docu-
ments are nalized. An estimate of eort based on
the WBS is made, which is then crosschecked with
the function point [7] based estimate. The estimation
group does this review, which also gives correction
factors to the assumptions made. The estimate is
nalized once the corrections are made to the
assumptions, in which a revised WBS is designed.
This revised WBS forms the basis for the scheduling
and phase wise breakdown of eort. These details
go into schedule document and the project proposal.
Once the proposal is accepted, the project managers
workbook and the SQA workbook become active.
These are updated as and when each phase or module
is completed.
3.1. Weakness in the current estimation process
The WBS based approach towards eort estimation
has many drawbacks. Some of them are listed below:
(i) The WBS based estimate involves experience-based
estimates at the module/phase level. This estimate
can be biased. The nature of the bias may range
from Availability bias, Representative bias, Over-
condence bias, Precedence bias, etc. [15].
(ii) Even though WBS biased estimate may be easier, it
is based on the assumption that the person doing
the estimate has had adequate domain experience
in the company and he/she will remain in the com-
pany forever. These assumptions can change at any
point of time.
Hence, there is a need for an estimating process,
which is predictable and not dependent on a single indi-
vidual. Further, it should be easy to use and reproduc-
ible. The COCOMO II.2000 model integrated with
function points will meet these requirements. Further,
this model can be calibrated for other development envi-
ronments in the company.
R. Dillibabu, K. Krishnaiah / International Journal of Project Management 23 (2005) 297307 299
4. The COCOMO II.2000 model
The model is dened in terms of scale factors (SF
j
)
and eort multipliers (EM
i
), which are used for estimat-
ing eort (cost) that is required to develop software pro-
ject. The response obtained from Project Managers
through a standard questionnaire is made use of for esti-
mating the parameters of the model.
4.1. Overview of the COCOMO II.2000 model
The Constructive Cost Model (COCOMO II [2]) for
cost estimation is based on three major stages of any
development project. It acknowledges the fact that the
lines of code are impossible to know early in the devel-
opment cycle. In stage I, the project usually builds pro-
totypes to resolve high-risk issues that involve user
interfaces, software and system interaction, perfor-
mance, or technical maturity. In this stage, COCOMO
II estimates size using object points. This technique cap-
tures size in terms of high-level eort, generates such as
number of server data tables, number of client data ta-
bles and the percentage of screens and reports reused
from previous projects.
In stage 2, COCOMO II employs function points as a
size measure. Function Points estimate the functionality
captured in the requirements, thereby oering a richer
system description than object points. By stage 3, devel-
opment would have begun and more information would
be made available. Sizing is done in terms of lines of
code and many cost factors can be estimated with some
degree of comfort.
4.2. Sizing and counting source lines of code (SLOC)
A good size estimate is very important for any good
model estimation. COCOMO II model uses size data
that inuences eort, i.e., new code and code that is re-
used and modied. All the projects analyzed in this
study are considered to have new code, since the projects
have very negligible or no reused code.
For new and reused code, a method is used to
make them equivalent so that they can be rolled up
into an aggregate size estimate. The baseline size in
COCOMO II is a count of new lines of code. The
adjustment takes into account the amount of design,
code and testing that has to be changed. It also con-
siders the understandability of the code and the pro-
grammer familiarity with the code. Code size is
expressed in thousands of source lines of code
(KSLOC). The goal is to measure the amount of intel-
lectual work that is put into program development.
Dening a line of code is dicult because of concep-
tual dierences involved in accounting for executable
statements and data declarations in dierent lan-
guages. Diculties arise while trying to dene consis-
tent measures across dierent programming languages.
In COCOMO II, the logical source statement has been
chosen as standard line of code.
4.3. Eort estimation
The COCOMO II eort estimation model is shown in
Eq. (1). This model is used for both the Early Design
and Post-Architecture cost models to estimate eort
for the Waterfall lifecycle models. The inputs are the
Size of software development, a constant A, an expo-
nent E, and a number of values called eort multipliers
(EM). The number of eort multipliers depends on the
model
PM
ns
A Size
E
P
n
i1
EM
i
;
where ns nominal schedule & A 2:94
for COCOMO II:2000: 1
The constant, A, approximates a productivity constant
in PM/KSLOC for the case where PM = Person Months
& E = 1.0 Productivity changes as E increases because
of the non-linear eorts on size. The constant A is ini-
tially set when the model is calibrated to the project
database reecting a global productivity average. The
COCOMO model should be calibrated to local data,
which then reects the local productivity and improves
the models accuracy.
4.4. Scale drivers
The scale drivers in the exponent, E, are used only at
the project level. The exponent E in Eq. (2) is an aggre-
gation of ve scale drivers that account for the relative
economies or diseconomies of scale encountered for
software projects of dierent sizes. Software projects
generally exhibit diseconomies of scale because of two
main factors:
1. Growth of interpersonal communication overhead.
2. Growth of large-system integration overhead.
Eq. (2) denes the exponent, E, used in Eq. (1).
Each scale driver has a range of rating levels, from
very low to extra high. Each rating level has a weight.
The specic value of the weight is called a scale factor
(SF). The projects scale factors, the selected scale dri-
ver ratings, are summed and used to determine a scale
exponent, E
E B 0:01
X
SF
j
;
where B 0:91 for COCOMO II:2000: 2
5. Methodology, tting and calibration to the local
environment
5.1. Methodology
To achieve the objectives proposed, the following
methodology is proposed. The COCOMO II software
cost and schedule model will be used to estimate the eort
of projects. These estimates will then be compared with
actual values of eort. The results will be analysed to
determine the overall eectiveness of the model. To aid
in understanding the process, the step-by-step description
of the proposed calibration of the model is described.
5.2. Model calibration by natural log approach
COCOMO II will be calibrated using the COCOMO
Equation (Eq. (3)). The number of data points consid-
ered is ten. In this approach, only the multiplicative cal-
ibration variable A will be calibrated
PM
ns
A Size
E
P
n
i1
Em
i
where E B 0:01
X
s
j1
SF
j
: 3
5.3. Analysis by curve tting
Most software cost models can be abstracted into a
function of ve basic parameters: size, process, person-
nel, environment and required quality [19]. The relation
among these parameters and the estimate eort can be
written as follows:
Effort PersonnelEnvironmentQualitySize
process
:
4
Eq. (4) describes a power equation of the form
Y L X
M
; 5
where Y = eort, X = size and L and M are constants.
When compared with the COCOMO II.2000 model
equation (Eq. (3)), we arrive at the following:
L A
X
EM
i
and M B 0:01
X
SF
j
: 6
By tting a power curve between the Logical SLOC
and the Actual Eort, by using Eqs. (5) and (6), a bet-
ter calibrated COCOMO II model can be arrived at.
These values of A and B will help in analysing the
software development process in the company in a
better manner.
5.4. Validation of methodology
The validation of the above analysis is done using
Contes Criteria [5], to determine the accuracy of the cal-
ibrated and uncalibrated model. This will be achieved
using Eqs. (7)(11).
5.5. Contes criteria
Here the Magnitude of Relative Error (MRE) is com-
puted (degree of estimating error in an individual esti-
mate) for each data point. This step is a precedent to
the next step and is also used to calculate PRED (e).
An MRE of 25% or less indicates satisfactory results
MRE jEstimate Actual=Actualj: 7
Calculate the mean magnitude of relative error (average
degree of estimating error in a data set) for each data
set. According to Conte, the MMRE should also have
a value of 25% or less
MMRE RMRE=n; 8
where n = total number of estimates.
Calculate the root mean square (Models ability to
accurately forecast the individual actual eort) for each
data set. This step is a precedent to the next step only.
Again, satisfactory results are indicated by a value of
25% or less
RMS 1=n REstimate Actual
2
0:5
: 9
Calculate the relative root mean square (models abil-
ity to accurately forecast the average actual eort) for
each data set. According to Conte, the RRMS should
have a value of 25% or less
RRMS RMS=RActual=n: 10
A model should also be within 25% accuracy, 75% of the
time. To nd this accuracy rate PRED (e), divide the to-
tal number of points within a data set that has an
MRE = 0.25 or less (represented by n). The equation
then is
PRED e k=n; 11
where e equals 0.25.
This methodology seems to be sound since it is based
on proven and accepted mathematical and statistical
analysis.
5.6. Fitting of the COCOMO model
The following steps are used in tting the model to
the practical data applicable for a software service
company:
1. Mining and normalizing of project data from the pro-
ject database.
2. Counting the lines of code.
3. Arriving at values for scale factors and eort
multipliers.
4. Formulation of the uncalibrated COCOMO Eort
Equation.
5. Calibrating the values of multiplicative constants to
arrive at a customised calibrated COCOMO Eort
Equation.
6. Validation of the calibrated and the uncalibrated
COCOMO equation.
5.7. Mining and normalizing of project data from the
project database
Data mining from 10 Embedded System Projects (8
Development Projects and 2 Porting Projects) are under-
taken in two phases. In the rst phase, preliminary data
from SQA workbook and PM workbook was explored
for each of the 10 projects and normalized. This pro-
vided an insight into various modules in a particular
project and details like the planned eort, actual eort,
size (module-wise). Other data about the projects are
collected into a Master Table Form. This master table
facilitates in nding data that is missed from the two
workbooks.
In the second phase, data for each of the project was
further explored and missing data was mined from the
various databases in the company, with the aid of this
master table. The data was normalized because of the
huge dierences in conguration management across
various projects.
5.8. Counting the lines of code
The lines of code for all projects are available in the
metrics database in the company in the form of physical
LOC. It is found that the company has used a code
counting utility called Measure. This utility followed
certain rules for counting, which are not in conformance
with the rules specied by the authors of the COCOMO
model.
For this purpose, the authors of the model have
also developed Code Count, a utility for counting
the lines of code as per the requirements of the CO-
COMO suite of models. This utility has been down-
loaded [21] and applied to all the projects (Table 2).
Shows the LOC count.
5.9. Scale factors and eort multipliers
The scale factors and eort multiplier have to be ob-
tained for each of the 10 projects. Due to limitation of
resources, it has been decided to administer a question-
naire based on various criteria such as spreadsheet soft-
ware package.
The responses have been analyzed using spreadsheet
software and the results arrived at are listed in Tables
35.
5.10. Formulation of the COCOMO equation
Based on these responses, the original COCOMO II
equation (Eq. (1)) is transformed as shown below
Effort 49:12 Size
1:08
: 12
The estimate of the eort for the 10 projects, based
on the above equation is shown in Table 6. It is seen that
the error between the estimated eort and actual eort is
very high for porting projects. This may introduce a very
serious error during the process of calibration. Hence,
the porting projects are not considered for the purpose
of calibration and calibration is done using eight devel-
opment projects only.
5.11. Calibration of the COCOMO equation
5.11.1. Natural log approach
The generic COCOMO II model discussed previously
has been used to estimate software development eorts
for a variety of project types. However, this model can
be tailored to a particular organisation or project do-
main to arrive at more accurate estimates. The COCO-
MO II Post-Architecture model is signicantly more
accurate when calibrated to an organization. All that
needs to be done is to calibrate the constants, A and/
or B, in the eort estimation equation. The intent of
calibration is to take the productivity and activity distri-
butions of the local development environment into
account.
Local calibration improves predication accuracies
because:
The rating scale is subjective, leading to inconsisten-
cies across dierent organizations.
The lifecycle activities as covered by COCOMO II
may be slightly dierent from the lifecycle activities
as covered by the company.
The denitions as used by COCOMO II may dier
from those being used by the company. For, e.g., CO-
COMO II denes 1 person-month (PM) as 152 per-
Table 2
LOC count using Code Count tool in KLOC
Project name Project type KLOC
PROJ1 Development 12.58
PROJ9 Porting 33.568
PROJ10 Porting 38.861
son-hours. But the company evaluates one person-
month as 176 person-hours (8 h/day
*
22 days/month).
The simplest way to account for these variations is to
calibrate the multiplicative constants, A and B.
First, the calibration is done using the Natural Log
Approach as described by Boehm et al. [4]. This is done
on the multiplicative constant A using data from com-
pleted projects. The authors of the model have recom-
mended at least 5 data point for the calibration of the
constant A where as here data from 8 projects have
been used (Table 7). Shows the outcome of this exercise.
The value of the multiplicative constant A is obtained
from Table 7. Its value is 2.79. The estimated eort for
the 8 projects is shown in Table 8.
5.11.2. Curve tting approach
Another approach adopted for the calibration of the
constants is the curve tting method. Power curves are
tted with the help of the following statistical tools using
Eqs. (13) and (14):
Excel 2000 from Microsoft.
SPSS v10 from SPSS Corporation.
Datat 8.0 from Oakdale Engineering.
Curve Expert 1.3.
It is found that the curve t given by Excel and SPSS
explains the dependency of actual eort on the LOC to
the extent of 43%. The other statistical tools have given
Table 4
Eort multipliers (product and platform factors)
RELY
a
DATA
b
CPLX
c
RUSE
d
DOCU
e
TIME
f
STOR
g
PVOL
h
0.82 0.90 0.87 1.00 1.00 1.11 1.00 0.87
1.00 0.90 0.87 1.00 1.00 1.11 1.05 0.87
0.82 0.90 0.87 1.00 1.00 1.11 1.00 0.87
0.82 0.90 0.87 1.00 1.00 1.11 1.00 0.87
0.82 0.90 0.87 1.00 1.00 1.11 1.00 0.87
0.82 0.90 0.87 1.00 1.00 1.11 1.00 0.87
0.82 0.90 0.87 1.00 1.00 1.11 1.00 0.87
1.00 1.00 0.87 1.15 1.00 1.63 1.00 0.87
0.92 0.90 1.74 1.24 1.00 1.63 1.00 0.87
0.82 0.90 1.74 0.95 1.00 1.00 1.00 0.87
a
Required software reliability.
b
Database size.
c
Product complexity.
d
Developed for reusability.
e
Documentation match to life-cycle needs.
f
Execution time constraint.
g
Major storage constraint.
h
Platform volatility.
Table 3
Scale factors
Project name Project type PREC
a
FLEX
b
RESL
c
TEAM
d
PMAT
e
PROJ1 Development 2.48 4.05 5.65 3.29 1.56
PROJ9 Porting 4.96 5.07 2.83 0.00 3.12
PROJ10 Porting 0.00 5.07 2.83 2.19 7.8
a
Precedentedness.
b
Development exibility.
c
Architecture/risk resolution.
d
Team.
e
SW-CMM level 15.
an equation with the coecient of determination as
36%. And the error between the estimated eort and
the actual eort is same for both the equations. Table
9. Shows the resulted equations.
5.12. Validation of the calibrated models
Tables 10 and 11 shows the eort estimation using
the Eqs. (13) and (14). Studies have shown that soft-
ware development activity shows diseconomy of
scale, when a relation is considered between the soft-
ware size and development eort [19]. The equation
arrived at by the tools Data t and Curve Expert
shows economies of scale, whereas the equation given
by Excel and SPSS shows diseconomies of scale.
Hence, the equation given by Excel and SPSS (Eq.
(12)) is taken as the outcome of the curve tting
approach.
Table 6
Eort estimation with original COCOMO Equation
Project name Project type KLOC Actual eort Actual eort (PM) COCOMO eort (PM)
PROJ1 Development 12.58 249.56 11.34375 12.37
PROJ9 Porting 33.568 762.73 34.669318 7.29
PROJ10 Porting 38.861 208.63 9.4829545 36.19
Table 7
Calibration using the natural log approach
Project name Project type PM
Actual
KSLOC Unadjusted estimate Ln(PM
Actual
) Ln(UE) Ln(Di)
PROJ1 Development 249.56 12.58 757.37 5.52 6.63 1.11
Average = 1.03
Eort = 2.79
Table 5
Eort multipliers (personnel and product factors)
ACAP
a
PCAP
b
PCON
c
APEX
d
LTEX
e
PLEX
f
TOOL
g
SITE
h
SCED
i
0.85 0.88 1.29 1.00 1.00 1.09 1.17 0.86 1.00
0.71 0.76 1.00 0.88 0.91 1.00 0.90 1.00 1.43
0.85 0.88 1.29 1.00 1.00 1.09 1.17 0.86 1.00
0.85 0.88 1.29 1.00 1.00 1.09 1.17 0.86 1.00
0.85 0.88 1.29 1.00 1.00 1.09 1.17 0.86 1.00
0.85 0.88 1.29 1.00 1.00 1.09 1.17 0.86 1.00
0.85 0.88 1.29 1.00 1.00 1.09 1.17 0.86 1.00
0.85 0.88 1.00 1.00 1.00 1.00 0.90 1.00 1.00
0.71 0.76 1.00 0.88 0.84 0.91 0.78 1.09 1.43
1.00 0.88 1.00 1.00 1.00 1.00 1.17 0.86 1.14
a
Analyst capability.
b
Programmer capability.
c
Personnel continuity.
d
Applications experience.
e
Language and tool experience.
f
Platform experience.
g
Use of software tools.
h
Multi-site development.
i
Required development schedule.
Therefore, Eq. (13) is the nal COCOMO II.2000
equation for the purpose of cost (eort) estimation in
the company:
Effort 0:93Size
1:0197
Reproduced: 13
This model has been validated using Contes Criteria as
described already.
Outcome of model validation:
(i) MMRE = 0.57,
(ii) RRMS = 0.20,
(iii) Pred(0.25) = 0.38,
(iv) Pred(0.5) = 0.63.
It is found that:
(i) Of the three criteria, this model satises only one,
viz., the relative root mean square of the error
(RRMS), which is less than 0.25.
(ii) It fails on the other two criteria:
(a) Mean magnitude of relative error is greater
than 0.25.
(b) The estimated eort is not within 25% of the
actual estimates for atleast 75% of the time.
Pred(0.25) is less than 0.75.
