April 11

Forecasting Techniques
Prepared By:
Ashraf S. Youssef; Ph. D.
Quality Assurance & Communications Manager
Electronic & Electrical Industries Corp

May 23 – 24, 2011

Training Approach

• Focus on Basics
• Interactive
• Exercises 
• Case Studies
• Role Plays
• Have Fun !

Learning Objectives
• Learn about Forecasting 
Techniques and how can you link 
it with SAP.
• Become Familiar with Different 
Techniques.
Techniques
• Design Your Own Model.
• Realize the Benefits of 
Forecasting techniques at Your 
Organization

1
 April 11

Agenda

1. Production/service planning & Control

2. Forecasting Overview & Objectives
3. Forecasting Techniques
4 Material
4. M t i l requirement
i t planning
l i
5. Inventory Policy
6. Case Study.
7. Integration with SAP.

Production Planning and Control

The Production Control System

Demand Sales
and order Customer
forecasting
entry

Materials Shop-floor
p Shipping
pp g
Aggregate requirement scheduling Production and
planning planning and control receiving

Inventory
management Inventory Vendors

Production Planning and Control: Main Functions

Forecasting to predict customer demand on various

products over a given horizon.

Aggregate Planning to determine overall resources

needed.

Materials Requirement Planning to determine all required

components and timing.

Inventory Management to decide production or purchase

quantities and timing.

Scheduling to determine shop-floor schedule of various

components.

2
 April 11

Production Planning and Control Purpose

Effectively utilize limited resources in the production

of goods so as to satisfy customer demands and
create a profit for stakeholders.

Resources include the production facilities, labors

and materials
materials.

Constraints include the availability of resources,

delivery times for the products, and management
policies.

Production Objectives
High
Profitability

Low High
Costs Sales

Low Unit Quality High Customer

Costs Product Service

High High Low Fast Many

Throughput Utilization Inventory Response products

Less Short Low High More

Variability Cycle Times Utilization Inventory Variability

Performance Measures

Throughput
WIP
Cycle time
Service quality (OTD)
Profit

3
 April 11

Forecasting

• Objective: predict demand for planning

purposes.
• Forecasting Criterion:
1. Accuracy
2. Simplicity
p y of Computation
p
3. flexibility
• Forecasting Tools:
– Qualitative: Delphi, Analogies
– Quantitative: Causal and time series models

Aggregate Planning
• Objective: generate a long-term production plan that
establishes a rough product mix, anticipates bottlenecks,
and is consistent with capacity and workforce plans.
• Issues:
– Aggregation: product families and time periods must be
set appropriately for the environment.
– Coordination: AP is the link between the high level
functions of forecasting/capacity planning and
intermediate level functions of MRP, inventory control,
and scheduling.
– Anticipating Execution: AP is virtually always done
deterministically, while production is carried out in a
stochastic environment.

Demand Management
• Objective: establish an interface between the customer
and the plant floor, that supports both competitive
customer service and workable production schedules.
• Issues:
– Customer Lead Times: shorter is more competitive.
– Customer Service: on-time delivery.
– Batching: grouping like product families can reduce
lost capacity due to setups.
– Interface with Scheduling: customer due dates are
an enormously important control in the overall
scheduling process.

4
 April 11

Material Requirement Planning

• Objective: Determine all purchase and production

components needed to satisfy the
aggregate/disaggregate plan.
• Issues:
– Bill of Materials: Determines components,
p ,
quantities and lead times.
– Inventory Management: Must be coordinated with
inventory.

Definitions
• Parts: a component, sub-assembly, or an assembly that
moves through the workstations.
• End Items: parts sold directly to customers; relationship to
constituent parts defined in bill of material.
• Work in Process (WIP): inventory between the start and
endpoints of a product routing.
• Raw Material Inventory (RMI): material stocked at beginning of
routing.
• Crib and Finished Goods Inventory (FGI): crib inventory is
material held in a stock point at the end of a routing; FGI is
material held in inventory prior to shipping to the customer.
• Order: request from customer.
• Job: transfer quantity on the line.

