Six-Sigma Quality, Process Capability,

and Statistical Process Control

Selected Slides from Jacobs et al, 9th Edition
Operations and Supply Management
Chapter 9 and 9A
Edited, Annotated and Supplemented by
Peter Jurkat

Total Quality Management (TQM)

Total quality
management is defined
as managing the entire
organization so that it
excels on all dimensions
of products and services
that are important to the
customer

Design quality: Inherent

value of the product in
the marketplace
Dimensions include:
Performance, Features,
Reliability/Durability,
Serviceability, Aesthetics,
and Perceived Quality.

Conformance quality:
Degree to which the
product or service design
specifications are met

9-3

Six Sigma Quality

A philosophy and set of methods

companies use to eliminate defects
in their products and processes
Seeks to reduce variation in the
processes that lead to product
defects

Gurus and their wisdom

McGraw-Hill/Irwin

Copyright 2009 by The McGraw-Hill Companies, Inc. All rights reserved.

Consensus
Gurus had considerable differences
After 30 years get some consensus
Senior level leadership
Customer focus
Work force involvement
Process analysis
Continuous improvement

9-6

Costs of Quality

Appraisal Costs

External Failure
Costs

Costs of
Quality

Internal Failure
Costs

Prevention Costs

9-7

Cost of Quality Example

At 20% of sales, represents about $2M sales, at 2.5% about $73M sales

9-8

Six Sigma Quality Measurement

Six Sigma allows managers to readily describe process

performance using a common metric: Defects Per Million
Opportunities (DPMO)

DPMO

Numberof defects

x1,000,000

Numberof
opportunities
for error per x No. of units
unit

Defect associated with a critical-to-quality characteristic: a

measurable quantity used to identify failure
Statistical six sigma goal is 3.4 failures per 1,000,000 opportunities 1
failure in about 300,000 DPMO (actually 1 in 294,118)

Process Capability
Shows to what extent (probability) parts are produced that meet and fall
outside of specifications
Achieved when process variation (s.d.) is so small that an acceptable
proportion are defects Six-Sigma goal is 3.4 out of one million

Bearing Diameter

9A-9

Why 3.4 DPMO?

Six sigma between mean and, say, upper
specification limit results 1 defect in
100,000,000 (see SixSigmaOrigin.xls)
Experience has shown that in long term
processes have a wider variation than in short
term studies, which results in defects with
probability beyond 4.5 sigma, i.e. 3.4 DPMO
(1.5 sigma less than 6 sigma)
See origin of this at http://en.wikipedia.org/wiki/Six_Sigma ( which bases
above on Tennant, Geoff (2001). SIX SIGMA: SPC and TQM in
Manufacturing and Services. Gower Publishing, Ltd., p. 25. ISBN
0566083744.)

9A-11

Mean shift during process improvement

Still an improvement but capability is now measured against closest of LTL and UTL
LTL = lower tolerance limit

UTL = upper tolerance limit

9-12

Measure Example

Example of Defects Per Million

Opportunities (DPMO) calculation.
Suppose we observe 200 letters delivered
incorrectly to the wrong addresses in a
small city during a single day when a
total of 200,000 letters were delivered.
What is the DPMO in this situation?

DPMO

200

So, for every one

million letters
delivered this
citys postal
managers can
expect to have
1,000 letters
incorrectly sent to
the wrong
address.

x1,000,000 1, 000

x200,000

Cost of Quality: What might that DPMO mean in terms of

over-time employment to correct the errors?

Service Blueprint, Failure Anticipation, and Poka-Yokes

Complete blueprint (p262-3)

identifies 16 failure opportunities

8-13

Toyota Dealer Service Example

Blueprint identified 16 failure opportunities
per customer
Assume 20 customers /day => 80,000
customers/year for 250 working days per year
At 3.4 failures per 1,000,000 opportunities this
would allow .272 failures/year, or 3 2/3 years
between failures
What is the critical-to-quality characteristic of
the first identified failure? Second failure?