This is to be expected for a company going in for such
a model-based approach for the rst time. The project
metrics data that has been presently is not sucient
for the purpose of eort estimation. Fine-tuning of the
project metrics programme should be carried out so as
to make the COCOMO eort equation satisfy the Con-
tes criteria. This equation is recommended to the com-
pany for the purpose of eort estimation. Further
calibration needs to be done using more project data
Table 10
Eort estimation using the curve tted equation (Eq. (13))
Project name Actual eort PM COCOMO eort MRE Eq. (13) MMRE RMS RRMS Pred(0.25) Pred(0.5)
PROJ1 11.34 12.37 0.09 0.09 1.03 0.09 0.13 0.13
PROJ2 14.73 13.12 0.11 0.10 0.41 0.03 0.30 0.30
PROJ3 6.57 5.11 0.22 0.14 1.18 0.11 0.38 0.38
PROJ4 21.20 31.07 0.47 0.22 3.92 0.29 0.38 0.50
PROJ5 21.19 10.69 0.50 0.28 1.19 0.08 0.38 0.63
PROJ6 21.64 7.55 0.65 0.34 6.84 0.42 0.38 0.63
PROJ7 2.05 5.18 1.53 0.51 5.15 0.36 0.38 0.63
PROJ8 3.54 7.03 0.98 0.57 3.58 0.20 0.38 0.63
Table 11
Eort estimation using the curve tted equation (Eq. (14))
Project name Actual eort PM COCOMO eort MRE Eq. (14) MMRE RMS RRMS Pred(0.25) Pred(0.5)
PROJ1 11.34 14.01 0.23 0.23 2.67 0.24 0.13 0.13
PROJ2 14.73 14.45 0.02 0.13 1.69 0.13 0.30 0.30
PROJ3 6.57 8.74 0.33 0.19 2.63 0.24 0.38 0.38
PROJ4 21.20 22.88 0.08 0.17 3.12 0.23 0.38 0.50
PROJ5 21.19 12.96 0.39 0.21 0.89 0.06 0.38 0.63
PROJ6 21.64 10.76 0.50 0.26 5.17 0.32 0.38 0.63
PROJ7 2.05 8.64 3.22 0.68 2.30 0.16 0.38 0.63
PROJ8 3.54 10.36 1.92 0.84 0.30 0.03 0.38 0.63
Table 8
Eort estimation using the natural log approach
Project name Actual eort COCOMO eort MRE MMRE RMS RRMS Pred(0.25)
PROJ1 249.56 834.65 2.34 2.34 585.09 2.34 0
PROJ2 324.03 888.10 1.74 2.04 574.63 2.00 0
PROJ3 144.57 327.13 1.26 1.78 480.91 2.00 0
PROJ4 466.50 2213.57 3.75 2.27 967.74 3.27 0
PROJ5 466.22 715.05 0.53 1.92 872.70 2.64 0
PROJ6 476.15 494.51 0.04 1.61 796.70 2.25 0.10
PROJ7 45.00 346.82 6.71 2.34 746.37 2.41 0.10
PROJ8 77.94 458.35 4.88 2.66 711.00 2.53 0.10
Table 9
Equations from the curve tting approach
Tool Equation Correlation coecient
Microsoft Excel 2000 Y = 0.933x
1.0197
R
2
= 0.4307 (13)
SPSSv10
Data t Y = 3.534x
0.543
R
2
= 0.3600 (14)
Curve expert
and for various types of projects like porting projects,
maintenance projects, etc.
6. Conclusions and recommendations
6.1. Conclusions
The project data from the project database in the
company are studied and the COCOMO Model applied
to the past data. It is then calibrated using two
approaches:
Natural log approach.
Curve tting approach.
Calibrating with these approaches and analyzing the
outcome has resulted the following conclusions:
1. The calibrated COCOMO Equation for which the
error is the least is obtained by the curve tting
approach (Eq. (13)). The equation is
Effort 0:933Size
1:0197
: 14
This equation has a coecient of determination (R
2
) va-
lue of 43%.
2. The scale factors, eort multipliers and the multipli-
cative constants A and B, are estimated using curve
tting approach, based on the average of the
responses from the Project Managers. These are:
I
X
SF
j
16:27 I
X
EM
i
16:893
IA 0:056 IB 0:857:
3. The value of the multiplicative constant A arrived at
by Eq. (4) is 0.056, which is very low. This implies
that there might be some other factors that have an
impact on the actual software development eort.
This aspect requires more study.
4. The model has been validated using Contes Criteria
and observed the following:
(a) Of the three criteria, this model satises only one
criterion, viz., the RRMS for which the error is less
than 0.25.
(b) It fails on the other two criteria.
This is to be expected for a company going in for such a
model-based approach for the rst time. The project met-
rics data being collected presently is not sucient for the
purpose of eort estimation. Fine-tuning of the project
metrics data should be carried out so as to make the CO-
COMO eort equation satisfy the Contes Criteria.
6.2. Recommendations
The company is presently not using any model-
based approach for arriving at an eort estimate. It
is recommended that the company may apply the pop-
ular and well-known model, COCOMO II.2000, for
the purpose of eort estimation. It would be particu-
larly helpful to the company to have a better idea
about the actual eort spent in developing software.
This would, in turn, form the basis for further process
improvement initiatives in the organization. The fol-
lowing initiatives have been recommended to the
company:
1. Recording of the actual eort spent on the software
development should be done in a systematic manner,
where, even if a programmer works overtime and/or
weekends, it would be duly recorded.
2. The use of Code Count tool for the purpose of count-
ing the Logical SLOC is recommended if the organi-
zation uses COCOMO II.2000 model.
3. The scale factors and eort multipliers that form a
part of the COCOMO Equation should be used as
a benchmark or baseline for process improvement
initiatives.
4. The present method of arriving at Scale Factors and
Eort Multipliers through the process of question-
naire should be done away with and the rating scales
for the various scale and cost drivers must be equated
to the similar metrics that are in use. This will reduce
the bias in the evaluation of the various factors by the
project managers. For example, the existing processes
for employee evaluation can be linked to the rating
scale of Analyst Capability and Programmers
Capability.
6.3. Scope for future work
1. This case study is based on a database of 10
projects only. Even though these projects represent
the nature of the software being developed, the accu-
racy of the multiplicative constant A can be
increased by considering all the projects in the com-
panys database.
2. The factors, which aect the development eort and
cost apart from the factors listed by the COCOMO
model, can be studied and accounted for to arrive
at more accurate estimates.
References
[1] Albrecht AJ, Ganey JE. Software function, source lines of codes,
and development eort prediction: a software science validation.
IEEE Trans Software Eng 1983;SE-9:63948.
[2] Bernheisel Wayne A. Calibration and validation of the
COCOMO II cost/schedule estimating model to the space and
missile systems centre database. Master thesis, Ohio, Air Force
Institute of Technology; 1997.
[3] Boehm BW, Abts C, Clark B, Devnani-Chulani S. COCOMO II
model denition manual, The University of Southern California;
1997.
[4] Boehm BW et al.. Software cost estimation using COCOMO II.
1st ed. Englewood Clis, NJ: Prentice-Hall; 2000.
[5] Conte SD, Dunsmore HE, Shen VY. Software engineering metrics
and models. California: Benjamin Cummings; 1986.
[6] Danfeng Hong. Software cost estimation. Department of Com-
puter Science, University of Calgary. Available from: http://
pages.cpsc.ucalgary.ca/~hangd/SENG/621/report.html.
[7] David Longstreet. Function Point Analysis Training Course;
2002. Available from: http://www.softwaremetrics.com/fpafund.
htm.
[8] Fenton NE, Peeger SL. Software metrics: a rigorous and
practical approach. PWS Publishing Company; 1997.
[9] Hughes RT. Expert judgement as an estimating method. Inform
Software Technol 1996;38(2):6775.
[10] IFPUG Counting Practices Committee, Function Point Counting
Practices Manual. Release 4.1; 1999. Available from: http://
www.careld.com.hk/document/software.engine/fpa.pdt.
[11] Norden Peter V. Curve tting for a model of applied research and
development scheduling. IBM J Res Develop 1958;2(3).
[12] Parkinson GN. Parkinsons law and other studies in administra-
tion. Boston: Haughton-Min; 1957.
[13] Park R. The Central Equations of the PRICE software cost
model. 4th COCOMO Users Group Meeting, November;
1988.
[14] Parr NA. An alternative to the raleigh curve model for software
development eort. IEEE Software Eng 1980(May).
[15] Peeters David, Dewey Gerorge. Reducing bias in software project
estimates. Crosstalk J Defense Software Eng 2000.
[16] Putnam LH. A general empirical solution to the macro software
sizing and solution to the macro software sizing and estimating
problem. IEEE Trans Software Eng 1978;4(4):34561.
[17] Putnam LH, Ann Fitzsimmons. Estimating software cost, Dat-
amation; 1979.
[18] Ramesh Gopalaswamy. Managing global software projects. 1st
ed. New Delhi: Tata McGraw Hill; 2001. p. 226-7.
[19] Royce W. Software project mangement: a unied frame-
work. Reading, MA: Addison Wesley; 1998.
[20] Terdiman R, Datar R, Chohan S. Emerging oshore trends: a
view from India. Markets Gartner Report M-15-1087; 2001.
[21] Available from: http://sunset.usc.edu/reserch/CODECOUNT/
instructions.html.
Designbuild pre-qualication and tendering
approach for public projects
Khaled Al-Reshaid, Nabil Kartam
*
Department of Civil Engineering, College of Engineering, Kuwait University Construction Program, Major Projects
Administration, Kuwait University, P.O. Box 5969, Safat 13060, Kuwait
Received 26 March 2004; received in revised form 19 October 2004; accepted 17 November 2004
Abstract
While the designbuild delivery system is gaining ground in the private sector, due to its proven success, public sector owners are
still skeptical and hesitant to adopt it. Selecting the designbuilder can be complex process, particularly for public-sector projects.
The selection of private-sector designbuild teams can be much less formal. This paper presents the result of a study on the
applications of the designbuild delivery method in the public sector using professional construction management services. The
methodology and implementation of selecting a contracting rm for designbuild project is described in detail. The process includes
pre-qualication, technical and nancial evaluation, key issues in the Terms of Reference, pre-tender meeting, tenders design-super-
vision consultant, and critical contract clauses. The analysis resulted in positive indicators for the adaptation of designbuild and its
importance in public projects.
Keywords: Designbuild; Construction contracts; Construction management; Project delivery; Joint venture
1. Introduction
Designbuild contracts combine design and construc-
tion under a single entity and have gained the attention
of the industry as a whole. In Kuwait, the traditional
delivery approach of Design-Bid-Build is proven to be
inadequate in meeting the public sector owners require-
ments and expectations of nishing the projects on time
and within budget [1]. As a result, alternative delivery
methods, such as designbuild and Build-Operate-
Transfer (BOT), are seriously being considered by public
sector owners to overcome the drawbacks of the design-
bid-build. For an owner, the primary benet includes
the simplicity of having one party responsible for the
development of the project and the time saving as a re-
sult of overlapping design and construction phases.
The private sector in Kuwait has widely used the de-
signbuild delivery system with a great deal of success.
But in the public sector, designbuild is still perceived
to be subjective and complicated despite its benets doc-
umented in vast literatures [27]. Unlike the traditional
low price awarding approach for public projects, public
owners nd it dicult to evaluate designbuild propos-
als and justify the evaluation process to the public. Its
subjectivity comes from the fact that designbuild con-
tract is not only competitively bid according to the pub-
lic tendering regulations but also negotiated at the same
time.
Although designbuild is one of the more popular
alternative project-delivery methods that also has a long
history, it does not appear to have well-established con-
tractor selection procedures. Lack of scientic study of
designbuild procurement applications in Kuwait neces-
sitated this research project. Proven success of design
build in the private sector should be the motivator for
doi:10.1016/j.ijproman.2004.11.004
*
Corresponding author. Tel.: +965 9719202; fax: +965 4817524.
E-mail address: kartam@kuc01.kuniv.edu.kw (N. Kartam).
PROJECT
MANAGEMENT
public projects owners to adopt it for governmental
projects. The designbuild pre-qualication process rep-
resents the initial stages that an owner should employ
once a management decision to deliver a project using
design/build delivery has been made. It allows an owner
to review and determine appropriate prequalication
criteria for use in evaluating designbuild teams. The
process enables an owner to dierentiate between the
attributes of the competing teams, thus pre-qualifying
the team that best meet the owners needs. It then per-
mits the owner to maintain a consistent evaluation pro-
cess during prequalication. Several key issues have
been developed from current literature and actual expe-
rience and rened through a review of a public-sector
owners selection procedures. Lessons learned from an
actual designbuild project were used to dene critical
issues for a public owner intending to utilize design
build contract. This research falls in lines with the gov-
ernments consideration to modify its procurement laws,
guidelines and practices to accommodate new delivery
systems such as designbuild and BOT projects.
2. The project: multi-story Car Park
A case study of a Car Park project currently planned
to be built at Kuwait University on a designbuild basis,
was used. The objective was to investigate and identify
the important factors to be included within the pre-qual-
ication criteria as well as in formulating the Terms of
Reference (TOR) so as to be able to select a capable de-
signbuild team who will deliver the project successfully.
Lessons learned during this study helped in shaping crit-
ical issues for public owners with regards to implemen-
tation of the process.
This designbuild Car Park project is to accommo-
date 2300 car spaces in a site of area 32,000 m
2
. The esti-
mated contract duration is 27 months for design and
construction. Kuwait University (the owner) commis-
sioned, a joint venture international and local consulting
rm as the Construction Manager (CM) for this project.
The TOR required the designbuild team to be led
by the Contractor and carry out its obligation to deliver
the project in conformity with all Regulatory Authority
(Kuwait Municipality, Ministry of Planning, State
Audit Bureau, Ministry of Electricity and Water, etc.)
requirements in addition to coordinating the work with
other construction works planned or underway within
the University campus.
3. CM role in a designbuild project
A key factor in implementing a project on a design
build process is to introduce a CM Consultant to super-
vise the entire process on behalf of the owner. In the
traditional Design-Tender-Construction process, the de-
signer or A/E rm acts as the owners representative,
whereas the designer is a part of the contractors team
in the designbuild process. Therefore, he/she cannot
supervise the work on behalf of the owner as he/she is
loyal to the contractor. It, therefore, becomes imperative
for the owner to employ an independent agency to look
after his/her interest. Specialized CM rms are available
in the construction industry with extensive knowledge,
experience and know-how in the pre-design, design, ten-
der, construction and operation phases of a project.
CMs duties include assisting the owner and designer/
builder in planning, reviewing plans and specications,
applying value engineering principles, determining con-
struction feasibility, providing budget estimates, and
allocating the work to facilitate bidding and award pro-
cedures. Such CM rms utilize special computer-based
software and programs to control cost, time and quality
and deliver the project to owners satisfaction.
The services of a CM were utilized by Kuwait Univer-
sity to establish the pre-qualication criteria according
to the space program and other performance require-
ments. Main attributes of existing pre-qualication pro-
cess were scoring system and selection criteria to achieve
optimal team integration, experience, and resources
from the evaluated designbuild teams. Also, the ser-
vices of this CM rm will be utilized to supervise and
closely monitor contractors performance since the de-
signbuild approach cannot exclusively rely on the
objective role of the design rm in supervision.
4. Pre-qualication
There is a need for a systematic approach to selecting
designbuild teams based on a strong program and an
outstanding professional pre-qualication process [8].
Existing literature discusses dierent designbuild deliv-
ery approaches for public projects, where clear and iden-
tied objective criteria are reected to minimize
possibility of risk to the public-sector owner [911].
The challenging avenue for Kuwait University, in the
selection of a proper designbuild team for its Car-Park
project, was to balance design, construction, supervi-
sion, and designbuild joint venturing experiences in
its pre-qualication process.
Criteria used in the pre-qualication process included
review of the credentials and experience of the various
design/supervision consultants proposed by the various
contractors. In addition, subsequent tender submissions
also included evaluation of the selected consultants
qualication and reputation in order to gain further
assurance of professional conduct from the consultant.
For the designbuild Car-Park project, the pre-qualica-
tion process went through three stages as shown in
Fig. 1.
310 K. Al-Reshaid, N. Kartam / International Journal of Project Management 23 (2005) 309320
The rst stage was the general pre-qualication crite-
ria imposed by the Government of Kuwait through a
Central Tenders Committee (CTC) on all bidders for
public projects. The CTC is the public agency in Kuwait,
in charge of evaluating and qualifying contractors inter-
ested in bidding for public projects. According to CTCs
criteria, general contractors for public projects are clas-
sied into four major categories or grades, based on
the volume of successfully executed public construction
works, measured in monetary value, and based on the
contractors nancial capabilities and technical creden-
tials. The output of this initial stage was more than 50
potential contractors.
The second stage of the evaluation process was car-
ried out by the owner, Kuwait University, through its
CM. Grade 1 General Contractors, the top level cate-
gory of the CTC listing, were pre-qualied for new
and upcoming construction projects for Kuwait Univer-
sity. The specic needs and priorities of Kuwait Univer-
sity were considered in the two main parts of the
evaluation process (Table 1): technical and nancial,
with the following criteria:
1. Technical evaluation
General completeness and quality of submission.
Structure and organization activities, experience,
anticipatedstrategy, procurement andorganization.
Personnel availability of technical, administra-
tive and eld personnel.
Plant and machinery availability of suitable con-
struction equipment.
Other resources subcontractors, fabrication facil-
ities, shop drawings, and hardware and software
availability.
Companys experience value and type of exe-
cuted projects.
Credentials of the autonomous design rm that is
part of the Designbuild consortium.
2. Financial evaluation:
Value of executed projects last 5 (+) years and
ongoing work.
Audited nancial statements.
Bank references and bond-ability proof.
Financial power ratios of assets/liabilities/share-
holders equity.
The third stage of pre-qualication focused on pro-
ject requirements (Table 2). In this stage, the following
general areas were considered:
Experience in designbuild projects in general.
Experience specically in Car-Park designbuild.
In-house vs. joint venturing of design and construc-
tion capabilities.
Project control methods used value Engineering,
quality and cost control.