Forecasting

xt

5
 April 11

Production Planning and Control

The Production Control System

Demand Sales
and order Customer
forecasting
entry

Materials Shop-floor Shipping

Aggregate Production
requirement scheduling and
planning
planning and control receiving

Inventory
Inventory Vendors
management

Forecasting

• Basic Problem: predict demand for planning

purposes.
• Forecasting Tools:
– Qualitative:
• Delphi
• Analogies
– Quantitative:
• Causal models (e.g., regression models)
• Time series models

Forecasting “Laws”
1) Forecasts are always wrong!
2) Forecasts always change!
3) The further into the future, the less reliable the
forecast!
40%
Trumpet
p of Doom
20%

+10%
-10%

Start of
season
16 weeks
26 weeks

6
 April 11

Why Forecasting?

• Most decisions made in PP&C are “plans for future”

– How much of a certain product should I make
next week?
– How much inventory should I keep in May 2011?
• If yyou forecast the future ((more accurately),
y) yyou will
(more likely) have better plans.
• Many decisions (plans) in PP&C depend on the
forecasts of demand.
• Forecasting demand versus forecasting sales.

Qualitative Forecasting
1. Delphi Technique: Iteratively tapping multiple
experts opinion with continuous comparisons and
revisions.

2. Market research: Panels; Questionnaires, Test

markets.

3. Product Life cycle analogy: Forecast based on

life cycle of similar products.

Quantitative Forecasting
• Goals:
–Predict future from past
–Smooth out “noise”
–Standardize forecasting procedure

• Methodologies:
–Causal
Causal Forecasting:
• regression analysis
–Time Series Forecasting:
• Moving average
• Weighted Moving Average
• Exponential smoothing
• Double Exponential smoothing
• Seasonal models

7
 April 11

The Forecasting Model

Historical Data

Forecast
Generation

Forecast Current
Forecast
Control Observation

Managerial
Judgment and
Experience

Modified Forecast

Time Series Forecasting

Historical Data Forecast

xt, t = 1, … ,T Time series model xˆ

T +τ
(T )

Observed value at Predicted value for

x time t
t
time T+τ made at time
T
T.

t
T T+τ

The Constant Model

Model: • xt
xt = a + εt, εt NID(0, σ2)

Assumptions: •
Noo ttrend,
e d, o
only
y random
a do noise
o se
Noise is normally distributed t
Noise is independent
Noise has a constant variance (doesn’t change with
time)
Noise averages out to zero

8
 April 11

The Constant Model

Moving Average (same as least squares)

V (M T ) = σ ,
2
1 T 1
M T
= ∑ ,
N t =T − N +x1 t N xˆ T +τ
(T ) = M T ± 2 1 +
N
σ

•Averages the latest N observations.

•N
N ranges from 2 to 12.
12
•Larger N provides stability, but less responsive.

¾Assigns equal weights to the latest N observations

¾Does not adapt to error

The Constant Model

Moving Average- Example
t xt Mt xˆt (t − 1) et |et|

1 20
2 22
21.67
3 23
21 67
21.67 21 67
21.67 -1.67
1 67 1 67
1.67
4 20
20.33 21.67 -3.67 3.67
5 18
20.00 20.33 1.67 1.67
6 22
20.67 20.00 2.00 2.00
7 22
22.33 20.67 2.33 2.33
8 23
21.67 22.33 -2.33 2.33
9 20
22.33 21.67 2.33 2.33
10 24
2.29

The Constant Model

Exponential Smoothing
α α
2 −α σ
=α + (1 − α ) S T −1 , V (S T ) = (T ) = S T ± 2 1 + σ
2
S T x T
, xˆ T +τ
2 −α

ƒCombines current observation and previous forecasts

for future predictions.
ƒα ranges from 0.1 to 0.5. (α = 2/(Ν+1))
ƒSmaller α provides stability, but less responsive.