DMAIC Cycle
GE developed
methodology
Overall focus is to
understand and
achieve what the
customer wants (Juran)
Identifies defects and
variation in processes
as underlying cause of
defects (Deming)
A 6-sigma program
seeks to reduce the
variation in the
processes that lead to
these defects

Define customers and

their priorities
Measure process and its
performance
Analyze causes of
defects
Improve by removing
causes
Control to maintain
quality

DMAIC in Action
We are the maker of a
2. Measure
cereal. Consumer Reports
a. How would we measure to
has just published an article
evaluate the extent of the
that shows that we
problem?
frequently have less than 16
b. What are acceptable limits
ounces of cereal in a box.
on this measure?
What should we do?
c. Lets assume that the
government says that we
1. Define
must be within 5 percent
a.

b.

What is the critical-toquality characteristic?

The CTQ (critical-toquality) characteristic in
this case is the weight of
the cereal in the box.

d.
e.
f.

of the weight advertised on

the box.
Upper Tolerance Limit = 16 +
.05(16) = 16.8 ounces
Lower Tolerance Limit = 16
.05(16) = 15.2 ounces
Survey: 1000 boxes have
mean weight = 15.875 oz
with s.d. = .529

9-17

Lower Tolerance
= 15.2

Process
Mean = 15.875
Std. Dev. = .529

Upper Tolerance
= 16.8

What percentage of boxes are defective (i.e. less than 15.2 oz)?
Z = (x Mean)/Std. Dev. = (15.2 15.875)/.529 = -1.276
NORMSDIST(Z) = NORMSDIST(-1.276) = .100978

Approximately, 10 percent of the boxes have less than 15.2

Ounces of cereal in them way out of six-sigma specs

DMAIC in Action
3. Analyze - how can we
improve the
capability of our cereal
box filling process?
a. Decrease Variation
b. Center Process
c. Tighten Specifications

4. Improve How good is

good enough?
a. Set center spec at goal
(16 oz in this case)
b. Set controls so that a
deviation of 6 s.d.
occurs only 3.4 times
out of a million

DMAIC in Action
5. Statistical Process
Control
a.
b.
c.

d.

Use data from actual

process
Estimate distributions
Calculate capability do
better if not adequate
(actually do better all the
time)
Statistically monitor the
process over time

e. Tools
1)
2)
3)
4)
5)
6)
7)

8)
9)

Process flow charts (e.g.,

Toyota service blueprint)
Run charts
Pareto charts
Check sheets
Cause-and-effect
diagrams
Opportunity flow
diagrams
Failure mode and effect
analysis (FMEA)
Statistical Process Control
(SPC) and Control charts
Design of Experiments
(DOE)

9-20

9-21

9-22

9-23

Failure Mode and Effect Analysis

Severity: cost of damage, rating number

Occurrence: observed relative frequency,
predicted probability
Detection: probability of detection
RPN = Occurrence X Severity X Detection

9-24

Upper
Control
Limit

Lower
Control
Limit

Part of Statistical Process Control

Uses statistical theory and practice to follow processes in order to determine if they are
within specification/control
Also used to predict if a process might be going out of control while still within specs
General approach is to sample a process at intervals, plot the results and compare
these to control limits

9A-25

Statistical Process Control

Based on statistical theory of variation (dispersion)
Defines process capability
Establishes process control limits
Controls process bases on periodic sampling (small samples
as opposed to inspecting/measuring everything)

Variation
Assignable variation is caused
by factors that can be clearly
identified and possibly
managed

Common variation is inherent

in the production process

Example: A poorly trained

employee that creates
variation in finished product
output.
Example: A molding process
that always leaves burrs or
flaws on a molded item.

9A-26

Taguchis View of Variation

Traditional view is that quality within the LS and US is good and that
the cost of quality outside this range is constant, Taguchi views costs as
increasing as variability increases, so seek to achieve zero defects and
that will truly minimize quality costs.
High

High

Incremental
Cost of
Variability

Incremental
Cost of
Variability

Zero

Zero
Lower Target Upper
Spec
Spec
Spec
Traditional View

Lower Target Upper

Spec
Spec
Spec
Taguchis View

Upper and lower specs are also called upper and lower tolerance limits (UTL and LTL)

9A-27

Process Capability Index, Cpk

Capability Index shows

how well parts being
produced fit into design
limit specifications.