In addition, Kuwait University introduced a new
criterion in its pre-qualication process which was to
nd out the total amount of claims submitted on each
project by each bidder including those which were ap-
proved. The objective was to identify potential bidders
who are prone to submitting claims by ascertaining
the extent of unapproved claims submitted by each
bidder on their projects and to send a clear message
to such bidders that a co-operative environment dur-
ing an on-going project will be a major consideration
for Kuwait University in its future pre-qualication
process.
Fig. 1. Pre-qualication stages.
K. Al-Reshaid, N. Kartam / International Journal of Project Management 23 (2005) 309320 311
Table 1
Contractors Pre-qualication (for a project with a budjet estimates between KD 4M and KD 8M
EVAL UAT IO N CRIT ERIA SH EET
EVAL UAT IO N CR IT ER IA Key Po in t s T o Con si der CRI T ERI A / PO INT S
100 -8 1 80 - 61 60- 41 40- 21 2 0- 0
Gen eral ( 5%) 5
1 Com plet ene ss of Subm is s ion ( 50%) 50 Fo rm at s , Cer t if i ca tes ( Exc lu ding F inan ci al Pr oof s ) Per s ona l Ju dgm e nt
2 Qua lit y of Pr es ent at ion ( 5 0%) 50 Fo llowing F or m at s as Requ ir ed Per s ona l Ju dgm e nt
A St r uct u re & O rga ni za t io n ( 15 %) 15
2 /F 1,2 Com pany's Pr i nc ipal Ac t ivit ies ( 15%) 15 Gove rn m en t Pr oj ec ts , Ed uc ati onal F ac ili ti es , Labo ra tor i es Exc el lent V. Goo d G ood Fai r Poor
3 Ex per ienc e as Ma in Cont r ac to r ( 65%) 65
a. In Kuwa it ( 80%) 80 Num ber of year s > 1 0 10 - 7 6 - 4 3 - 2 1 - 0
b. In GC C Ar ea (1 0%) 10 Num ber of year s > 5 4 3 2 1 - 0
c . In Midd le Eas t ( 5%) 5 Num ber of year s > 5 4 3 2 1 - 0
d. W o rl dwi de (5 %) 5 Num ber of year s > 2 1 -- - - 0
5 ,6 Par ent , Subs id iar ies & J oi nt Vent ur e Par tn 5 Sat is f ac t or y sc op e o f wor k de sc r ip ti on f or eac h ent it y Exc el lent V. Goo d G ood Fai r Poor
7 Ant ic ipat ed Str a tegy f or p r oj ec t ac c om plis 5 Sat is f ac t or y appr oac h Exc el lent V. Goo d G ood Fai r Poor
8 Pr oc edur es of Pr oc ur em e nt ( 5%) 5 Sat is f ac t or y appr oac h Exc el lent V. Goo d G ood Fai r Poor
9 Com pany's Or gani zat ion Char t (5 %) 5 Pr ovide d a nd Sa ti sf ac to ry Exc el lent V. Goo d G ood Fai r Poor
B Fi nan ci al I nf or mat io n ( 25%) 25
2 T ot al val ue of c ons tr uc t ion i n l as t 4+ year s 10 T ot al Am ount i n t he la st f our yea rs + c ur r ent year > 75 Mil lio n 75 - 56 M 5 5 - 38M 37 - 19M < 19 Mi lli on
3 T ot al val ue of ongo ing bac k log of wor k ( 5% 5 Cur r ent co ns tr . ba ck lo g wor k as a % of t he Ann. Avg. f r om B2 < 10 0% 100 - 119%120 - 1 39%1 40 - 15 9% 1 60 - 200 %
4 Com pany's Au dit ed Fi nanc ial St at em e nts 10 Pr ovide d a s r eque st ed Pr o vi ded - - - N ot Pr ovide d
5 ,6 Ref er enc e ( 35%) 3 5
5. Ba nk Ref er enc e ( 50%) 50 Pr ovide d / Not Pr ovide d Pr ovide d - - - N ot Pr ovide d
6. Bo ndabi lit y Pr oof ( 50%) 50 Pr ovide d / Not Pr ovide d Pr ovide d - - - N ot Pr ovide d
8 Fi nanc ia l Rati os ( 4 0% ) 40
a. Cur r ent Ra ti o ( 30%) 30 Cur r ent As s et s / C ur re nt Li abil it ies > 200 200 - 15 0 14 9 - 100 99 - 50 49 - 0
b. Qui ck Rati o (3 0%) 30 Cur r ent As s et s -I nven tor y- Ca pit al Ex pens es - Pr epai d- W I P/Cu r re nt Li abil it ies 150 - 1 20 119 - 90 8 9 - 60 59 - 30 <3 0
c . Debt Ra ti o ( 20 %) 20 T ot al Li abil it y - Shar eh older s Equit y / T ot al T angib le Ass et s 0 - 20 21 - 40 4 1 - 60 61 - 80 81 - 90
d. Net W or k ing Capi ta l t o Equit y 20 Cur r ent As s et s - Cur r ent Li abi lit ies / Shar eho lder s Equit y 60 - 4 5 44 - 30 2 9 - 20 19 - 10 < 10
C Pe rso nne l ( 1 0%) 1 0
1 .1 T ec hn ic al & F iel d ( 25%) 2 5 No. o f T ec hni c al & F ield s t af f wit h c om pany > 6 0 60 - 46 4 5 - 31 30 - 15 < 15
1 .2 Adm i nis t r ati ve/Ex ec ut ive s ta f f ( 25%) 25 Num ber of Adm i nis t r ati ve a nd Ex ec ut ive s t af f with c om pany > 1 0 10 - 9 8 - 7 6 - 5 < 5
2 .0 Manage ri al s ta f f C.V. 's ( 20%) 20 I ncl udi ng En gine er ing & N on- Engin eer ing De par tm ent s > 5 5 - 4 3 2 1 - 0
4 .1 Engin eer s ( Ex per ie nc e > 15 Year s ) ( 15%) 15 Num ber wit h re sp ec t to ex pe ri enc e > 4 4 3 2 1 - 0
4 .2 Engin eer s ( Ex per ie nc e 10 - 1 5 Year s ) (10%) 10 Num ber wit h re sp ec t to ex pe ri enc e > 4 4 3 2 1 - 0
4 .3 Engin eer s ( Ex per ie nc e 5 - 10 Year s ) (05%)% 5 Num ber wit h re sp ec t to ex pe ri enc e > 5 5 - 4 3 2 1 - 0
D Pl an t & M a chi ner y ( 10 %) 10
1 Conc r et e Mi x er s ( 10%) 10 Num ber of equi pm e nt > 8 8 - 7 6 - 5 4 - 1 0
2 Ex cava tor s ( 10%) 10 Num ber of equi pm e nt > 6 6 - 5 4 - 3 2 - 1 0
3 T r ac t or s & Loade r( s ) ( 1 0%) 10 Num ber of equi pm e nt > 1 0 10 - 8 7 - 5 4 - 1 0
4 T r uc k s ( 10%) 10 Num ber of equi pm e nt > 1 0 10 - 7 6 - 4 3 - 1 0
5 T ower Cr an es ( 15 %) 15 Num ber of equi pm e nt 2 1 -- - - 0
6 Mobil e Cr anes ( 1 5%) 15 Num ber of equi pm e nt > 4 3 2 1 0
7 Power Gen er ato r s & C om pr es s or s (1 0%) 10 Num ber of equi pm e nt > 1 0 10 - 6 5 - 3 < 3 0
8 Fo rm wor k & Sc af f oldi ng, o r Ot her , ( 2 0%) 20 Sat is f ac t or y se ts of Fo rm Syst em s & Sc af f ol din g, or Ot her Maj or I t em s Exc el lent V. G ood G ood Fai r Poor
E Ot her Re sou rces ( 5%) 5
1 Sub- Cont r ac t or s (2 0%) 2 0 Sat is f ac t or y Sub- c ont ra ct or s Exc el lent V. G ood G ood Fai r Poor
2 Fa br ic ati on Fac i lit ies ( 20%) 20 Sat is f ac t or y Fabr i ca ti on Pl ant s be long s t o Co m pa ny Exc el lent V. G ood G ood Fai r Poor
3 Pr oc ur em e nt (2 0%) 20 Degr ee of Sa ti sf ac to ry an swer Exc el lent V. G ood G ood Fai r Poor
4 Shopd ra wing s Pr oduc t ion s our c e ( 20%) 20 Sat is f ac t or y so ur ce f or Sho pdr awi ngs pr oduc t ion E xc el lent V. G ood G ood Fai r Poor
5 Com put er r es ou rc es ( 20%) 20 Sat is f ac t or y avail able Co m put er sys. ( Har dwar es & Sof twar e) Exc el lent V. G ood G ood Fai r Poor
F Comp any' s Ex peri en ce ( 30 %) 30
1 Pr oj ec t s' Va lue, (L as t 10 Year s ) ( 55 %) 55 Num ber of Pr oj ec t s wi th valu e: 4 M KD and over > 8 8 - 6 5 - 4 3 - 2 1 - 0
2 Pr oj ec t s' T ype , ( Las t 10 Year s ) ( 3 0%) 30 Num ber of Pr oj ec t s of Sim il ar T ype > 6 6 - 5 4 - 3 2 - 1 0
3 Com pany's Ex pe ri enc e in Sim ilar W or k 15 Year s of Exp er ienc e in Sim il ar W or k > 1 0 10 - 7 6 - 4 3 - 2 1 - 0
3
1
2
K
.
A
l
-
R
e
s
h
a
i
d
,
N
.
K
a
r
t
a
m
/
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
P
r
o
j
e
c
t
M
a
n
a
g
e
m
e
n
t
2
3
(
2
0
0
5
)
3
0
9
3
2
0
Table 2
Designbuild contractors evalutation matrix
Category Allocated
%
Points
0 1 2 3 4 5 6 7 8 9 10
A design build experience
A-2 Annual KD volume of designbuild
work (ave. of last 10 yrs)
0 1 2 3 4 5 6 7 8 9 10+
A-3 Appr. % annual average of design
build vol. to total vol.
0 +0 to 10 +10 to 20 +20 to 30 +30 to 40 +40 to 50 +50 to 60 +60 to 70 +70 to 80 +80 to 90 90+
A-4 Design experience
In-house design experience No Yes
In-house/design rm sta
(a) Number of A/E in Eng. Dept. 0 +0 to 2 +2 to 4 +4 to 8 +8 to 10 10+
(b) No. of Draftsmen, Technical,
Clerical & Secretarial
0 +0 to 2 +2 to 4 +4 to 8 +8 to 10 10+
(c) Department Head; Named (Y/N) No Yes
(c) Department Head; Degree (Y/N) No Yes
(d) Years as Department Head 0 1 2 3 4 5 6 7 8 9 9+
(e) Company licensed for design in
Kuwait
No Yes
(f) Scope of license (full or restricted)
A-5 Designbuild projects experience + 4 to 6
No. of designbuild projects
completed by the rm
0 +0 to 2 +2 to 4 +6 to 8 +8 to 10 +10 to 1 2 +12 to 14 +14 to 16 +16 to 18 +18
Average volume of designbuild
projects executed
0 3+
Time/cost, performance assessment Fail Poor Good V.Good Excellent
A-6 Car-Park experience
No. of Car-Park projects completed
by the company
0 +0 to 2 +2 to 4 +4 to 8 +8 to 10 10+
No. of design build Car-Park
projects completed
0 1 2 3 4 5+
Average volume of Car-Park projects
executed
0 +.5 to 1 + 1 to 1. 5 1. 5 to 2 2 to 2.5 3+
Average vehicle capacity of Car-Park
projects
0 500 2000 3500+
Time/cost, performance assessment Fail Poor Good V. Good Excellent
A-7 Current design-bulid projects
No. of current designbuild projects 0 1 2 3 4 +5
Ave. vol. of current designbuild
projects (Mill. KD)
0 1 2 3 4 +5
Ranking of current DB projects (in
size, Quality)
Fail Poor Good V. Good Excellent
(continued on next page)
K
.
A
l
-
R
e
s
h
a
i
d
,
N
.
K
a
r
t
a
m
/
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
P
r
o
j
e
c
t
M
a
n
a
g
e
m
e
n
t
2
3
(
2
0
0
5
)
3
0
9
3
2
0
3
1
3
Table 2 (continued)
Category Allocated
%
Points
0 1 2 3 4 5 6 7 8 9 10
A-8 Company carries professional
indemnity for designbuild
No Yes
A-9 Experience of outside Architect/
Engineering rm
No. of years of experience (or in
business relation)
0 +0 to 1 +1 to 2 +2 to 3 +3 to 4 +4 to 5 +5 to 6 +6 to 7 +7 to 8 +8 to 9 +9
Rank the Architect/Engineering
rm capability
Sub-consulting rms ranking for
the following:
01: Architectural Fail Poor Good V. Good Excellent
02: Structural design works Fail Poor Good V. Good Excellent
03: Mechanical design works Fail Poor Good V. Good Excellent
04: Electrical design works Fail Poor Good V. Good Excellent
05: Trac Engineering Fail Poor Good V. Good Excellent
06: Landscaping Fail Poor Good V. Good Excellent
07: Quantity surveyors Fail Poor Good V. Good Excellent
08: Value Engineering Fail Poor Good V. Good Excellent
09: Others (full scale for local
sub-consultant)
A-10 Rank inspection testing services
consultant
A-11 Rank project cost control
procedures
A-12 Rank value Eng. cost/benet
analysis capability
A-13 Sub-contract works
Estimated % (based on total
Project Cost)
+ 90 + 80 to
90
+ 70 to
80
+ 60 to
70
+ 50 to
60
+ 40 to
50
+ 30 to
40
+ 20 to
30
+ 1 0 to
2 0
+ 0 to
10
0
A-14 Claims on KU projects
(Explanatory Assessment)
3
1
4
K
.
A
l
-
R
e
s
h
a
i
d
,
N
.
K
a
r
t
a
m
/
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
P
r
o
j
e
c
t
M
a
n
a
g
e
m
e
n
t
2
3
(
2
0
0
5
)
3
0
9
3
2
0
5. Pre-qualications results
Grade 1 contractors list (more than 50 contractors)
was obtained from the CTC and included 25 leading
contracting rms in Kuwait for general construction
works. Kuwait University sent pre-qualication ques-
tionnaire to all 25 Grade 1 contractors. 20 contractors
responded positively to the questionnaire, but only 12
contractors passed the minimum requirements, as
shown in Table 3. Those 12 contractors had experience
on a wide range of designbuild project types, as shown
in Fig. 2. This process helped in narrowing the bidders
to those with experience in similar projects which in
turns reduces the risk to the owner. These short listed
Table 3
Designbuild contractors evaluation summary ranking
S. No. Firm name Evaluator-1 Evaluator-2 Evaluator-3 Average P/F
1
1 Contractor A 31.59 32.54 28.41 30.847 P
2 Contractor B 25.71 26.35 34.29 28.783 F
3 Contractor C 23.02 24.60 22.54 23.387 F
4 Contractor D 51.59 46.83 53.02 50.480 P
5 Contractor E 35.56 35.08 48.41 39.683 P
6 Contractor F 10.48 10.16 11.90 10.847 F
7 Contractor G 63.81 60.32 57.30 60.477 P
8 Contractor H 9.52 10.63 10.32 10.157 F
9 Contractor I 33.49 33.17 30.79 32.483 P
10 Contractor J 55.40 53.33 59.05 55.927 P
11 Contractor K 27.94 29.68 28.25 28.623 F
12 Contractor L 48.71 42.49 48.84 46.680 P
Rank S. No Firm name Evaluator-1 Evaluator-2 Evaluator-3 Average P/F
1
Final ranking
1 7 Contractor G 63.81 60.32 57.30 60.477 P
2 10 Contractor J 55.40 53.33 59.05 55.927 P
3 4 Contractor D 51.59 46.83 53.02 50.480 P
4 5 Contractor E 35.56 35.08 48.41 39.683 P
5 9 Contractor I 33.49 33.17 30.79 32.483 P
6 1 Contractor A 31.59 32.54 28.41 30.847 P
7 2 Contractor B 25.71 26.35 34.29 28.783 F
8 11 Contractor K 27.94 29.68 28.25 28.623 F
9 3 Contractor C 23.02 24.60 22.54 23.387 F
10 12 Contractor L 18.89 15.24 16.98 17.04 F
11 6 Contractor F 10.48 10.16 11.90 10.847 F
12 8 Contractor H 9.52 10.63 10.32 10.157 F
1
Maximum score is 100 and minimum pass score required is 30 points.
Fig. 2. DB experience of dierent types of projects done by the 12 contractors.
rms were invited to submit proposals, based on the
Terms Of Reference (TOR) that was prepared by the
CM, and to attend a pre-tender meeting.
6. Evaluation criteria
One critical aspect of the evaluation was the selection
criteria for the designbuild contract. In Kuwait, the
designbuild evaluation procedure is not yet well estab-
lished for public projects, and the awarding of design
build tenders will still be based on the lowest bid as is
the practice with the traditional design-bid-build.
Though it seems that it will combine the advantages of
the two delivery systems, the lowest price of the tradi-
tional Design-Bid-Build and the better technical quality
of the designbuild, it has a major drawback; the owner
will not have the option of selecting the architectural de-
sign. The designbuild team of the lowest bid will deter-
mine the outlook of architectural design, which may not
be preferred by the owner.
The only scoring matrix will be whether or not the de-
sign had the minimum required technical standards. As
a result, Kuwait University faced with the challenge of
drafting the TOR to have the project scope well de-
signed with all the minimum expected technical stan-
dards. In addition, xed provisional sums, which will
be controlled by the owner, were included in the con-
tract for later enhancement to some items that are not
nalized in terms of specications, quantities or design.
It can also help in minor improvements in the contract
for the selected lowest bid, should that be needed.
7. Request for proposals
Limiting the number of pre-qualied bidders in a de-
signbuild contract would increase the contractors
interest in participation, since a larger cost is associated
with preparing a designbuild tender [12]. Even though
it is desirable to limit the competition among nalists
to be between three to ve to ensure fairness to partici-
pating teams, Kuwait University decided to send Re-
quest for Proposal (RFP) to a larger number of
design-builder for the following reasons:
The project is under very tight budget and higher
number of competitors would help drive submitted
prices down.