¾Assigns decaying weights to all observations

¾Adapts to error

9
 April 11

The Constant Model

Exponential Smoothing- Example
t xt St xˆ t (t − 1) et |et|

20 21.67
1
22 21.83 21.67
2
23 22.42 21.83 1.17 1.17
3
20 21 21
21.21 22 42
22.42 -2.42
2 42 2 42
2.42
4
18 19.60 21.21 -3.21 3.21
5
22 20.80 19.60 2.40 2.40
6
22 21.40 20.80 1.20 1.20
7
23 22.20 21.40 1.60 1.60
8
20 21.10 22.20 -2.20 2.20
9
24 22.55 21.10 2.90 2.90
10
2.14

The Linear Model

xt
• Model:
xt = a + bt + εt, εt NID(0, σ2)
• Assumptions:
Linear trend
Noise is normally distributed
Noise is independent t

Noise has a constant variance (doesn’t change with

time)
Noise averages out to zero

The Linear Model

Least Squares

T T T
N ∑t x x
− ∑ x ∑t x
bˆ = t =T − N +1
T
t =T − N +1 t =T − N +1
T
, aˆ = x − bˆt
∑t ∑t )
2
N 2
−(
t =T − N +1 t =T − N +1

1 (T + τ − t )2
xˆ T +τ
(T ) = aˆ + bˆ(T + τ ) ± 2 1 + + σ
N S tt

10
 April 11

The Linear Model

Double Moving Average

1 T
( N − 1)
= ∑M , E[ M T ] = E[ xT ] − ( N − 1)b = E[ M T ] −
[ 2] [ 2]
M T
N t =T − N +1
t
2
b

2τ 2τ
(T ) = 2 M T − M T , (T ) = ( 2 + − (1 +
[ 2] [ 2]
xˆ T xˆ T +τ
N −1 M T
)
N −1 M T
) , ( point estimate )

2τ 2τ 1 (T + τ − t ) 2
(T ) = ( 2 + − (1 + ± 2 1+ + σ
[ 2]
xˆ T +τ
N −1 M T
)
N −1 M T
)
N S tt

The Linear Model

Double Exponential Smoothing
2β β
=α + (1 − α ) S T −1 , E[S T ] = E[ xT ] − b = E[S T ] −
[2] [2] [2]
S T S T
α α
b

ατ ατ [ 2 ]
(T ) = 2 S T − S T , (T ) = (2 + − (1 + ) S T , (point estimate )
[2]
xˆ
T xˆ T +τ
β ST
)
β

ατ ατ 1 (T + τ − t )2
(T ) = ( 2 + − (1 + ± 2 1+ + σ
[ 2]
xˆ T +τ
β ST
)
β ST
)
N S tt

Example: Linear Model

Year
Month 1 2 3 4
Jan 816 870 842 898
Feb 817 890 896 869
Mar 686 795 737 787
Apr 594 664 694 749
May 596 625 732 777
June 526 634 618 756
Jul 540 610 653 642
Aug 602 617 658 673
Sept 665 708 682 840
Oct 747 738 789 914
Nov 756 674 703 935
Dec 868 870 875 1020

11
 April 11

The Seasonal Model

Model:•
xt = (a + bt)cL + εt, εt NID(0, σ2)
Example
The quarterly sales of company XYZ over the past five years are
given below. Use this data to forecast the sales in every quarter of
years 6 and 7. Use the same data to forecast the total sales of
years 6 and 7.
Quarter
Year I II III IV
1 17.51 40.24 35.76 20.08
2 21.33 47.81 45.45 24.11
3 25.26 50.04 53.92 29.65
4 27.38 65.50 56.58 26.51
5 32.99 67.27 65.94 34.34

The Seasonal Model

Example
70.00

60.00

50.00

40 00
40.00

30.00

20.00

10.00

0.00
1

The Seasonal Model

Example
Year I II III IV
1 17.51 40.24 35.76 20.08 28.40
2 21.33 47.81 45.45 24.11 34.67
a = 23.55
b = 5.28.
3 25.26 50.04 53.92 29.65 39.72
4 27.38 65.50 56.58 26.51 43.99
5 32 99
32.99 67 27
67.27 65 94
65.94 34 34
34.34 50 14
50.14