LTL/UTL = lower/upper tolerance limit

X LTL
UTL - X

C pk = min
or

3
3

As a production process
produces items small
shifts in equipment or
systems can cause
differences in
production
performance from
differing samples.

Shifts in Process Mean

9A-28

The Cereal Box Example

We are the maker of this cereal. Consumer reports has

just published an article that shows that we frequently
have less than 16 ounces of cereal in a box.
Lets assume that the government says that we must
be within 5 percent of the weight advertised on the
box.
Upper Tolerance Limit = 16 + .05(16) = 16.8 ounces
Lower Tolerance Limit = 16 .05(16) = 15.2 ounces
We go out and buy 1,000 boxes of cereal and find that
they weight an average of 15.875 ounces with a
standard deviation of .529 ounces.

9A-29

Cereal Box Process Capability

Specification or Tolerance
Limits
Upper Spec = 16.8 oz
Lower Spec = 15.2 oz

Observed Weight
Mean = 15.875 oz
Std Dev = .529 oz

X LTL UTL X
C pk Min
;

3
3
15.875 15.2 16.8 15.875
C pk Min
;

3
(.
529
)
3
(.
529
)

C pk Min.4253 ; .5829
C pk .4253

9A-30

What does a Cpk of .4253 mean?

An index that shows how well the

units being produced fit within the
specification limits.
This is a process that will produce a
relatively high number of defects.
Many companies look for a Cpk of 1.3
or better 6-Sigma company wants
2.0!

9A-31

Types of Statistical Sampling in SPC

Attribute (overall acceptable or not)

Defectives refers to the acceptability of product across a

range of characteristics.
Defects refers to the number of defects per unit which
may be higher than the number of defectives.
p-chart application (p for proportion)

Variable (Continuous)

Usually actual dimensions (length, weight, hardness, )

Usually measured by the mean and the standard
deviation.
X-bar and R chart applications (x-bar for mean and R for
range much easier to measure than s.d.)

9A-32

Control Charts

9A-33

Statistical Process Control Formulas:

Attribute Measurements (p-Chart)

Given:

T o ta l N u m b e r o f D e fe c tiv e s
p =
T o ta l N u m b e r o f O b s e rv a tio n s
sp =

p (1- p)
n

Compute control limits:

UCL = p + z sp
LCL = p - z sp

9A-34

Example of Constructing a p-chart

1. Calculate the sample

proportions, p (these
are what can be plotted
on the p-chart) for each
sample
2. Calculate the
average and s.d. of the
sample proportions

55
p =
= 0.036
1500

sp =

p (1 - p)
=
n

Sample
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

n
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100

Defectives
4
2
5
3
6
4
3
7
1
2
3
2
2
8
3

.036(1- .036)
= .0188
100

p
0.04
0.02
0.05
0.03
0.06
0.04
0.03
0.07
0.01
0.02
0.03
0.02
0.02
0.08
0.03

9A-35

Example of Constructing a p-Chart

Calculate control limits and plot the individual sample proportions, the
average of the proportions, and the control limits .036 3(.0188)

UCL = 0.0924
LCL = -0.0204 (or 0)