The Car-Park project is simple, conventional and
well understood to be of generic pre-cast concrete
structure.
Highly regarded contractors might not be interested
in bidding for the project and expressed interest only
for their public image; therefore, they may be inclined
to submit high bids that are not competitive.
Before sending out the RFP, Kuwait University
needed approval from appropriate Public Authorities
for solicitation of designbuild proposals through the
CTC. As a result, a series of meetings were held with
concerned Public Authorities and approval was granted.
It was agreed that the proposals would be invited by the
CTC who would then forward the complete proposals
to Kuwait University for review and evaluation. Subse-
quently, the contract will be awarded to the selected
bidder based on the recommendation of Kuwait
University.
The TOR for the project required the tenderers De-
sign-Supervision Consultant to act as an independent
and autonomous entity. Upon award of the contract,
this requirement will be re-emphasized, whereby the se-
lected Consultant shall be obliged, through a declara-
tion to render its services and exercise its authority in
an unbiased and impartial manner with full allegiance
to the sole interest of the project and the owner. The
items enumerated below are precautionary measures to
assist the owner in the administration and management
of the project in a proper and ecient manner. It is ex-
tremely important to incorporate these requirements in
the Tender Documents so that the Consultant is fully
aware of his/her obligations during the execution of
the contract.
The selected Consultant is required to provide Pro-
fessional Indemnity Insurance at the end of design
stage to indemnify the owner against damages and
losses arising out of errors and omissions in the
design. This will hold the Consultant fully responsible
for the accuracy of his design.
The Consultants principal site representative (Resi-
dent Engineer on site) is designated as the Engineers
Representative with specic authority and responsi-
bility delegated to him by the owner. This will further
serve to emphasize the authority of the Consultant, as
dictated and controlled by the owner. In this type of
set-up the Engineers Representative will be held fully
accountable in the discharge of his duties, pursuant to
and to the extent of powers delegated to him by the
owner.
Only the owner has the authority to approve the sta
proposed by the Consultant and also retains subse-
quent power to remove any of his sta found unsuit-
able to carry out his duties during supervision.
The owner has full control on release of Consultants
payment.
8. Pre-tender meeting
The objective of a pre-tender meeting is to clarify and
respond to questions and ambiguities in tender docu-
ments as identied by the bidders. During the pre-tender
meeting of the designbuild Car Park project for the
Kuwait University, two major and important issues as
pointed out by the bidders were:
1. A Trac Study Report for the campus master
plan prepared by a specialized consultant was
requested, as it was not included in the tender
documents. This was necessary to enable the
bidders in developing the design concept with
respect to entry and exit points of the Car Park
building.
2. A query on the extent of existing underground ser-
vices within the building site led to the discovery of
additional chilled and hot water pipe lines connected
to the Campus network mains. This was missed out in
the original tender documents.
Consequently, two separate addenda were issued to
the bidders clarifying the above two situations.
9. Submitted tenders
Nine out of the 12 invited tenders submitted their
bids to CTC before the closing deadline; the tender re-
sults are listed in ascending order in Table 4. Contractor
D submitted the lowest tender; the other tender
amounts exceed the lowest bid by the percentages
shown in Table 5. The bids were compared to the lowest
bid in terms of both the total tender amount and the cost
distribution into design, construction and supervision
(Fig. 3).
10. Compliance with tender requirements
The three lowest bidders, contractors D, A&I, respec-
tively, were analyzed for compliance with the tender
requirements according to the following four categories:
(1) Submitted requirements, (2) structural, (3) architec-
tural, and (4) mechanical and electrical. The owners
CM conducted a thorough examination of the three sets
Table 4
Tenders submitted
1. D KD 3,269,569.000
2. A KD 3,747,000.000
3. I KD 3,924,000.000
4. F KD 4,444,000.000
5. J KD 4,971,000.000
6. H KD 5,670,267.000
7. L KD 5,750,598.000
8. B KD 6,492,000.000
9. K KD 7,984,000.000
The following contractors did not respond:
1. G
2. C
3. E
Table 5
Tender amounts of bidders exceeding the lowest bidder
SL Bidder Cost % exceeding the lowest bidder
1. D 0.00
2. A 14.60
3. I 20.02
4. F 35.92
5. J 52.04
6. H 73.43
7. L 75.88
8. B 98.56
9. K 144.19
Fig. 3. Submitted tenders.
Table 6
Comparison of arithmetically checked & discount adjusted tender sums of three lowest tenders
SL NO REMARKS
KD FILS KD FILS KD FILS 1 vs 2 1 vs 3
1 DW/1-A Stage 1, Phase I, II & III 15,000 000 14,400 000 5,638 200 -4% -62%
2 DW/1-B Stage 2, Phase I 75,000 000 52,800 000 9,397 000 -30% -87%
3 DW/1-C Stage 2, Phase II & III 50,000 000 19,200 000 75,177 000 -62% 50%
4 DW/1-D Shop Drawings etc. 10,000 000 10,020 000 4,698 500 - -53%
Bill No. 01 - Total 150,000 000 96,420 000 94,910 700 -36% -37%
5 S/1 Division 1 - General Requirements 15,000 000 166,262 936 178,133 170 1008% 1088%
6 S/1 Division 2 - Site Construction 107,330 472 275,105 831 208,932 049 156% 95%
7 S/1 Division 3 - Concrete 1,861,540 213 1,945,163 328 2,166,688 059 4% 16%
8 S/1 Division 4 - Masonry 8 ,833 530 3,925 063 4,495 760 -56% -49%
9 S/1 Division 5 - Metals 22,870 670 130,114 097 125,291 201 469% 448%
10 S/1 Division 6 - Wood & Plastics - - - - - - - -
11 S/1 Division 7- Thermal and Moisture Protection 130,381 545 155,933 750 216,444 745 20% 66%
12 S/1 Division 8 - Doors & Windows 25,503 633 16,933 358 14,707 305 -34% -42%
13 S/1 Division 9 - Finishes 131,350 038 161,057 617 202,273 680 23% 54%
14 S/1 Division 10 - Specialties 6,248 281 15,886 032 13,813 590 154% 121%
15 S/1 Division 11 - Equipment - - - - - - - -
16 S/1 Division 12 - Furnishings - - - - - - - -
17 S/1 Division 13 - Special Construction 65,465 019 - - - - - -
18 S/1 Division 14 - Conveying System 154,844 080 139,917 641 99,834 728 -10% -36%
19 S/1 Division 15 - Mechanical 103,090 704 151,046 631 149,951 643 47% 45%
20 S/1 Division 16 - Electrical 232,298 100 220,500 721 205,313 922 -5% -12%
Bill No. 02 - Total 2,864,756 285 3,381,847 005 3,585,879 226 18% 25%
Bill No. 03- Supervision
21 236,265 000 260,066 832 235,880 414 10% -
Bill No. 03 - Total 236,265 000 260,066 832 235,880 414 10% -
22 Operation & Maintenance 33,600 000 - - 5,638 200 - -83%
23 8,400 000 8,666 163 1,691 460 3% -80%
Bill No. 04 - Total 42,000 000 8,666 163 7,329 660 -79% -83%
- - - - - - - -
Total Tender Cost 3,293,021 285 3,747,000 000 3,924,000 000 14% 19%
Special Discount
Final Tender Cost 3,293,021 285 3,747,000 000 3,924,000 000 14% 19%
Bill No. 01 - Design Works
Bill No. 02 - Construction & Commissioning
B.O.Q. BILL REFERENCE ANDDIVISION Contr. D Contr. A Contr. I % VARIANCE
Adjustment
Add/Omit
Consultant's Supervision
Bill No. 04- Operation & Maintenance
Ditto - Optional
Adjusted Adjusted Adjusted Adjusted
3
1
8
K
.
A
l
-
R
e
s
h
a
i
d
,
N
.
K
a
r
t
a
m
/
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
P
r
o
j
e
c
t
M
a
n
a
g
e
m
e
n
t
2
3
(
2
0
0
5
)
3
0
9
3
2
0
of tender documents and concluded that all three have
complied with the overall tendering requirements, except
for minor deviations and discrepancies in all of the ma-
jor four categories. Such minor discrepancies would be
required to be rectied through a declaration by the rec-
ommended bidder prior to formal award of contract.
11. Financial appraisal
The nancial appraisal of the three lowest tenders
were check for arithmetical and tendering errors, and
were dealt with either by absorbing the arithmetical er-
rors in the lump sum tender amount, or by adding omit-
ted tendering sums to the total lump sum amount in the
case of the provisional sum and the optional operation
and maintenance sum (Table 6). The nal corrections
are as shown in Fig. 4. As well, the nancial appraisal
included comparison between the three lowest bidders
on design accumulated cost vs. phase, arithmetically
checked and discount adjusted according to the CSI 16
divisions, concrete works, cumulative and monthly work
in place.
12. Contract awarding
The evaluation criteria were designed in a concise and
precise way to eectively select the lowest responsible
bidder in designbuild projects. Based on the compre-
hensive analysis that included evaluation of the bidders
compliance with the tender requirements, technical as-
pects and nancial criteria, it has been concluded that
the lowest bidder contractor D had, for the most
part, fullled the requirements of the Tender and oered
the best price. The recommendation by Kuwait Univer-
sity is to award the designbuild contract to the lowest
bidder, after its accounts for the minor deviations and
discrepancies outlined in a declaration.
13. Conclusions
Construction industry professionals predict that the
designbuild project delivery system will soon be the
dominant method of project delivery. While design
build can be used with any project, it is most prevalent
in private sector work but is growing in acceptance for
public sector work. It is characterized by its single con-
tract with the owner and by the overlapping of design
and construction phases. Designbuild concept works
best when the owner is knowledgeable, vigilant and par-
ticipates actively in project activities in all its phases
from design development through construction and
commissioning. This can be achieved either through
his own team of professionals or through a specialized
CM rm acting as Owners Representative under direc-
tion from the owner.
Fig. 4. Comparison of project cost as proposed by tenderers.
This study demonstrated the stages of tendering and
contracting in the designbuild delivery system from
an owners point of view with CM consultant. The pa-
per presented the methodology and results of selecting
designbuild contractor for a public project. It de-
scribed comprehensive criteria for contractors pre-
qualication, contractors evaluation, contractors
compliance with submitted requirements as per TOR,
and contractors nancial appraisal. The described
contractors evaluation criteria include: Annual volume
of designbuild work (average of last 10 years), % an-
nual average of designbuild volume to total work vol-
ume, in-house versus joint venturing design and
construction capabilities, current and previous de-
signbuild projects experience, similar projects experi-
ence, inspection testing services, project control
procedures (quality and cost control, value engineering
cost/benet analysis capability), % sub-contract works,
claims history against Kuwait University. Such com-
prehensive pre-qualication and evaluation process is
vital to choosing a competent, yet competitive de-
signbuild contractor who can eectively and econom-
ically serve the owners interest. The proposed process
will provide an even more structured approach and as-
sist in formulating consistent and reusable guidelines
for owners in contractor selection for designbuild
projects. Added benets include optimized costs (for
both owners and contractors) and fewer chances of
bids being challenged by unsuccessful bidders.
It is evident from the above example that pre-tender
meetings very often lead to clarication of doubts and
ambiguities in the tender documents, resulting in a more
accurate set of tender documents. It is of prime impor-
tance to ensure that the bidders understood the scope
of the work, the design intent, technical requirements
and other contractual terms and conditions of the con-
tract very well, so that the tenders they submit are com-
prehensive and accurate. This will minimize future
complications and claims during construction. Thus,
pre-tender meetings provide useful assistance in this re-
gard, particularly in the case of designbuild tenders,
where the detailed scope and specications are not de-
scribed in the tender documents and have to be devel-
oped by the bidders, along with their design concept.
The other important factor that can enhance the
quality of a tender is to introduce Provisional Sums,
which will be controlled by the owner as and when
deemed necessary. Provisional Sums are meant for
inclusion of works or items that cannot be specied at
the time of tender or those that are not within the nor-
mal scope or expertise of bidders. This gives an opportu-
nity for the owner to add specialized requirements
during the construction stage without going through
the process of variations which often leads to delays
and extra costs.
Acknowledgements
The authors gratefully acknowledge the nancial sup-
port of Kuwait University Research Administration
Grant #EV 03/03.
References
[1] Kartam N, Al-Daihani T, Al-Bahar J. Professional construction
management practices in Kuwait issues, diculties and recom-
mendations. Int J Project Manage 2000;18(2):28196.
[2] Songer A, Molenaar K. Project characteristics for successful
public-sector designbuild. J Constr Eng Manage
1997;123(1):3440.
[3] Molenaar K, Songer A. Model for public sector design
build project selection. J Constr Eng Manage 1998;124(6):
46779.
[4] Molenaar K, Songer A, Barash M. Public-sector design/
build evolution and performance. J Manage Eng 1999;15(2):
5462.
[5] Ndekugri I, Turner A. Building procurement by design and build
approach. J Constr Eng Manage 1994;120(2):24356.
[6] Ernzen J, Schexnayder C. One companys experience with design/
build: labor cost risk and prot potential. J Constr Eng Manage
2000;126(1):104.
[7] Flora G, Ernzen J, Schexnayder C. Field-level managements
prespective of design/build. Pract Period Struct Des Constr
1998;3(4):1805.
[8] Potter K, Sanvido V. Design/build prequalication system. J
Manage Eng 1994;10(2):4856.
[9] Potter K, Sanvido V. Implementing a design/build prequalica-
tion system. J Manage Eng 1995;11(3):304.
[10] Palaneeswaran E, Kumaraswamy M. Contractor selection for
design/build projects. J Constr Eng Manage 2000;126(5):3319.
[11] Gurry W. Documenting designbuild. ASCE Civil Eng
1995;65(9):479.
[12] Gordon J. Unit price contractor or designbuild. Hydropower
Dams 2001;5:99101.
Evaluation of stakeholder inuence in the
implementation of construction projects
Stefan Olander
*
, Anne Landin
Division of Construction Management, Lund University, P.O. Box 118, S-221 00 Lund, Sweden
Received 6 July 2004; received in revised form 27 August 2004; accepted 11 February 2005
Abstract
A negative attitude to a construction project by stakeholders can severely obstruct its implementation. Such obstruction will
cause cost overruns and exceeded time schedules due to conicts and controversies concerning project design and implementation.
A case study consisting of two projects has been undertaken to investigate how the problems of managing the concerns of stake-
holders present themselves in an actual construction project. A method of stakeholder mapping, together with the power/interest
matrix, has been used to identify stakeholders and their inuence on the projects studied. Which problems arose, how were they
resolved, and what were the consequences of the solution? The case study shows that an evaluation of stakeholder demands and
inuence should be considered as a necessary and important step in the planning, implementation, and completion of any construc-
tion project.
Keywords: Stakeholders; Management; Construction; Project; Power/interest matrix
1. Introduction
In any project, and especially in construction pro-
jects, many dierent and sometimes discrepant interests
must be considered. Representatives of these interests
are referred to as the project stakeholders. A project
stakeholder is a person or group of people who have a
vested interest in the success of a project and the envi-
ronment within which the project operates [1]. The
implication is that a stakeholder is any individual or
group with the power to be a threat or a benet [2].
The demands of the community exert pressure on
organisations to develop new methods of working and
communicating with stakeholders [3]. A negative percep-
tion by stakeholders can severely obstruct a construction
project. Inadequate management of the concerns of
stakeholders often leads to conicts and controversies
about the implementation of a construction project. To
avoid this, project managers should try to acknowledge
the concerns of all stakeholders and in a dialogue seek
to reconcile conicting interests [46].
A construction project aects stakeholders in both
positive and negative ways. The positive aects can be,
better communication, better housing or higher stan-
dards of living. The negative side of a construction pro-
ject can be deterioration of the physical environment for
the aected stakeholders [7]. The demands of dierent
stakeholder groups are various. A construction project
can, for instance, be of use to one stakeholder group
and have a negative impact on another. Understanding
each others viewpoints helps to build relationships, thus
avoiding preconceived ideas and assumptions [3]. Thus,
project management must be able to analyse the various
demands presented by stakeholders so that communica-
tion between them is facilitated.
doi:10.1016/j.ijproman.2005.02.002
*
Corresponding author. Tel.: +46 046 222 7420; fax: +46 046 222
4414.
E-mail addresses: stefan.olander@bekon.lth.se (S. Olander), anne.
landin@bekon.lth.se (A. Landin).
PROJECT
MANAGEMENT
2. Methods
A case study [8] was conducted, for which two pro-
jects in Sweden, were chosen to investigate how the
problem of managing stakeholders presents itself in
a construction project. The projects are a housing pro-
ject and a railroad project. The main reasons for select-
ing these is that they have both had diculty in the
management of demands and needs of external stake-
holders, and that they are dierent in size and nature.
This allows the identication of factors in the external
stakeholder management that are independent of the
size and nature of the project.
The main source of information in this case study was
interviews with various stakeholders in the project,
project managers, project owners, architects, local
authorities, the aected residents, politicians, and repre-
sentatives of various interest groups. The interviews were
conducted as semi-open interviews. The structure of the
interview and relevant issues were predened. In addition
to the interviews, ocial documents and investigations
concerning the project were examined, in order to acquire
the ocial view of the project, and to gain input to the
structure of the interviews. Articles fromlocal newspapers
were also examined to gain a picture of the role of the
media. The information obtained was then structured
and analysed to accumulate input for the evaluation tools
that are presented.
2.1. Stakeholder mapping and the power/interest matrix
Various stakeholder mapping techniques exists, see
for instance [1,9,10]. Mendelow [9] has presented a model
of environmental scanning in the context of the stake-
holder concept, and includes the dynamism of the envi-
ronment and the power of the stakeholder relative to
the organisation or, as in this case, the project. Accord-
ing to Mendelow [9], the basis on which stakeholders
possess power relative to an organisation is liable to
change depending on the impact which the stakeholders
environment has on the stakeholders basis of power.