Year I II III IV
1 0.6 2 1.4 2 1 .26 0 .71
2 0.6 2 1.3 8 1 .31 0 .70
Seasonal 3 0.6 4 1.2 6 1 .36 0 .75
Indices 4 0.6 2 1.4 9 1 .29 0 .60
5 0.6 6 1.3 4 1 .32 0 .68
A verag e 0.6 3 1.3 8 1 .31 0 .69

12
 April 11

The Seasonal Model

Forecasted Values
Year 6 average = 23.55 + 6 * 5.28 = 55.23
Year 7 average = 23.55 + 7 * 5.28 = 60.51

Seasonal Q u a rte r I II III IV

Indices In d e x 0 .6 3 1 .3 8 1 .3 1 0 .6 9

Year I II III IV
Forecasted 6 3 4 .7 7 7 6 .0 5 7 2 .1 1 3 7 .9 5
Values
7 3 8 .0 9 8 3 .3 3 7 9 .0 0 4 1 .5 8

Criteria for Selecting a Forecasting Model

1. Give the smallest bias, as measured for cumulative

forecast error; or
2. Gives the smallest mean absolute deviation; or
3 Support management’s beliefs about the underlying
3.
pattern of demand.
4. If the demand inherently stable; then higher vales of
N are suggested.

Materials Requirement Planning

Booster

Pump Motor

13
 April 11

Production Planning and Control

The Production Control System

Demand Sales and

order entry Customer
forecasting

Materials Shop-floor Shipping

Aggregate
requirement scheduling Production and
planning
planning and control receiving

Inventory
Inventory Vendors
management

Material Requirements Planning

Given a master schedule, what raw
materials and components quantities
Master Production
are needed and when. Scheduling
Bills of
Material
Material Requirements
Planning
Inventory
Status

Quantities and
Requires: Order Times

•Bill of materials (Manufacturing tree)

•Inventory status
•Lead times

Assumptions
1. Known deterministic demands.

2. Fixed, known production lead times.

3. Infinite capacity.

Idea is to “back out” demand for components by using

lead times and bills of material.

14
 April 11

MRP Procedure

1. Netting: net requirements against projected inventory

2. Lot Sizing: planned order quantities

g p
3. Time Phasing: planned orders backed out by
y lead time

4. BOM Explosion: gross requirements for components

Inputs

– Master Production Schedule (MPS): due dates and

quantities for all top level items

– Bills of Material (BOM): for all parent items

– IInventory
t Status:
St t (on
( hand
h d plus
l scheduled
h d l d receipts)
i t ) ffor
all items

– Planned Lead times: for all items

Material Requirements Planning: Example

Consider the manufacturing tree Booster

(BOM) and the accompanying
information.

Pump Motor
Period GR
1 10
2 15
3 12
4 16 Scheduled Order Receipts (SOR)
5 15
6 12
Period 1 2 3 4 5 6 7 8
7 18 Pump 12 14 0 0 0 0 0 0
8 14 Motor 8 2 2 0 0 0 0 0

15
 April 11

Material Requirements Planning: Example

Time 1 2 3 4 5 6 7 8
GR 10 15 12 16 15 12 18 14
Pump SOR 12 14 0 0 0 0 0 0
BI 0 2 1 0 0 0 0 0
LT = 2 NR -2 -1 11 16 15 12 18 14
POR 0 0 11 16 15 12 18 14
EI 2 1 0 0 0 0 0 0
PORE 11 16 15 12 18 14 ? ?

T ime 1 2 3 4 5 6 7 8
GR 22 32 30 24 36 28 ? ?
SOR 8 2 2 0 0 0 0 0
Motor BI 14 0 0 0 0 0 0 0
NR 0 30 28 24 36 26 ? ?
LT = 1 POR 0 30 28 24 36 26 ? ?
EI 0 0 0 0 0 0 0 0
PORE 30 28 24 36 26 ? ? ?