0.16
0.14
0.12

UCL

0.1
p 0.08
0.06
0.04
0.02
0

LCL
1

O b servation

10

11

12

13

14

15

9A-36

Example of x-bar and R charts: Steps: Calculate x-bar Chart and Plot Values

UCL = x + A 2 R 10.728 - .58(0.2204 ) = 10.856

LCL = x - A 2 R 10.728 - .58(0.2204 ) = 10.601
1 0 .9 0 0
1 0 .8 5 0

UCL

M ea n s

1 0 .8 0 0
1 0 .7 5 0
1 0 .7 0 0
1 0 .6 5 0
1 0 .6 0 0

LCL

1 0 .5 5 0
1

S am p le

10

11

12

13

14

15

9A-37

Example of x-bar and R charts: Calculate R-chart and Plot Values

UCL = D 4 R ( 2.11)( 0.2204 ) 0.46504

LCL = D 3 R (0)( 0.2204 ) 0
0 .8 0 0
0 .7 0 0
0 .6 0 0
0 .5 0 0
R

UCL

0 .4 0 0
0 .3 0 0
0 .2 0 0
0 .1 0 0

LCL

0 .0 0 0
1

8
S a m p le

10

11

12

13

14

15

9A-38

Common criteria for concluding

process is out of control or in
danger of being so

Acceptance Sampling
Purposes
Determine quality level of acquired goods or services
(after the fact) when no sampling of production
process is available
Ensure quality is within predetermined level

Advantages

Economy
Less handling damage
Fewer inspectors
Upgrading of the
inspection job
Applicability to destructive
testing
Entire lot rejection
(motivation for
improvement)

Disadvantages

Risks of accepting bad

lots and rejecting good
lots
Added planning and
documentation
Sample provides less
information than 100percent inspection

9A-40

Risk

Acceptable Quality Level (AQL)

Max. acceptable percentage of defectives defined

by producer

The a (Producers risk)

The probability of rejecting a good lot

Probability of Type I error based on consumers
null hypothesis that lot is good

Lot Tolerance Percent Defective (LTPD)

Percentage of defectives that defines consumers

rejection point

The (Consumers risk)

The probability of accepting a bad lot

Probability of Type II error based on consumers
null hypothesis that lot is good

9A-41

Operating Characteristic Curve

Probability of accepting lots

with give % of defectives

The OCC brings the concepts of producers risk, consumers

risk, sample size, and maximum defects allowed together
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

a = .05 (producers risk

= reject good lot)
n = 99
c=4

=.10(consumers risk = accept bad lot)

AQL
H0: Lot is good
Ha: Lot is bad

LTPD
Percent defective

10 11 12

The shape
or slope of
the curve is
dependent
on a
particular
combination
of the four
parameters

n = sample size
c = acceptance number (max
defectives allowed before lot
is rejected)

9A-42

Example: Acceptance Sampling Problem

Zypercom, a manufacturer of video interfaces,

purchases printed wiring boards from an outside
vender, Procard. Procard has set an acceptable
quality level of 1% and accepts a 5% risk of rejecting
lots at or below this level. Zypercom considers lots
with 3% defectives to be unacceptable and will assume
a 10% risk of accepting a defective lot.
Develop a sampling plan for Zypercom and determine
a rule to be followed by the receiving inspection
personnel.

9A-43

Example: What is given and what is not?

LTPD = Lot tolerant percent defective (buyers)
AQL = Acceptable quality level (seller)

In this problem, AQL is given to be 0.01 and LTDP is given to be

0.03. We are also given an alpha of 0.05 and a beta of 0.10.
What you need to determine is your sampling plan is c and n.
For a give allowed sampling error SE and confidence C = 1 - the sample
size is determined by:

pqz2 / 2
n=
SE 2

where p = probability of a defective and q = 1 - p

9A-44

Example: Step 2. Determine c

LTPD = Lot tolerant percent defective (buyers)
AQL = Acceptable quality level (seller)

First divide LTPD by AQL

LTPD
.03
=
= 3
AQL
.01

Then find the value for c by selecting the value in the QA-12 (on disk)
n(AQL)column that is equal to or just greater than the ratio above.
Exhibit QA-12

c
0
1
2
3
4

LTPD/AQL
44.890
10.946
6.509
4.890
4.057

So, c = 6.

n AQL
0.052
0.355
0.818
1.366
1.970

c
5
6
7
8
9

LTPD/AQL
3.549
3.206
2.957
2.768
2.618

n AQL
2.613
3.286
3.981
4.695
5.426