The model that is presented consists of a grid where
power and dynamism are relevant factors. Power ranges
from low to high, and dynamism ranges from static to
dynamic. A static environment implies that there is little
likelihood of the stakeholders to alter their power base,
and a dynamic environment may lead to alterations in
the bases from which stakeholders derive their power.
Johnson and Scholes [10], simplied and adapted
Mendelows [9] model and changed the axes of dyna-
mism to instead measure interest, and thus formulated
the power/interest matrix (see Fig. 1) which analyses
the following questions:
How interested is each stakeholder group to impress
its expectations on the project decisions?
Do they mean to do so? Do they have the power to
do so?
By grouping stakeholders in the power/interest ma-
trix, project management can produce a better picture
of how communication and relationships between stake-
holders has aected the project and its implementation.
In combination with the power/interest matrix, Bonke
and Winch [11] developed the stakeholder map, which
also analyses the problems and the proposed solutions
the dierent stakeholders have in the implementation
of the project. The stakeholder map includes: stakehold-
ers, divided into proponents and opponents, problems
identied by the stakeholders, and their suggested solu-
tions to the problems. If the stakeholder map, as Bonke
and Winch [11] present it, is added to the actual outcome
of project decisions and the consequences of the out-
come, a tool for evaluating the stakeholder management
process for a project, can be created.
From the information gained from the projects stud-
ied the stakeholders have been identied, and from this
the most relevant stakeholders have been evaluated in
the stakeholder map and the power/interest matrix, at
dierent stages of the project, up to the point when con-
struction work on site was given approval to start. The
stages that will be used in the analysis are; the feasibility
and conceptual design stage, the formal planning stage,
and the stage of appeals. To place the stakeholders in
the power/interest matrix both their relative power over
the project, and their interest to impose their expecta-
tions on the project have been judged on a scale from
0 to 10. The grades of power and interest are shown
and motivated in the analysis of the cases studied.
3. The case study
In this section, the inuence stakeholders have had on
project decisions is described by the stakeholder map
Keep
Satisfied
Minimal
Effort
Keep
Informed
Key
Players
Level of interest
Power
low
low
high
high
Fig. 1. Stakeholder mapping, the power/interest matrix [10].
322 S. Olander, A. Landin / International Journal of Project Management 23 (2005) 321328
and analysed with the power/interest matrix for each of
the cases studied.
3.1. Project 1: Lund, Sweden, a housing project consisting
of 60 apartments
The developer purchased the property in the late
1980s. The property was previously an old residence sit-
uated in a large park, which had been inhabited for
many years and was partly in disrepair. However, both
the park and the old residential building had a cultural
value for the community, and in addition, the park also
had some recreational value.
The initial stakeholder analysis of the project identi-
ed six major stakeholder groups. The real estate devel-
oper and their project managers. The Municipality,
which was interested of how this development would af-
fect the community as a whole, and in addition was
responsible for the formal planning process. Interest
groups for senior citizens, since the project was targeted
at senior citizens as the future tenants. The residents in
the vicinity of the project, who were concerned about
how the project would aect their living environment.
The National Government, that acted as the last instance
of appeal in the formal planning process. Interest groups
for the preservation of the cultural and historical image of
the city.
3.1.1. The feasibility and conceptual design stage (phase
1) 19881990
Before 1988, the property had been inhabited for a
long time and in 1988 plans to build a 12-storey hotel
were rejected by the municipality. A real estate devel-
oper then purchased the property in the winter of
1989. The real estate developer then developed a pro-
ject of two 9-storey apartment buildings for senior cit-
izens. They informed the municipality, the interest
groups for senior citizens and the residents in the
vicinity of the project. The real estate developer met
a positive attitude from the municipality and the inter-
est groups for senior citizens, mainly because there
was a need for new housing in the community.
However, from the residents in the vicinity they met
a strong opposition even at this early stage. The
neighbouring residents had fought hard against the
12-storey hotel, because the height of the building
would aect the immediate environment, which con-
sisted largely of small, one family houses. When the
developer, just one year later, presented a project of
two 9-storey buildings many of the residents in the
vicinity believed it to be even worse than the hotel
plans. In spite of this the real estate developer pur-
sued the project and together with the municipality
began work on a local community plan for the
project.
3.1.2. The formal planning stage (phase 1) 19901992
When the work on the local community plans was
started, the height of the buildings was the key issue.
The residents in the vicinity acted forcefully in an at-
tempt to stop the project. Their main arguments were
that the buildings would greatly aect their immediate
environment with regard to shadow, and that their fu-
ture neighbours would be able to see into their homes
and gardens. At this stage another inuential, and
opposing, group interested in preserving the cultural
and historical image of the city came on to the scene.
The city of Lund is approximately a thousand years
old, which means that there are a strong interests to pre-
serve the old image of the city. This project would block
the view of the 12th century cathedral, and thus change
the cultural image of the city. In spite of the growing
concerns about the negative aspects of the project, the
municipality still believed that the positive aspects out-
weighed the negative. The future tenants, represented
by the interest groups for senior citizens argued for the
need of the project, and supported the views of the real
estate developer.
3.1.3. The stage of appeals (phase 1) 19921993
After the local community plan had been approved
by the municipality the residents in the vicinity appealed
against the decision. The appeal was rejected in the rst
instance, and the residents appealed again to the second
and last instance of appeal, the National Government.
At this stage there were growing concerns about the pro-
ject from the municipality and some politicians ques-
tioned the decision to approve the plan, mainly on the
grounds of the eect of the project on the cultural and
historical image of the city. The debate continued during
the whole process of appeals, and the real estate devel-
oper, pressured by the opposition, presented alternative
solutions for the project. In December 1993, the Na-
tional Government presented its decision. The suitabil-
ity of the project was strongly questioned and grounds
were found to approve the appeal. The municipality
had at a late stage of the planning process made changes
to the plan, without displaying the change publicly,
which for the National Government, was a sucient
reason to reject the local community plan, and thus
the project. The real estate developer was thus forced
to present a totally revised proposal.
3.1.4. The feasibility and conceptual design stage (phase
2) 19941995
Since the decision of the National Governments to re-
ject the local community plan was based on a formal er-
ror, and not on the design of the buildings, the real estate
developer could have kept the initial design and seek its
approval. However, due to the strong criticism against
the project at the earlier stage, and because the real
estate development company had made a change in their
S. Olander, A. Landin / International Journal of Project Management 23 (2005) 321328 323
management sta, they decided to start afresh. The new
project manager hired an architect who earlier had been
employed by the municipality, to present alternative
solutions for the project. In addition to this, a nearby
petrol station had been removed, which allowed for the
possibility to build more buildings on the property. Thus,
the new solution was ve 56 storey buildings. This solu-
tion harmonised better with neighbouring buildings and
the cultural and historical image of the city. After this
new solution was presented the only remaining oppo-
nents was the residents in the vicinity. They were still
of the opinion that the buildings were too high, and with
the increase in the number of buildings, they also felt that
the area would become too densely built up. These argu-
ments, in addition to their distrust of the real estate
developer, decided them to continue with their strong
opposition to the project.
3.1.5. The formal planning stage (phase 2) 19951998
As stated above, the only strong opposition that re-
mained was the group of neighbouring residents. The
community had at this stage no problem with approving
the local community plan in accordance with the sugges-
tion of the real estate developer. However, the munici-
pality was now concerned with avoiding the kind of
formal errors on which the earlier plan had been re-
jected. Thus, the municipality saw the importance of
conducting a thorough analysis before approving the
project. After considering the various views the munici-
pality approved the plan in February 1998.
3.1.6. The stage of appeals (phase 2) 1998
The neighbouring residents were still not in agree-
ment with the decision to approve the project, and they
again appealed to the highest instance, the National
Government. However, none of the instances saw any
problems with the plan and the project, and thus re-
jected the appeals.
3.1.7. Consequences
The main consequence for the project was the time
delay; the project took 12 years from the rst ideas to
completion. The time delay and the extensive changes
to the project also had the eect that a signicant
amount of already committed resources became obso-
lete, representing approximately 510% of the total pro-
ject costs. However, the developer identies at least one
important positive eect, which is that, by taking the
views of the majority of the stakeholders, with the
exception of the neighbouring residents, into consider-
ation the nal solution is better than the original
proposal.
3.1.8. The power/interest matrix
The power and interest (see Fig. 2) of the stakehold-
ers were very dierent between phase (1) and phase (2).
Level of interest
Power
1
1
10
10
6 3
2
1
Level of interest
Power
1
1
10
10
5
6
3
2
1
Level of interest
Power
1
1
10
10
5
4
3
2 1
Level of interest
Power
1
1
10
10
3
2
1
Level of interest
Power
1
1
10
10
3
2
1
Level of interest
Power
1
1
10
10
4
3
2
1
5 5
The feasibility and conceptual
design stage (phase 1) 1988-1990
The formal planning
stage (phase 1) 1990-1992
The stage of appeals
(phase 1) 1992-1993
design stage (phase 2) 1994-1995
The formal planning
stage (phase 2) 1995-1997
(phase 2) 1998
1. The real estate developer, 2. The Municipality, 3. Residents in the vicinity, 4. The national government, 5. Interest groups for the preservation of the
cultural and historical image of the city, 6. Interest groups for senior citizens
Fig. 2. The power/interest matrix for project 1.
In phase (1), the proposed project was more controver-
sial, which added to the number of interested stakehold-
ers. This led to the eect that one stakeholder group
could use another in order to extend their power base.
This was evident in the relation between the residents
in the vicinity and the National Government, where
the residents put forward all arguments against the pro-
ject using certain personal channels in order raise the
interest of the National Government to reject the project
proposal. This led to the eect that the project managers
lost their power base. However, the project managers
had learned their lesson. In phase (2), they tried to
acknowledge as many concerns as possible when they
outlined the revised project. This had the eect that all
opposing stakeholders, except the neighbouring resi-
dents, were satised. The remaining process was to deal
with one single stakeholder group, the neighbouring res-
idents, whose power base had now dramatically
decreased.
3.2. Project 2: Lund, Sweden, railroad project, the con-
struction of a two way railroad track through the town
centre
In the late 1980s, the Swedish government decided to
transform the west coast railway (between Malmoe and
Gothenburg) from a single to a double track railroad.
The whole route passes through a number of communi-
ties, and one of these is the city of Lund.
The initial stakeholder analysis of the project identi-
ed seven major stakeholder groups. The National Rail-
road Administration the developer of the project. The
Municipality, which was interested of how this develop-
ment would aect the community as a whole, and in
addition was responsible for the formal planning pro-
cess. The railroad companies, which would be responsi-
ble for the trac along the west coast railroad. The
residents in the vicinity of the expanded railway, which
were aected by the increase in railroad trac. The Na-
tional Government, which was both the initiator of the
project, and the last instance of appeal for the formal
planning process. The National Board of Housing and
the Swedish Rescue Services, were two of the national
authorities to which the project was referred for
consideration.
3.2.1. The feasibility and conceptual design stage, 1990
1994
In 1990, the rst investigations for the expansion of
the west-coast railroad through the city of Lund,
according to the requirements set out by the National
Government. At this early stage the National Railroad
Administration made the decision that the best alterna-
tive was to expand the railroad along the existing
single-track route. However, this decision was made
on the basis of insucient analysis of other possible
alternatives, and of how the proposed alternative af-
fected the various stakeholders. The only stakeholders
really being considered at this early stage were the rail-
road companies, which would manage the trac on this
railroad, and for them an expansion along the existing
route was the most rational choice. The problem with
this alternative was that it would aect a large number
of residents along the railroad, who would experience a
signicant increase of railroad trac in the vicinity of
their homes. However, at this early stage they were
not involved in the process. The rst public consulta-
tions were held in November 1991, when a proposal.
Essentially already decided on, was presented to the
neighbouring residents. Due to the lack of public
involvement an interest group consisting of concerned
residents had been formed in 1990 to follow the pro-
cess. At this early stage they had little or no inuence
on the process. In 1993, the municipality was involved
because a community plan was required. By acting
through the municipality the residents raised the ques-
tion for the need of alternative solutions. As a base
for the forthcoming local community plan, the munici-
pality demanded additional investigations concerning
alternative solutions from the project developer, the
National Railroad Administration. An investigation
of ve alternatives was undertaken by an independent
consultant, who showed that the proposed alternative
was the best and most rational solution. One of the
most compelling arguments for the expansion along
the existing route was the developed infrastructure al-
ready in place with train stations, and that large areas
of the city had been developed to coordinate with this
existing infrastructure. The negative aspect was still
that the proposed alternative would aect a large num-
ber of residents along the existing route.
3.2.2. The formal planning stage 19941997
Based on the investigation undertaken the munici-
pality decided to produce a local community plan based
on the proposed alternative along the existing route. At
this stage the municipality had sole inuence on how
the project would develop, and the project developers
role in this stage was to provide the information re-
quired by the municipality to make their decisions.
However, at this stage the number of interested resi-
dents increased dramatically. This was probably due
to the lack of information at the early stages and that
it was rst at this stage that the public became aware
of the proposed development. The interest group of
concerned residents increased their member base and
also changed their perspective to actively oppose the
proposed alternative and to put forward an alternative
solution. This solution had been investigated as one of
the ve solutions and had been rejected because it
necessitated a whole new infrastructure and because
of its technical and economic disadvantages. These
ndings were questioned by the residents arguing that
the alternatives had not been fairly investigated. In
1995, the municipality and the project managers
decided to make an additional investigation to compare
the alternative proposed by the project developer with
that proposed by the residents. Again this investigation
showed that it was more rational to expand along the
existing route. However, at this stage of the project
the relation between, on the one hand the project devel-
oper and the municipality, and on the other the resi-
dents along the existing route, deteriorated radically.
The outcome of this was that every investigation made
from this point on, which proposed the expansion
along the existing route, was regarded as biased by
the residents. In addition to this, two national author-
ities, the National Board of Housing and the Swedish
Rescue Services had raised concerns about living condi-
tions and safety issues along the proposed route. In
spite of these concerns, the municipality approved the
local community plan for the expansion along the exist-
ing route in the summer of 1997.
3.2.3. The stage of appeals 19972003
Directly after the municipalitys approval of the pro-
ject the residents in the vicinity of the proposed route
appealed against this approval. The main arguments
for the appeal were that all possible alternatives had
not been objectively investigated and that a better solu-
tion existed. As a supplement to the appeal the residents
had conducted their own investigation that showed the
benets of their alternative relative to that proposed
by the developer and the municipality. In the autumn
of 1997, the rst instance of appeal decided to reject
the appeals and approve the municipalitys decision.
However, the residents were not satised and appealed
to the second and nal instance of appeal, the National
Government. This raised a dilemma for the National
Government, as on the one hand it was supposed to pro-
cess the appeal objectively, and on the other hand it was
the initiator of the project. The dilemma grew worse
when two national authorities, the National Board of
Housing and the Swedish Rescue Services, stated that
the appeal should be approved and that the local com-
munity plan should be revised. At this stage, the munic-
ipality considered their job as completed and did not
engage in the process of appeals, while the project devel-
oper found themselves in a situation where they could
only wait. The National Government was now in a situ-
ation where they could not approve the proposed devel-
opment because of the concerns that were raised by the
opposing stakeholders, but they could not reject it either
because the expansion of the west-coast railroad was a
prioritised national concern. This situation together
with an active opposition from the residents, led to a
process of appeal that lasted for six years. Six years that
simply prolonged the project with out producing any
added value.
3.2.4. Consequences
The main consequence for the project was a delay of
approximately seven years, mainly because of an adverse
public opinion, expressed by the residents in the vicinity,
which led to a time-consuming juridical process. In addi-
tion to this, the media coverage was almost always from
the viewpoint of the residents in the vicinity, which led
to bad a press and a generally bad reputation for the
project.
3.2.5. The power/interest matrix
The power and interest of the stakeholders changed
during the course of the project (see Fig. 3). In the
rst stage the initiative lay largely with the developer.
They had the interest and also the power, and the
only other key were. However, the project managers
did not acknowledge the concerns of other stakehold-
ers, which basically raised both their interest and their
power base. In the second stage, this development led
to an increase of stakeholders with both an interest
and the power to inuence the project. The formal
power lay by legislation with the municipality, but
Level of interest
Power
1
1
10
10
5
4
3
2
1
Level of interest
Power
1
1
10
10
5
4
3
2
1
Level of interest
Power
1
1
10
10
5
4
3
2 1
6
6
design stage 1990-1994
The formal planning
stage 1994-1997
1997-2003
1. The National Railroad Administration, 2. The Municipality, 3.The Railroad companies, 4. Residents in the vicinity, 5. The national government
6. The National Board of Housing, and the Swedish Rescue Services Agency
Fig. 3. The power/interest matrix for project 2.
the project managers lost some of their power by act-
ing reactively instead proactively. The residents in the
vicinity of the proposed expansion dramatically in-
creased their interest to inuence, due to a their per-
ception of having been disregarded by the project
developer. They used this interest, in the stage of ap-
peal, to increase their power base and became a rele-
vant key stakeholder. In addition to this there were
stakeholders with a relatively low interest to inuence,
but with a high power to do so. The project developer
now found that they lacked control and had to de-
pend on other stakeholders in order to be able to pro-
ceed with the project.
4. The role of the media
The media hold a unique position in the process.
They cannot really be dened as a stakeholder, be-
cause they have no actual stake in the project. How-
ever, the media can have a tremendous eect on the
projects outcome, both positive and negative. Thus,
the media are an important factor that must be con-
sidered, which in practice gives them the status of an
important stakeholder.
The media coverage of both projects studied has been
extensive, and mainly of an oppositional nature. The
opponents used the media eectively to express their
opinions, and many prominent local spokespersons,
architects, professors, politicians, etc., were eager to give
their views on the project. The result was that the project
management had to handle a negative press coverage, in
addition to an already dicult process of managing the
views of the various stakeholders, with the media having
a strong inuence on decision makers, politicians, and
local and national authorities.