Capacity Requirements Planning (CRP)

– Uses routing data (work centers and times) for all items

– Explodes orders against routing information

– Generates usage profile of all work centers

– Identifies overload conditions

– More detailed than RCCP

– No provision for fixing problems

– Lead times remain fixed despite queuing

MRP Results

• 1980 Survey (Over 1,100 US and Euro Firms):

– Less than 10% of companies recoup MRP investment within
two years

• 1982 Survey of 679 APICS Members:

– 9.5% regarded their companies as Class A users
– 60% reported their firms as Class C or D users

• 1986 Survey of 33 S. Carolina MRP Users:

– Similar responses to 1982 survey
– Avg. investment in hardware, software, and training for an
MRP system was $795,000 (std dev. of $1,191,000)

16
 April 11

MRP Results
• MRP is capacity-insensitive
– But, lead times do depend on loading when capacity is finite
– Incentive to inflate lead times
– Results
• increased congestion,
• increased WIP,
• decreased customer service

Inventory and Materials Management

Inventory Management

Production Planning and Control

The Production Control System
Sales
Demand and order Customer
forecasting entry

Materials Shop-floor Shi i

Shipping
Aggregate requirement scheduling Production and
planning planning and control receiving

Inventory
management Inventory Vendors

17
 April 11

Demand Management
• Basic Problem: establish an interface between the customer and the
plant floor, that supports both competitive customer service and
workable production schedules.

• Issues:
– Customer Lead Times: shorter is more competitive.
– Customer Service: on
on-time
time delivery.
– Batching: grouping like product families can reduce lost
capacity due to setups.
– Interface with Scheduling: customer due dates are an
enormously important control in the overall scheduling
process.

Why Manage Inventory?

• Total $ investment on in inventories is $1.37 trillion

(last quarter of 1999)
• 34% in Manufacturing
• 26% in Retail
82% of the total
• 22% in Wholesale
• 8% in Farm
• 10% in Other

Why Manage Inventory?

• In 1998, American companies spent $898 billion in logistics-
related activities (or 10.6% of Gross Domestic Product).
– Transportation 58%
– Inventory 38%
– Management 4%
• By effectively managing inventory:
– Xerox eliminated $700 million inventory from its supply chain
– Wal-Mart became the largest retail company utilizing efficient
inventory management
– GM has reduced parts inventory and transportation costs by
26% annually

18
 April 11

Customers,
Field demand
Sources: Regional Warehouses: centers
plants Warehouses: stocking sinks
vendors stocking points
ports points

Supply

Inventory &
warehousing
costs
Production/ Transportation
purchase costs Transportation
costs Inventory & costs
warehousing
costs

Why Manage Inventory?

• By not managing inventory successfully

– In 1994, “IBM continues to struggle with shortages in their
ThinkPad line” (WSJ, Oct 7, 1994)
– In 1993, “Liz Claiborne said its unexpected earning decline is
the consequence of higher than anticipated excess inventory”
(WSJ, July 15, 1993)
– In 1993, “Dell Computers predicts a loss; Stock plunges. Dell
acknowledged that the company was sharply off in its
forecast of demand, resulting in inventory write downs” (WSJ,
August 1993)

Inventory
• Where do we hold inventory?
– suppliers and manufacturers
– warehouses and distribution centers
– retailers
• Types of Inventory
– WIP and subassemblies
– raw materials
– Spare parts
– finished goods
• Why do we hold inventory? (Short answer)
– Economies of scale
– Uncertainty in supply and demand

19
 April 11

Why do we hold inventory?

• Economies of scale
• Uncertainty in supply and demand
• Speculation
• Transportation
• Smoothing production/purchasing
• Logistics
• Cost of controlling inventory

Decisions to Make
• We have to decide
– How often we review the inventory
– When we should issue a (replenishment/production) order
– How large the order should be

Characteristics of Inv. Systems

• Demand
– Constant (level) or variable
– Deterministic (known) or Stochastic (random or uncertain)

• Lead Time

• Review Time
– Continuous or periodic review

• Excess Demand
– Backordered or lost

• Changing inventory

20
 April 11

Relevant Costs

• Unit value or unit variable cost (c)

– Cost of making a part available for usage
• Purchase + Freight + Mfg. Costs
– Usually different from “accounting” cost
– Should include more than just book value

Relevant Costs

• Holding cost (cost of carrying in inv.)

– Opportunity costs of the money tied to inventory (I = ic),
where i is the available rate of return on investment (may use
IRR).
– Warehousing and Handling (cost of providing space to store
items, counting and moving items in the warehouse)
– Deterioration, damage, obsolescence
– Insurance and taxes
W: Warehousing cost, $ per item per year

Relevant Costs

Inv.
Avg. inv. level

Time, t
1 2

21
 April 11

Relevant Costs

• Ordering or Setup Cost (P)

– Fixed cost
• Independent of the size of the replenishment or
production order
– Ordering forms, phone calls, other communication costs,
g, inspection,
receiving, p , cost of interrupted
p p
production,,
opportunity cost of lost time, etc.

EOQ History

– Introduced in 1913 by Ford W. Harris, “How Many Parts to

Make at Once”

– Interest on capital tied up in wages, material and overhead

sets a maximum limit to the quantity of parts which can be
profitably
y manufactured at one time; “set-up” costs on the jjob
fix the minimum. Experience has shown one manager a way
to determine the economical size of lots.

– Early application of mathematical modeling to Scientific

Management

EOQ Modeling Assumptions

1. Production is instantaneous – there is no capacity
constraint and the entire lot is produced simultaneously.
2. Delivery is immediate – there is no time lag between
production and availability to satisfy demand.
3. Demand is deterministic – there is no uncertainty about
the quantity or timing of demand.
4. Demand is constant over time – in fact, it can be
represented as a straight line, so that if annual demand is
365 units this translates into a daily demand of one unit.
5. A production run incurs a fixed setup cost – regardless of
the size of the lot or the status of the factory, the setup
cost is constant.
6. Products can be analyzed singly – either there is only a
single product or conditions exist that ensure spar-ability of
products.

22
 April 11

EOQ Model
• Time unit: one year
• Total Cost = setup cost + opportunity cost + Warehousing cost,
total cost is calculated per unit.
• Purchase Cost Constant
• Opportunity cost is always based on average quantity
• Warehousing cost may be based on average quantity for mixed
storage areas, or on maximum quantity for dedicated storage.
• Goal: Find the order quantity that minimizes total costs

• General Equation

P ITQavg
TC = + + WT , for dedicated
Q Q storage
P ( I + W )TQavg
TC = + , for mixed storage
Q Q

EOQ Model
Assumptions:
• No Stockouts
• Order when no inventory
• Order size determines
policy
Inventory
y
Qavg = Q/2

Order Avg. Inventory

Quantity (Qavg)
Q

EOQ Model
160
140
120 Total Cost
100
Cost

80
Holding Cost
60
40 Order Cost
20
0
0 500 1000 1500
Optimal Order Order Quantity
Quantity, Q*

23
 April 11

EOQ Model

P ITQavg
TC = + + WT , for dedicated
Q Q storage
P ( I + W )TQavg
TC = + , for mixed storage
Q Q
By differentiation:

2 PD
EOQ = Q * = for dedicated
( I + 2W ) storage

2 PD
EOQ = Q * = for mixed storage
(I + W )

EOQ Observations

• Batching causes inventory (i.e., larger lot sizes translate into

more stock).
• Under specific modeling assumptions the lot size that
optimally balances holding and setup costs is given by the
square root formula:
• Total cost is relatively insensitive to lot size (so rounding for
other reasons, like coordinating shipping, may be attractive).

2 PD
Q* = carrying cost (cc) = I + W or cc = I + 2W
cc

EOQ Trade-off

• Two interpretations:
– If you order more (larger Q), you incur higher
inventory cost, but less setup cost
– If you order less frequently, you incur larger
inventory cost,
cost but less setup cost
• The trade-off is not linear!

24
 April 11

Costs in EOQ Model

ABC – FMS/R Matrix

Moving F M S (R)
Item Class
A
B
C

The Inventory Management Could be better

based on the accuracy of ABC-FMS matrix.

25