It is dicult to estimate the eect of the media cover-
age, but an open and trustworthy communication with
the media and the aected stakeholders is essential. In
both cases, the project management has found itself
ghting a rearguard action against negative public opin-
ion and a negative press coverage, mainly because the
stakeholders had become dissatised with the informa-
tion they had been given, and due to the lack of trust
this created.
5. Conclusions
An important issue for a project management team is
to identify those stakeholders who can aect the project,
and then manage their diering demands through good
communication in the early stages of a project. In both
of the projects studied various stakeholder groups have
inuenced project decisions beyond the control of the
project management team.
Four major lessons for project managers could be
learned from the projects studied here. (1) To investigate
all possible alternatives and solutions to realise the objec-
tives of the project, not only from the quantitative as-
pects of technology and economy, but also from the
more qualitative aspects of potential inuence from
stakeholders. (2) To clearly dene all the positive and
negative arguments about the chosen alternative in rela-
tion to the other alternatives investigated, in order to be
regarded as trustworthy by those stakeholders who are
negatively aected by the project. (3) The stakeholders
base of inuence is not static. The stakeholder analysis
must be conducted and updated during the entire life cy-
cle of the project, with the purpose of gaining knowledge
about the potential inuence various stakeholders have
at dierent stages of the project. (4) Prior to any major
decision to proceed into a new phase of the project an
analysis of how the decision aects the dierent stake-
holders should be made in order to be proactive in the
stakeholder management process.
The model presented here should be seen as a tool to
identify and evaluate the inuence of project stakehold-
ers on project decisions. The purpose of this is, for
forthcoming projects, to gain a knowledge of how pro-
ject decisions aect stakeholders and how stakeholders
aect project decisions in the construction process. A
hypothesis is that the model of stakeholder mapping
and the power/interest matrix could be a useful tool
to, a priori, judge the potential inuence of identied
project stakeholders. One way to achieve this would
be to add the dynamism factor as described by Mende-
low [9], and thus be able to judge the intensity of the
stakeholder scanning process for the project. However
there is a need for further studies in order to determine
how useful these models are in that context. This would
indicate that a more ambitious strategy in the external
stakeholder management process would increase the
possibility of resolving conicts of interests. However,
a more ambitious strategy also requires more resources.
Hence, the dilemma for the project management is to
balance the use of resources with the appropriate strat-
egy towards each individual stakeholder group.
6. Further research
Further research is needed to formulate a more gen-
eral model of how an external stakeholder management
process should be conducted for construction projects.
The model should be able to combine identication of
stakeholders, assessment of their needs and their de-
mands on project implementation, assessment of their
potential inuence on project decisions, choice of stake-
holder strategy, assessment of consequences of project
decisions, and evaluation of the stakeholder manage-
ment process. Furthermore, studies need to be conducted
of how the stakeholder management process relates to
other project management processes, for instance, cost
and risk management.
Acknowledgements
We thank SBUF (Swedish Builders Development
Fund), the National Railroad Administration and the
National Road Administration for their nancial sup-
port and all those who at dierent stages of the project
gave their time and eort with interviews and discus-
sions. We thank the anonymous referees for their valu-
able comments which have had a major impact on the
structure of the article.
References
[1] McElroy B, Mills C. Managing stakeholders. In: Turner RJ,
Simister SJ, editors. Gower handbook of project management.
3rd ed.. Gower publishing limited; 2000. [Chapter 42].
[2] Gibson K. The moral basis of stakeholder theory. J Bus Ethics
2000;26:24557.
[3] Watson T et al.. Issues negotiation investigating in stakehold-
ers. Corp Commun: An Int J 2002;7(1):5461.
[4] Susskind L, Cruikshank J. Breaking the impasse consensual
approaches to resolving public disputes. USA: Basic Books Inc;
1987.
[5] Susskind L, Field P. Dealing with an angry public the mutual
gains approach. New York: Free Press; 1996.
[6] Davy B. Essential injustice. Wien: Springer Verlag; 1997.
[7] Olander S. Consensual approaches to siting controversy. In:
Proceedings from the 10th international symposium on the
organization and management of construction. University of
Cincinnati, CIB: CRC Press; 2002.
[8] Yin RK. Case study research design and methods, applied social
research methods series, vol. 5. Sage Publications; 1994.
[9] Mendelow A. Environmental scanning: the impact of stake-
holder concept. In: Proceedings of the second international
conference on information systems, December 1981. Cambridge,
Mass; 1981.
[10] Johnson G, Scholes K. Exploring corporate strategy. Lon-
don: Prentice Hall Europe; 1999.
[11] Winch G, Bonke S. Project stakeholder mapping: analysing the
interests of project stakeholders. In: The frontiers of project
management research. USA: Project Management Institute
(PMI); 2002. [Chapter 23].
Framework for project managers to manage construction safety
Evelyn Ai Lin Teo
*
, Florence Yean Yng Ling, Adrian Fook Weng Chong
Department of Building, National University of Singapore, 4 Architecture Drive, Singapore 117566, Singapore
Received 24 February 2004; received in revised form 22 June 2004; accepted 3 September 2004
Abstract
This study proposes a Policy, Process, Personnel and Incentive (3P + I) framework which may help project managers manage
construction site safety. A postal survey of contractors in Singapore was conducted to test the framework. It is found that site acci-
dents are more likely to happen when there are inadequate company policies, unsafe practices, poor attitudes of construction per-
sonnel, poor management commitment and insucient safety knowledge and training of workers. It is recommended that project
managers pay more attention to the important factors identied in this study to help them enhance performance at construction sites
and reduce the frequency of accidents.
Keywords: Site safety; Safety policy; Site accidents; Incentive
1. Introduction
Construction sites are generally complex and some-
times unsafe. It is complex because of extensive use of
sophisticated plants, equipment, modern methods of
construction, multidisciplinary and multitasked aspects
of its project workforce [1,2]. Generally, construction
sites are still one of the most dangerous workplaces be-
cause of high incidence of accidents. In Singapore, the
number of deaths and injuries of the construction indus-
try has not been satisfactory (see Fig. 1) although gov-
ernment eorts have been made to deal with this
problem. Construction sector contributes less than
10% to the gross domestic product, but take up more
than 37% of all industrial accidents in Singapore. There-
fore, there is an urgent need to improve construction
safety.
The aim of this paper is to investigate how project
managers may increase the safety levels of construction
sites. The specic objectives of this paper are to: (1)
identify variables that signicantly aect construction
safety and (2) propose a framework for project manag-
ers to use and manage construction safety.
The rst objective is important because the root
causes of accidents have to be identied for project man-
agers, so that they can take preventive and remedial ac-
tions to minimise accident occurrences. The proposed
framework may help project managers to manage con-
struction safety. When safety aspects are well managed,
the frequency of accident occurrences may be reduced.
The shortcoming of this framework is that it does not
indicate how safe a site is at a given point in time. A fu-
ture study to develop a scoring system to ascertain the
safety level of a construction site will be carried out to
develop such a system.
The next section is a brief literature review of vari-
ables that aect the construction safety as a whole.
The research method is then presented, followed by
the results. A discussion follows, whereby the factors
that aect construction safety signicantly are identi-
ed. A framework to manage construction safety is
then proposed, followed by conclusions and
recommendations.
doi:10.1016/j.ijproman.2004.09.001
*
Corresponding author. Tel.: +65 68741008; fax: +65 67755502.
E-mail address: bdgteoal@nus.edu.sg (Evelyn Ai Lin Teo).
PROJECT
MANAGEMENT
2. Literature review on factors aecting site safety
From the literature, many factors that may contrib-
ute to site safety were identied. These areas are now
briey reviewed.
2.1. Policy factor
The safety legislation and policies have a great impact
upon the safety level of a construction worksite. Legisla-
tion forms a framework in which health and safety is
regulated and controlled [3]. All project managers have
to follow the rules and regulations duly and punish-
ments to be meted out to those who out them.
In Singapore, the construction site safety legislation is
governed by the requirements stipulated under the Fac-
tories Act (Chapter 104) and the Factories (Building
Operations and Work of Engineering Construction)
Regulation, 1994. The Factories (Building Operations
and Works of Engineering Construction) Regulation re-
quires all occupiers of construction worksites, which
have contract values of S$10 million (US$1 = S$1.75)
or more to implement a Safety Management System
(SMS) specied under the 1999 Code of Practice for
Safety Management System for Construction Worksites
(CP 79).
Apart from the weak safety organisation culture with
ill-dened safety roles and inadequately developed
safety procedures, the lack of sound safety policies
would result in poor safety performance of construction
sites. Legislation and its enforcement do aect construc-
tion safety to a considerable extent [4]. As such, safety
legislation has to be taken seriously when planning job
activities and setting up company policies.
Another policy-induced factor is the promotion of
the safety recognition of construction rms through
the Occupational Health and Safety Management Sys-
tem (OHSMS) certication. This is achieved by ensuring
that the rms full the requirements of the Occupational
Health and Safety Assessment Scheme (OHSAS) 18001.
The main objective of OHSMS certication scheme is to
encourage and enhance safety awareness, promote safe
work practices and raise the safety standards of the con-
struction industry.
With an increase in the reliance on insurance, con-
tractors tend to shift the liability for compensation to
insurance companies [4]. They may not provide proper
site safety training and supervision for the workers.
2.2. Process factor
Process factors refer to the process of carrying out
works by construction personnel that may eventually
be harmful to their well-being and safety. Some organi-
sations or individuals had in fact, committed an unsafe
act unknowingly while carrying out their tasks and jobs
[5].
A major concern for managing process factors is the
eectiveness of control over the large numbers of sub-
contractors on construction sites due to diversication
of activities. Thus, with higher numbers of subcontract-
ing, the chances of accident occurrences will be more fre-
quent [1]. In Singapore, a chain of subcontractors (third/
fourth party subcontractors) is commonly observed. As
such, the probability of the lack of communication,
coordination and control will increase [1,3]. Further-
more, main contractors may shift all safety responsibil-
ities to subcontractors and may not ensure that the
Fig. 1. Industrial accidents by degree of incapacity and industry (19942003).
330 Evelyn Ai Lin Teo et al. / International Journal of Project Management 23 (2005) 329341
subcontractors are capable of providing a safe working
environment [6].
Dierent construction methods need to meet dierent
safety standards and expectations. Accidents on con-
struction sites are caused by: unsafe working conditions
at heights, stepping on, striking against or tripping over
objects, poor lighting conditions, burial by earth col-
lapse during excavation, collapse of scaoldings and
working platforms, hazards in lifting operations, elec-
trocution, re hazards, lack of proper access and inade-
quate education and training. In addition, engagement
of poor tools and equipment can also cause accidents
[7].
Eective communication and information transfer
between management and employees will yield better
safety standards and enhance the achievement of safety
policies [8]. The eectiveness of such transferring medi-
ums is subject to the dierent means, like information
transfer. The safety commitment of the management
has been found to be a determinant to the various means
of communication and information transfer to all levels
of the construction project [9].
In addition, other process activities like the han-
dling of hazardous materials including toxic and
chemical wastes could spell danger. Thus, personnel
should be given proper training, provided with ade-
quate information on the type of chemicals which he
is handling and supplied with the correct protective
equipment [10]. Other hazardous materials including
explosives, ammable liquids, corrosive and acidic
substances need to be handled with professional train-
ing and care.
2.3. Personnel factor
Personnel factors refer to issues pertaining to the hu-
man aspect of the construction activities. One such fac-
tor points to the safety behaviours and attitudes of the
management and workers within an organisation. The
safety behaviour and attitudes represent its safety
culture. Safety culture is a subset of organisational cul-
ture, where the beliefs and values refer specically to
matters of health and safety [11]. Thus, the safety cul-
ture in an organisation is dependent upon the safety
commitment of the management and workers towards
its safety promotions and campaigns. One of the ways
to investigate the organisational safety culture is
through conducting employee perception surveys as a
tool to detect dierences in their attitudes and to test
the eectiveness of a safety programme [12].
The provision of safety training for employees is an-
other important aspect for consideration under the per-
sonnel factors. A study on the eect of rst aid training
on Australian construction workers concluded that
training has a positive preventive eect on workers to
avoid injury [13]. It has also been found that workplace
injuries would be reduced if workers received rst aid
training [14].
Accidents may occur because of poor attitudes and
bad behaviours of the workers, which are dicult to
monitor and control. There is a positive link between
safety performance and workers attitudes [15]. In addi-
tion, negative behaviours and attitudes have prompted
most workers not to put on their personal protective
equipment whilst working on site [5]. In this regard,
the workers need to possess the correct skills and knowl-
edge for the nature of work and to be motivated to be-
have safely [16].
Another important aspect towards enhancing safety
culture as mentioned earlier is through the safety com-
mitment displayed by the top management of the
company. Surveys commissioned by the Health and
Safety Executive (HSE) revealed that 75% of all fatal
accidents in the building and civil engineering indus-
tries in the United Kingdom are generally caused by
ineective management action taken [17]. It is found
that large-scaled construction companies generally
having better safety performance due to the high level
of safety support and commitment shown from the
top management [18]. Research found that the reduc-
tion in accidents would be achieved when top manage-
ment takes an active interest and is dedicated to safety
enhancement as well as maintaining good safety
standards [19].
2.4. Incentive factor
Previous studies have shown the advantages and dis-
advantages of introducing incentives to improve site
safety. Some studies in favour of the scheme have pro-
ven that a reduction in construction site accidents and
injuries has been achieved [20,21]. Other studies showed
that safety indices did not improve despite of the intro-
duction of safety incentives [22]. This led to the deduc-
tion that only certain types of incentives will improve
site safety [23].
It has been argued that incentives do not have a pos-
itive impact on safety performance [24]. Some rms with
no incentive programmes have been found to have bet-
ter safety records than those with incentive programmes.
Incentives may be eective in reducing workplace inju-
ries, but this may depend on how the incentives are
being allocated.
At least, not all contractors agree that incentives are
vital in improving safety performance. Those contrac-
tors who agree to the introduction of incentives are gen-
erally in favour of monetary incentives over other types
of incentives. These monetary incentives are either used
for giving monetary returns for safety performance of
personnel and for supervisory safety incentives plan
[25]. Nonetheless, safety incentives do not necessarily
produce the desired outcome; it is dependent upon the
Evelyn Ai Lin Teo et al. / International Journal of Project Management 23 (2005) 329341 331
dierent relationships of the groups and individuals
expectations and reactions towards safety incentives
[26].
2.5. Summary of literature review
The review above uncovered some 50 potential attri-
butes that may aect construction safety. These studies
usually investigated a few specic factors which aect
construction site safety and the safety performances
of construction personnel. These previous works did
not provide a holistic framework that may help project
managers handle the various policy, process, personnel
and incentive aspects that may aect construction
safety. This study therefore aims to ll this gap, by
proposing and testing a framework to manage con-
struction safety.
Besides the potential attributes identied from the
literature, a review of the best practices of the Safety
Management Systems adopted in the USA, UK, Ja-
pan, Hong Kong and Australia was undertaken, and
more attributes were identied. Using content analysis,
the potential variables were grouped into four major
headings: policy, process, personnel and incentive.
These four major headings form the framework for
managing site safety (see Fig. 2). Policy, Personnel
and Incentive are organisational factors. Process is a
technical factor. The draft framework and the poten-
tial attributes were shown to the local authority in
charge of safety, three safety consultants/auditors,
three developers and three contractors to ensure that
important issues/attributes have not been left out. This
wide consultation ensures that the data collection
instrument developed is robust.
3. Research method
A questionnaire was designed with the objective of
determining the more important variables that aect
site safety. In the questionnaire, respondents were re-
quested to provide information relating to safety as-
pects for one of the recently completed projects.
Respondents were asked to rate the extent to which
each of the 50 variables inuenced the site safety per-
formance of the project, on a ve-point Likert scale,
where 1 = not important at all and 5 = very impor-
tant. The questionnaire also invited respondents to
state any other factors that aect construction safety,
and to rate these factors.
The population comprised contractors that are oper-
ating in Singapore. The sampling frame comprised the
1469 general building contractors registered with the
Singapores Building and Construction Authority
Policy Factors (Po)
1. Understanding and implementation of SMS
2. Understanding and participation in OHSMS
3. Understanding and implementation of permit-to-work
system
Incentive
Factors (I)
1. Monetary
incentives
2. Non-
monetary
incentives
3. Disciplinary
action
Process Factors
(Pr)
1. Quality of
subcontractors
2. Understanding
and
implementation of
safety procedures
3. Carrying out
work in a safe
manner
4. Carrying out
work in a
professional
manner
5. Type and
method of
construction
Personnel
Factors (Pe)
1. Management s
attitude towards
safety
2. Supervisors and
workers attitudes
towards safety
3. Contextual
characteristics of
workers
Fig. 2. 3P + I framework for managing construction safety.
(BCA). Of these, questionnaires were sent by post, with
self-addressed and pre-stamped envelopes, to 420 ran-
domly selected building contractors.
4. Data sample characteristics
From the 420 questionnaires sent out, 61 responses
were received within two months (response rate of
15%). One questionnaire that was not substantially com-
pleted was discarded. Data from the remaining 60 usa-
ble returned questionnaires were checked, edited,
coded and analysed.
The designation of the respondents were: upper
management (20%), middle management (48%): safety
personnel (17%): technical sta (157%). Upper man-
agement respondents comprised managing directors,
directors and general managers and senior project
managers. The middle management respondents were
project managers and assistant general managers.
Safety personnel consisted of environmental heath
and safety ocers/ managers, safety ocers, safety
supervisors and safety auditors. Technical sta refers
to supervisors, site coordinators and clerk-of-works.
The average working experience in the construction
industry of the respondents is 12 years. The minimum
and maximum working experience are one and 34
years, respectively. In addition, 60% of the respond-
ents have more than 10 years of working experience.
The prole of the projects is shown in Table 1. The
respondents did not provide any new factors to be in-
cluded in the study.
From the prole of respondents, it can be seen that
the majority of them are middle and senior manage-
ment, and have extensive experience in the construction
industry. It is expected that the data collected from them
are reliable.
5. Statistical analysis
Statistical analysis was conducted using the Statisti-
cal Package for Social Sciences (SPSS) software pack-
age. This paper reports the t test results and factor
analysis results. The results of the multivariate analy-
sis involving regression procedures are reported in an-
other paper [39].
Statistical t tests of the mean were carried out to
check the entire populations (contractors in general)
likely response to the issues raised in the questionnaire,
based on the samples ratings. The null hypothesis
H
0
:l = l
0
and the alternative hypothesis H
1
:l > l
0
was
set out, where l is the population mean. l
0
is the critical
rating above which the variable plays a very important
part in inuencing site safety of the project. In this
study, l
0
was xed at 3 because by the denitions given
in the rating scale, ratings above 3 represented impor-
tant or very important. The signicance level for the
one-tailed test was set at 0.05. The t test results are
shown in Table 2.
Besides the t test, factor analysis was undertaken.
Factor analysis is motivated by the fact that meas-
ured variables can sometimes be correlated in such
a way that their correlation may be reconstructed
by a smaller set of parameters, which could represent
the underlying structure in a concise and interpretable
form.
The 50 variables were input into the SPSS software
and factor analysis was carried out, to ascertain if there
is any further relationship among the many proposed
safety strategies to help the construction industry
achieve higher safety standards. Appendices ABCD
show the results of the factor analysis, factor loadings,
commonalities and eigenvalues for the extracted
factors.
The factor loadings are the correlation coecient be-
tween an original variable (the statements) and an ex-
tracted factor. Commonality (h
2
) is the proportion of
the total variance of a variable accounted for by the
common factors. The eigenvalue is a measure of how
Table 1
Prole of projects
Prole Frequency Percentage
Project type
Residential 24 40.0
Institutional and religious 11 18.3
Commercial 7 11.7
Industrial 4 6.7
Civil engineering 6 10.0
Unknown 8 13.3
Project size (USD)
Not exceeding $1.5M 20 33.3
$1.5M$4.99M 7 11.7
$5M$14.99M 16 26.7
$15M$30M 10 16.7
>$30M 7 11.7
Project duration
Up to 12 months 19 31.7
1324 months 29 48.3
>24 months 12 20.0
Full time workers
125 workers 20 33.3
2650 workers 12 20.0
5175 workers 8 13.3
>75 workers 20 33.3
Projects safety performance
Unsafe site 0 0.0
Permanent disability happened 1 1.7
Temporary disability happened 8 13.3
Minor accidents happened 26 43.3
Very safe site 25 41.7
Mean safety level 4.25
standard variables contribute to the principal
component.
The results of the t test and factor analysis are now
discussed.
6. Discussion on policy aspect
6.1. Findings of t test results
The t test results showed that all 10 variables under
policy (Po1.1 to Po1.10) are very important in ensuring
site safety (p < 0.000) (see Table 2). The three most
important variables are:
(i) Proper implementation of in-house rules and regu-
lations (Po1.7),
(ii) Understanding of in-house rules and regulations
(Po1.6), and
(iii) Proper implementation of SMS (Po1.3).
The results show that proper implementation and
understanding of in-house rules and regulations are
the two most important policy factors that inuence site
safety. This agrees with a previous study which found
that the absence of safety rules and regulations would
greatly aect the proper enforcement of site safety [27].
The safety performance of an organisation is inuenced
by the extent of which the workers perceived safety
rules and procedures [28].
However, rules and regulations will be rendered inef-
fective if they are not properly implemented and exe-
cuted. These must be understood by all and properly
implemented to prevent accidents. The setting up of a
monitoring regime and feedback system on the eective-
ness of safety rules and regulations should be the rms
top priority.
There are two levels of understanding by individuals:
(i) The understanding by the users (i.e., the workers
and personnel who are required to obey/follow
the rules and regulations);
(ii) The understanding by the implementers (i.e., the
originators and creators of the rules and regulations
who expect users to put them into practice within
the organisation).
Besides basic safety trainings received, the users
superiors should always remind users to strictly comply
with the safety rules and regulations and also be in-
formed on any safety updates and the general conse-
quences of breaking the rules and regulations.
On the other hand, the implementers are normally
the management and/or the safety committee of the
rm. They have to ensure that what they propose
must be realistic and achievable. Besides complying
with the requirements of the safety authorities, they
have to update themselves constantly on all new safe-
ty-related legislative laws, requirements and genuine
feedback from the users in order to further improve
the current rules and regulations and tailor them to
suit each specic characteristics of the construction
project.
6.2. Results of factor analysis
The factor analysis revealed that the 10 variables can
be classied into three factors; understanding and imple-
mentation of (1) safety management system (SMS); (2)
OHSMS certication; and (3) permit-to-work system
(see Appendix A).
The rst factor is important because of the obliga-
tory need to implement a SMS on the construction site
if the project value is S$10 million or more. As such,
the understanding of the SMS is fundamental for con-
struction rms, which implements the system and to
pass a periodic safety audit checks. Besides, the
achievement of OHSMS certication acts as a motivat-
ing factor to the rm and shows the rms eligibility for
participating in the tendering for public projects in
Singapore.
Permit-to-work system aects safety on site because
highly specialised and hazardous activities require per-
mits to operate and work on a regular basis. This is be-
cause of the need to monitor all hazardous activities and
only the qualied employees are issued with a permit to
work. Such permits are normally issued on a daily basis
(for example; working in conned spaces, hot-works,
sand blasting works, working from heights and such).
Previous studies have also found that the issuance of
such permits needs to be under the control of an author-
ised person [3,8]. Having short-term permits in place
would eectively monitor the safety performance of
the contractor and at the same time, enhance coordina-
tion between all occupiers of the worksite when involv-
ing in hazardous activities.
7. Discussion on process aspect
The t test results showed that all 16 variables relating
to process (Pr2.1 to Pr2.16) are very important to ensure
site safety (p < 0.000) (see Table 2). The three most
(i) Familiarity with type and method of construction
by safety ocers/supervisors (Pr2.5),
(ii) Identication of unsafe practices on site (Pr2.9),
and
(iii) Proper handling of tools, equipment and plants
(Pr2.13).
The results show that it is very important for safety
ocers to be familiar with dierent types and methods
of construction which yield dierent safety require-
ments. Thus, safety ocers must be well trained and
be experienced enough to deal with safety concerns of
dierent construction activities. A key prerequisite for
a safety programme to be successful is to incorporate
Table 2
Mean ratings and t test results
Rank Ref Factors inuencing site safety Mean t value p
Policy aspect
1 Po1.7 Proper implementation of In-house rules and regulations 4.53 15.89 0.000
2 Po1.6 Understanding of In-house rules and regulations 4.38 14.50 0.000
3 Po1.3 Proper implementation of SMS 4.42 14.19 0.000
4 Po1.1 Understanding of Factories Act 4.35 13.84 0.000
5 Po1.2 Understanding of CP79: 1999 (SMS) 4.36 13.30 0.000
6 Po1.9 Proper implementation of Permit-to-work system 4.29 13.29 0.000
7 Po1.8 Understanding of Permit-to-work system 4.30 12.44 0.000
8 Po1.10 Understanding of insurance policies 4.22 12.03 0.000
9 Po1.4 Understanding of OHSMS 4.03 9.27 0.000
10 Po1.5 Companys participation in OHSMS 3.85 6.31 0.000
Process aspect
1 Pr2.5 Familiarity with type and method of construction by safety ocers/supervisors 4.53 19.6 0.000
2 Pr2.9 Identication of unsafe practices on site 4.97 18.74 0.000
3 Pr2.13 Proper handling of tools, equipment and plants 4.47 18.22 0.000
4 Pr2.3 Co-ordination, control and management of sub-contractors 4.47 17.47 0.000
5 Pr2.6 Communication and information ow 4.43 17.14 0.000
6 Pr2.7 Understanding of safety procedures 4.45 16.64 0.000
7 Pr2.8 Proper implementation of safety procedures 4.53 16.40 0.000
8 Pr2.12 Good house-keeping 4.53 15.89 0.000
9 Pr2.4 Type and method of construction 4.30 15.59 0.000
10 Pr2.16 Tight control of hazardous activities on site 4.37 13.81 0.000
11 Pr2.11 Identication of hazardous and dangerous activities 4.30 13.54 0.000
12 Pr2.10 Proper implementation of safe practices on site 4.55 12.76 0.000
13 Pr2.14 Maintenance regime of tools, equipment and plants 4.08 12.50 0.000
14 Pr2.15 Technical competency of specialist subcontractors 4.22 11.71 0.000
15 Pr2.1 Total number of subcontractors 4.00 10.86 0.000
16 Pr2.2 Selection of subcontractors 4.07 9.38 0.000
Personnel aspect
1 Pe3.3 Adoption of safe work behaviours by workers and supervisors 4.55 18.49 0.000
2 Pe3.11 Managements roles and responsibilities towards safety and health promotion 4.50 17.84 0.000
3 Pe3.2 Attitudes of workers and supervisors towards safe work practices 4.55 17.79 0.000
4 Pe3.12 Managements safety culture 4.40 16.21 0.000
5 Pe3.1 Safety and health training 4.38 15.5 0.000
6 Pe3.8 Inuence of managers and supervisors over workers 4.39 14.83 0.000
7 Pe3.9 Safety committees roles and responsibilities 4.22 12.03 0.000
8 Pe3.7 Work experience of workers and supervisors 3.98 11.5 0.000
9 Pe3.10 Understanding of safety committees aims and objectives by employees 4.13 11.09 0.000
10 Pe3.6 Workers language and communication barriers 3.93 9.28 0.000
11 Pe3.5 Workers adaptation to working environment 3.90 8.53 0.000
12 Pe3.4 Workers cultural backgrounds 3.78 6.85 0.000
Incentive aspect
1 In4.8 Degree, level and type of punishments in terms of nes (Monetary) 4.25 10.38 0.000
2 In4.1 Introduction of incentives 3.95 9.45 0.000
3 In4.7 Introduction of penalties and punishments 4.07 8.99 0.000
4 In4.9 Degree, level and type of punishments in terms of suspension from work 3.93 8.80 0.000
5 In4.2 Level and type of incentives in terms of bonus (Monetary) 3.63 5.13 0.000
6 In4.6 Level and type of incentives in terms of employee of the month award 3.58 4.61 0.000
7 In4.4 Level and type of incentives in terms of certicate of recognition 3.43 3.77 0.000
8 In4.3 Level and type of incentives in terms of promotion (position) 3.32 2.80 0.004
9 In4.5 Level and type of incentives in terms of rewards in kind 3.35 2.79 0.004
10 In4.12 Degree, level and type of punishments in terms of reporting to relevant authorities 3.23 1.60 0.057
11 In4.10 Degree, level and type of punishments in terms of demotion (position) 3.15 1.17 0.123
12 In4.11 Degree, level and type of punishments in terms of termination of service 3.14 0.84 0.202
a safety ocer at the project level [25]. Regular safety
inspections need to be done and it is very important
for a safety ocer to be vigilant and knowledgeable en-
ough to supervise the safety on-goings of daily opera-
tions and construction methods employed. If safety
ocers know their job well, potential accidents could
be avoided or minimised.
The identication of unsafe practices on site is a
very important way of elimination of potential acci-
dents and injuries. About 90% of workplace injuries
can be traced to unsafe work practices and behaviours
[10]. Active employee participation in understanding
safety responsibilities is vital in achieving a successful
safety programme by correcting any unsafe work prac-
tices and reporting these unsafe work actions and con-
ditions to the supervisors [29]. On the other hand,
supervisors and safety ocers must also be competent
enough to identify unsafe practices and to encourage
fellow workers to follow the right procedures. For
example, workers working on scaoldings have to put
on their safety belts and harnesses. Some workers
may feel that they do not need any safety belt and har-
ness with their numerous years of experience and skills.
The insistence of supervisors and safety ocers may
play a huge role in curbing unsafe practices and
achieve accident-free status.
One of the most important safety practices is the
proper handling of construction tools, equipment
and plants. Many workers suered injuries and even
death when they handle the tools and equipment
improperly. All sophisticated tools and equipment that
require special attention must be handled, maintained
and repaired by qualied personnel including larger
equipment and plants like cranes, excavator and bor-
ing machines.
7.2. Results of factor analysis
The factor analysis revealed that the 16 variables
can be condensed into ve factors: (1) quality of sub-
contractors; (2) understanding and implementation of
safety procedures; (3) carrying out work in a safe
manner; (4) carrying out work in a professional man-
ner; and (5) type and method of construction (see
Appendix B).
Subcontractors generally form a large part of the
workforce for dierent trades of work in a construction
project. Thus, subcontractors play a big role in uphold-
ing site safety as shown by other studies [1,3,30]. Fur-
thermore, site safety responsibilities have to be borne
by both the main contractor and the subcontractors in
order to reduce the numbers of site accidents [3]. None-
theless, subcontracting is both an economic imperative
and enemy of safe construction [3].
Safety procedures are created to be followed and
not merely to satisfy the auditors. Having proper
safety procedures not only entails an eective safety
programme, but also serves to identify known hazards
associated with specic work activities and to dene
the process by which to control those hazards within
the context of the work being performed [30]. Further-
more, procedures act as a framework to allow super-
visors and workers to adequately plan various
dangerous activities before works are being carried
out, such as working in conned spaces or at heights
[30].
Carrying out work safely and professionally also
contribute to the process aspect of safety management.
This is because well-trained and qualied persons
know what they are doing and thus, less prone to
accidents. Generally, qualied persons are much more
familiar with their job procedures, take proper precau-
tions when dealing with dangerous activities and mate-
rials, follow instructions, well informed and
experienced in their jobs. A research study found that
the experience level of the operatives depends on their
age [31]. Thus, the older an operative is, the more
experienced he becomes and the higher his level of
safety performance will be.
Finally, type and method of construction will also af-
fect the level of site safety to a reasonable extent. This is
because of the dierent aspects on the usage of tools,
plants, equipment and materials with regards to the pro-
posed type and method of construction. Most workers
being familiar with a certain type and method of con-
struction tend to employ the same techniques and
knowledge even though the nature of the project is dif-
ferent. As such, injuries and accidents will happen.
Retraining and project familiarisation is one way of pre-
venting unnecessary injuries.
8. Discussion on personnel aspect
The t test results showed that all 12 variables relating
to personnel (Pe3.1 to Pe3.12) are very important to en-
sure site safety (p < 0.000) (see Table 2). The three most
(i) Adoption of safe work behaviours by and workers
supervisors (Pe3.3),
(ii) Managements roles and responsibilities towards
safety and health promotion (Pe3.11), and
(iii) Attitudes of workers and supervisors towards safe
work practices (Pe3.2).
The results show that it is very important for both
workers and supervisors to adopt safe work behaviours.
The willingness of workers to adopt safe work practices
also depends largely on their perception of safety, level
of safety education and training received, cultural back-
grounds and communication between fellow workers,
supervisors and superiors. Supervisors play an impor-
tant role in inuencing their workers in the adoption
of safe work behaviours. This nding accords with a
previous one which found that the supervisors safety
behaviours and actions have a direct impact and inu-
ence on workers [29]. To portray positive safety behav-
iours, supervisors are required to: (1) support and view
safety as something worthwhile and workable; (2) lead
by example; (3) integrate safety into all phases of the
job; (4) never ignore near-miss occurrences; (5) being
considerate and take pride in workers jobs; and (6) have
no blame cultural mindsets.
However, the confusion between behaviours and atti-
tudes must be dierentiated. Behaviours are external,
observable acts or activities, while attitudes are inter-
nal ways of thinking or believing through years of
experience and personal views [10].
The results also show that the inuence of manage-
ment on the safety and health standards on site is very
crucial. The responsibilities and roles that management
plays will determine the overall safety performance of
the entire site as was also discovered in other studies
[32,33]. Safety culture has to be inculcated into the orga-
nisation through strong support and encouragement by
senior management. By managements endorsement,
safety promotion will boost the morale and motivate site
personnel to think and work safely.
Many studies have also shown that managements
commitment to safety is one of the major factors to-
wards the enhancement of safety performance
[10,34,35]. When the management is committed to
enhancing safety, zero tolerance will be given to person-
nel who frequently break safety rules and regulations.
Hence, the site safety performance will be dependent
upon the support of the management as well as the
cooperation of the site sta.
The attitudes of site personnel towards safe work
practices are found to be very important. This nding
is consistent with a previous study which who found that
if site personnel develop a concern towards their own
personal safety in their work, they tend to have better
safety records [31]. Thus, safe work practices have to
be maintained at all times and negative work practices
must be eliminated. For example, a worker smokes
while working in a non-permitted smoking area is
endangering the lives of himself and others if a re or
explosion results from his actions. Such irresponsible ac-
tions and attitudes must be changed in order to protect
the well being of all personnel and properties.
8.2. Findings of factor analysis
The factor analysis showed that the 12 variables may
be grouped into three factors: (1) managements attitude
towards safety; (2) supervisors and workers attitude to-
wards safety; and (3) contextual characteristics of work-
ers (see Appendix C).
Managements attitude towards safety has always
been the foremost priority in enhancing the safety per-
formance of an organisation [3]. It is more important
to assess workers safety by using the performance ap-
proach than the compliance approach [33]. The perfor-
mance approach requires a long-term commitment and
understanding towards improving construction pro-
cesses and work activities. As such, change in culture
is inevitable. Thus, the strong support in favour of such
corporate culture change highlights the attitude of the
management and their willingness to move towards a
more desirable and holistic view in workers safety and
health.
Supervisors and workers attitude is another cause for
concern in enhancing safety performance. The impor-
tance of having positive safety attitude has been well
documented [10,32]. Attitudes are personal traits where
an individual has to learn, train, develop and cultivate
over time. In addition, employees safety attitudes could
be one of the criteria for developing a positive safety
climate [36].
The made-up of the worker, in terms of his contex-
tual character includes his cultural background, adapta-
tion to working environment and communication
barriers. Many general workers and operatives in the
Singapore construction industry are foreigners [1]. In
this study, the respondents indicated that the majority
of their workers come from Thailand, India, Bangla-
desh, China and Malaysia. Thus, having diculties in
adapting to dierent work environments, dierent cul-
tures and dierent languages are the norm and have to
be taken into consideration when looking into workers
safety and health as these factors aect the well being
and attitudes of the workers to a large extent [1,27].
9. Discussion on incentive aspect
The t test results showed that seven variables (In4.1
to In4.12) relating to incentives/disincentives are impor-
tant in ensuring site safety (see Table 2). The two most
(i) Degree, level and type of punishments in terms of
nes (Monetary) (In4.8), and
(ii) Introduction of incentives (In4.1).
Implementation of monetary nes is an important
measure. If the nes are harsh and require the violators
to be nancially strained, then these violators may think
twice about committing the same safety oence again.
This appears to be a common practice for many compa-
nies to curb errant workers and even subcontractors in
Singapore. However, such punishments are not desir-
able as the primary objective of an incentive/ disincen-
tive system is to promote safety and not to discourage
or dampen the workers morale.
Another way of enhancing safety is through the intro-
duction of incentives [3,37]. One advantage of incentives
introduction is steered towards the goal-setting trend
within a company. The results do not agree with some
studies which found incentives to be ineective in pro-
moting safety [22,26]. However, incentive is one way
to reward the non-violators as a source of motivation
for their continual eorts, determination and their active
participation in promoting safety on the construction
worksites. For example, workers try to avoid accidents
by complying with safety rules and regulations and tak-
ing necessary safety precautions while working so that
they will be rewarded with incentives for chalking up
an accident free record.
Three variables (In4.10, In4.11, In4.12) are found to
be not important at p < 0.05. These relate to punishment
in terms of demotion, termination of service and report-
ing to the relevant authorities. The reasons may be that
these measures are drastic and should only be consid-
ered when the violator has committed a serious safety
and health oence or is a repeat oender. Discipline
should be the last recourse to reinforce the application
of safe work practices when all else (training, guidance
and encouragement) failed [30].
9.2. Findings of factor analysis
The factor analysis on incentive aspect shows that the
seven important variables may be reduced to three cate-
gories: (1) monetary incentives; (2) non-monetary incen-
tives; and (3) disciplinary action (see Appendix D).
The nature of people is such that they tend to be moti-
vated by the introduction of incentives, benets rewards
and recognition. Thus, people will behave in ways that
lead to desired outcomes (of the company) if compliance
to a source or medium (safety rules and regulations) will
provide them with returns (incentives) [38].
The most common forms of incentives available in
any reward system will probably be in monetary form
and bonuses. However, such monetary incentive would
result in personal gain rather than on promoting safety
[29]. They suggested that the use of such incentive pro-
grammes should be targeted at the whole workforce as
opposed to only a few workers.
Non-monetary incentives include personal recogni-
tion, promotion through ranks, attending reward func-
tions and pre-paid holiday packages. The most
eective incentive programmes centred at encouraging
team spirit and individual recognition [29]. Thus, in or-
der for an incentive system to be eective, the following
nine components have to be considered: upper manage-
ment support; incentives, programme rules, programme
promotion, communication of safety information, moti-
vators, employee participation, communication of pro-
gramme feedback and recognition [10].
Some forms of disciplinary actions should be im-
posed on site, especially to deal with recalcitrant work-
ers. However, following the t test results, these
punishments should not be too harsh, as it may give rise
to further problems.
10. Framework for managing construction safety
Based on the results of the factor analysis, a frame-
work for managing construction safety is proposed for
use by project managers (see Fig. 2). This framework
shows the four main factors (policy, process, personnel
and incentive) and attendant subfactors. This framework
is dierent from other previous safety related studies be-
cause it provides a holistic view of safety management,
and project managers can use it to achieve accidents
reduction. Such framework is especially useful for Singa-
pores construction industry because currently there is no
such tool to help project managers manage safety on site.
The framework is also a new invention because it incor-
porates the requirements of CP79 and views of the local
authority and safety experts. As it is development incor-
porated Singapores Code of Practice (CP79) on Con-
struction Safety Management System (SMS), the new
framework will act as a tool for failure detection in the
SMS implementation. The framework shows that under-
standing and comprehending the 3Ps alone will not guar-
antee accidents reduction. Project managers need to
introduce the Incentive factor (I) as a prime motivational
force, so that fullling the 3Ps may be easier. It is hoped
that this framework may assist in accident reduction.
The product of this study is the 3P + I framework
shown in Fig. 2. Project managers can use this frame-
work to improve construction safety by ensuring that
their construction sites comply with the core factors
and variables identied in the framework. In addition,
project managers can use the checklist in Appendices
ABCD to evaluate whether their sites comply with the
safety practices highlighted there. It acts as a guide
and checklist to signify any negative factors/variables
that a construction site may encounter.
An example of how the framework is used is now gi-
ven. If a worker is seen working on site without his per-
sonal protective equipment, the safety ocer can
quickly identify and classify the problem under the pro-
cess (Pr) and personnel (Pe) factors. The safety violation
could be due to several signicant variables: familiarity
with type and method of construction (Pr2.5); commu-
nication and information ow (Pe2.6); understanding
of safety procedures (Pr2.7); adoption of safe work prac-
tices (Pe3.3); and attitude of worker (Pe3.2). Thereafter,
the safety ocer is able to deal with the violation e-
ciently. This saves eort in trying to determine the root
cause of the problem.
The originality of this product is that it provides a
holistic framework that identies for project managers
the core factors and variables, classied under policy,
process, personnel and incentive aspects, which have
an eect on construction safety. It acts as a guide
and checklist to signify any negative factors/variables
that a construction site may encounter. This is its
contribution to practice. The framework also contrib-
utes to knowledge by establishing the important fac-
tors that are most likely the culprit of safety
violations.
This study suers from several limitations. The rst is
the fairly low response rate. However, since 60 sets of
usable questionnaires were received, this allowed robust
statistical analysis to be carried out. The second limita-
tion of this framework is that it does not indicate how
safe a site is at a given time. Therefore, future studies
will focus nding out how important each variable is
and how each of the items may be measured on site. This
will lead to specic measuring scales (a scoring system)
to ascertain how safe construction sites are.
11. Conclusion
Construction worksites are one of the most danger-
ous places in Singapore because of the high incidence
Appendix A. Details of the factor analysis on policy aspect
Ref Variables that inuence site safety Communalities (h
2
) Factor loading Factor (eigenvalue)
Po1.1 Understanding of Factories Act 0.822 0.890 Understanding and
implementation of SMS
(4.940)
Po1.2 Understanding of CP79: 1999 (SMS) 0.876 0.890
Po1.3 Proper implementation of SMS 0.824 0.783
Po1.6 Understanding of In-house rules and regulations 0.547 0.609
Po1.4 Understanding of OHSMS 0.587 0.636 Understanding and
participation in OHSMS
(1.312)
Po1.5 Companys participation in OHSMS 0.461 0.558
Po1.10 Understanding of insurance policies 0.752 0.861
Po1.7 Proper implementation of In-house rules and regulations 0.793 0.760 Understanding and
implementation of
permit-to-work system
(1.138)
Po1.8 Understanding of Permit-to-work system 0.844 0.647
Po1.9 Proper implementation of Permit-to-work system 0.887 0.910
Appendix B. Details of the factor analysis on process aspect
2
Pr2.2 Selection of subcontractors 0.793 0.864 Quality of subcontractors
(6.502) Pr2.3 Co-ordination, control and management of
subcontractors
0.812 0.707
Pr2.15 Technical competency of specialist subcontractors 0.693 0.678
Pr2.11 Identication of hazardous and dangerous activities 0.668 0.750 Understanding and
implementation of safety
procedures (1.994)
Pr2.7 Understanding of safety procedures 0.890 0.906
Pr2.8 Proper implementation of safety procedures 0.835 0.839
Pr2.9 Identication of unsafe practices on site 0.681 0.700 Carrying out work in a safe
manner (1.649) Pr2.10 Proper implementation of safe practices on site 0.804 0.612
Pr2.12 Good house-keeping 0.717 0.534
Pr2.13 Proper handling of tools, equipment and plants 0.793 0.814
Pr2.16 Tight control of hazardous activities on site 0.851 0.639
Pr2.1 Total number of subcontractors 0.717 0.663 Carrying out work in a
professional manner (1.273) Pr2.5 Familiarity with type and method of construction by
safety ocers/supervisors
0.773 0.859
Pr2.6 Communication and Information ow 0.822 0.524
Pr2.14 Maintenance regime of tools, equipment and plants 0.780 0.503
Pr2.4 Type and method of construction 0.809 0.888 Type and method of
construction (1.021)
of accidents. This study investigated ways that project
managers may take, to make their sites safer. This study
found that many variables signicantly aect site safety
(see Table 2). With the help of Factor analysis, the vari-
ables have been reduced to 14 main components (see
Appendices ABCD).
The main nding of this study is that site safety is af-
fected by four main factors: company safety policy; con-
struction process; personnel management with regard to
safety; and incentives. In terms of safety policies, it is
recommended that project managers implement policies
relating to SMS and OHSMS. Project managers should
also manage the construction process well, especially in
the day-to-day activities on site. Next, three parties must
be committed to safety: management, supervisors and
workmen. Underpinning these three factors is the incen-
tive factor. The incentive factor acts as a source of moti-
vation to the enhancement of safety for the workers. It
may not necessarily be the best tool to enhance safety
performance level of a worksite, but this study has
shown that some form of incentive is important.
The ndings of this study are important because the
factors aecting safety have been identied. If these fac-
tors are addressed and monitored closely, accident
reduction may be realised in construction sites. The
implication of the ndings is that project managers
can use the framework which has been tested in this
study, to manage construction safety with the view of
reducing accidents. This framework provides the nuts
and bolts which project managers can rely on to man-
age safety on their construction sites. It can help in the
detection of failure of implement SMS.
To conclude, safety aects all levels of the construc-
tion organisation: government; company management;
Appendix C. Details of the factor analysis on personnel aspect
2
Pe3.9 Safety committees roles and responsibilities 0.780 0.812 Managements attitude towards
safety (6.232) Pe3.10 Understanding of safety committees aims and
objectives by employees
0.805 0.782
Pe3.11 Managements roles and responsibilities towards
safety and health promotion
0.777 0.723
Pe3.12 Managements safety culture 0.716 0.745
Pe3.1 Safety and health training 0.632 0.529 Supervisors and workers
attitudes towards safety (1.387) Pe3.2 Attitudes of workers and supervisors towards
safe work practices
0.732 0.770
Pe3.3 Adoption of safe work behaviours by
workers and supervisors
0.862 0.832
Pe3.7 Work experience of workers and supervisors 0.647 0.746
Pe3.8 Inuence of managers and supervisors over workers 0.532 0.532
Pe3.4 Workers cultural backgrounds 0.691 0.813 Contextual characteristics of
workers (1.083) Pe3.5 Workers adaptation to working environment 0.746 0.731
Pe3.6 Workers language and communication barriers 0.783 0.690
Appendix D. Details of the factor analysis on incentive aspect
2
In4.1 Introduction of incentives 0.670 0.594 Monetary incentives (2.657)
In4.2 Level and type of incentives in
terms of bonus (Monetary)
0.779 0.878
In4.4 Level and type of incentives in terms of
certicate of recognition
0.796 0.892 Non-monetary incentives (1.592)
In4.6 Level and type of incentives in terms of employee
of the month award
0.750 0.851
In4.7 Introduction of penalties and punishments 0.849 0.867 Disciplinary action (1.105)
In4.8 Degree, level and type of punishments in
terms of nes (Monetary)
0.806 0.897
In4.9 Degree, level and type of punishments in terms
of suspension from work
0.704 0.626
supervisor; and worker. This paper found that the most
important level is the management level of the hierarchy
because its mandate and commitment ensures that safety
and health standards are eectively met. Without the
managements support, enhancing safety and health on
construction worksites may not be possible.
From this study, future research can be carried out.
The rst is to test the framework in dierent regions
of the world to determine its validity in a diverse range
of cultural settings. The second research is to develop a
detailed multi-attribute value model, comprising lower
levels of attributes, to assist in computing the sites Con-
struction Safety Index.
References
[1] Debrah YA, Ofori G. Subcontracting, foreign workers and job
safety in the Singapore construction industry. Asia Pac Bus Rev
2001;1(8):14566.
[2] Teo AL, Chong FW. Understanding construction fatalities in
Singapore: prevention versus cure a dynamic approach. Joint
international symposium of CIB working commissions of W55:
building economics; W65: organisation and management of
construction; W107: construction in developing countries, 2224
October 2003, Singapore.
[3] Rowlinson S. Hong Kong construction site safety manage-
ment. Hong Kong: Sweet and Maxwell Asia; 1997.
[4] Lingard H, Rowlinson S. Construction site safety in Hong Kong.
Constr Manage Econ 1994;12:50110.
[5] Dedobbeler N, Beland F. Risk acceptance and safety performance
on construction sites. J Occup Med 1991;29:8638.
[6] Wilson Jr JM, Kohen E. Safety management: problems encoun-
tered and recommended solutions. J Constr Eng Manage
2000;126(1):779.
[7] Alistair M, Gibb AGF, Haslam R. Causes of accidents and
priorities for intervention. In: Dias ALM, Coble R, editors.
Proceedings of CIB W99, Balkema AA, Rotterdam, Netherlands,
1997, p. 4-7.
[8] Holt ASJ. Principles of construction safety. London: Blackwell
Science; 2001.
[9] Baxendale T, Jones O. Construction design and management
safety regulations in practice-progress on implementation. Int J
Proj Manage 2000;18(1):3340.
[10] Heberle D. Construction safety manual. New York: McGraw-
Hill; 1998.
[11] Clarke S. Perceptions of organizational safety: implications for
the development of safety culture. J Org Behav 1999;20(2):18598.
[12] Toole M. The relationship between employees perceptions of
safety and organisational culture. J Safety Res 2002;33:23143.
[13] Lingard H. The eect of rst aid training on objective safety
behaviour in Australian small business construction rms. Constr
Manage Econ 2001;19:21727.
[14] McKenna SP, Hale AR. The eect of emergency rst aid training
on the incidence of accidents in factories. J Occup Accid
1981;4:4759.
[15] Hinze J. Human aspects of construction safety. J Constr Div,
Proc Am Soc Civil Eng 1981;107:6172.
[16] Lindell MK. State of the art reviews. Occup Med 1994;9(2):21140.
[17] Health and Safety Executive (HSE). Blackspot construction: a
study of ve years of fatal accidents in the building and civil
engineering industries. London: HMSO; 1988.
[18] Hinze J, Raboud P. Safety on large building construction projects.
J Constr Eng Manage 1988;144(2):28693.
[19] Mattila M, Rantanen E, Hyttinen M. The quality of work
environment, supervision and safety in building construction.
Safety Sci 1994;17(4):25768.
[20] Sulzer-Azaro B, Harris TC, McCann KB. Beyond training:
organisational performance management techniques. Occup Med
1994;9(2):32139.
[21] Geller ES. Behaviour-based safety: confusion, controversy and
clarication. Occup Health Safety 1999;68(1):409.
[22] McAfee RB, Winn AR. The use of incentives and feed-back to
enhance work place safety: a critique of the literature. J Safety Res
1989;20:719.
[23] Hofmann DA, Stetzer A. A cross-level investigation of factors
inuencing unsafe behaviours and accidents. Pers Psychol
1996;49:30739.
[24] Hinze J. Safety incentives: do they reduce injuries?. Pract
Periodical Struct Des Constr 2002;7(2):814.
[25] Hinze J, Wilson G. Moving toward a zero injury objectives. J
Constr Eng Manage 2000;126(5):399403.
[26] Hinze J, Gambatese J. Factors that inuence safety performance
of specialty contractors. J Constr Eng Manage
2003;129(2):15964.
[27] Kartam NA, Flood I, Koushki P. Construction safety in Kuwait:
issues, procedures, problems and recommendations. Safety Sci
2000;36:16384.
[28] Cox SJ, Cheyne AJT. Assessing safety culture in oshore
environments. Safety Sci 2000;34:11129.
[29] Peyton RX, Rubio TC. Construction safety practices and
principles. New York: Van Nostrand Reinhold; 1991.
[30] Hislop RD. Construction site safety: a guide to managing
contractors. New York: Lewis Publishers; 1999.
[31] Sawacha E, Naoum S, Fong D. Factors aecting safety perfor-
mance on construction sites. Int J Proj Manage
1999;17(5):30915.
[32] Mattila M, Hyttinen M, Rantanen E. Eective supervisory
behaviour and safety at the building site. Int J Ind Ergon
1994;13:8593.
[33] Haupt TC. A study of management attitudes to a performance
approach to construction worker safety. J Constr Res
2003;4(1):87100.
[34] Langford D, Rowlinson S, Sawacha E. Safe behaviour and safety
management: its inuence on the attitudes of workers in the UK
construction industry. Eng Constr Archit Manage
2000;7(2):13340.
[35] Mohamed S. Safety climate in construction site environments. J
Constr Eng Manage 2002;128(5):37584.
[36] Cox SJ, Cox TR. The structure of employee attitude to safety: a
European example. Work Stress 1991;5:93106.
[37] Zakeri M, Olomolaiye P, Holt GD, Harris FC. Factors aecting
the motivation of Iranian construction operatives. Build Environ
1997;32(2):1616.
[38] Vredenburgh AG. Organizational safety: which management
practices are most eective in reducing employee injury rates. J
Safety Res 2002;33:25976.
[39] Ling FYY, Ofori G, Teo EAL. Predicting safety levels of
construction project sites. In Proceedings of the CIB World
Building Congress: building for the future, Toronto, 17 May
2004.

Factors Affecting Cost Performance of Construction in India

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Factors Affecting Cost Performance of Construction in India

Uploaded by

Copyright:

Available Formats

1.

You might also like