Demand Forecasting

1
Demand Forecasting
2
Demand Forecasting

• Pivotal to operations demand management and PSI
  planning
• An unbelievable amount of information exists
• Multiple methods always deepen understanding
  and lower risk.
• Precision is usually more apparent than real
• Goal get close and have contingency plans

3
Forecasting Approaches

• Statistical analysisRegression, Time Series,
  etc.
• Market research
• Conceptual models
• Expert judgment

Complementary not mutually exclusive
4
Quantitative
Qualitative
Judgment
Numbers
Qualitative Methods
Quantitative Methods

• Used when situation is vague little data exist
• New products
• New technology
• Intuition, experience
• e.g., Internet sales

• Used in stable situations when historical data
  exist
• Existing products
• Current technology
• Math / stats techniques
• e.g., color televisions

5
Demand Forecasting
Judgment
Numbers
Quantitative
Qualitative
Model
Disaggregate
Bottom-up Top-down
Extrapolate
Roll-up
6
Demand Forecasting
Judgment
Numbers
Quantitative
Qualitative
Model
Disaggregate
Bottom-up Top-down
Extrapolate
Roll-up
7
Top Down Disaggregation
Industry
Category

Product

Item
8
Top Down Disaggregation
Industry
Company

Product

Item
9
Tyranny of 100
Share gains must come at the expense of specific
competitors (who are very likely to retaliate)
Which competitor(s)? Why? How?
10
Demand Forecasting
Judgment
Numbers
Quantitative
Qualitative
Model
Disaggregate
Bottom-up Top-down
Extrapolate
Roll-up
11
Bottom-up Aggregation

Customer 1
Customer 2
Customer 3

Item
12
Demand Forecasting
Judgment
Numbers
Quantitative
Qualitative
Model
Disaggregate
Bottom-up Top-down
Extrapolate
Roll-up
13
Time Series Analysis
80 70 60 50 40 30 20 10
Actual Projected
Penetration
0 1 2 3 4 5
6 7 8 9 10

Years
14
Time Series AnalysisAnalogous Products
15
Demand Forecasting
Judgment
Numbers
Quantitative
Qualitative
Model
Disaggregate
Bottom-up Top-down
Extrapolate
Roll-up
16
Intent Translation Model

ILLUSTRATIVE
L
TRANSLATION PROSPECTS PERCENT WEIGHT
PROFILE BUYERS Definitely
9 0 1 0 9 Probably 4 0 2 0 8 Migh
t or might not 10 2 0 2 Probably
not 0 1 5 0 Definitely not 0 3 5
0 19
Source Thomas, p.206
17
Linear Regression Model

• Shows linear relationship between dependent
  explanatory variables
• Example Diapers Babies (not time)

Slope
Y-intercept

Y
X


a
b
i
i
Dependent (response) variable
Independent (explanatory) variable
18
Regression Issues

• Illusory correlation
• No cause and effect
• Meaningless coefficients
• Unexplainable variations

19
Sequential Factoring
Total TV Households
Cable Homes
Wired For Cable
Cable/ Baseball Population
Baseball Fanatics
Baseball Pay Per View Market
Premium Service Buyers
A.K.A. Factor Decomposition, Factor Analysis
20
For example

• How much dog food sold annually in the
  U.S.?
• Express answer in

21
Sequential FactoringHow much dog food?

• How many people?
• How many homes?
• Homes with dogs?
• Number of dogs per home?
• Proportion of big little dogs ?
• Daily consumption ? (ounces)
• Ounces per can ?
• Price per can ?

22
How Much Dog Food ?
Big Eats
Little Eats
Dog Food
Big/little split
Big
Dogs / Home
Popul- ation
Little
Dogs
Homes
Homes w/ dogs
People / House
Dogs
23
Demand ForecastingMarket Factoring
MARKET POTENTIAL
MARKET SIZE
MARKET PENETRATION
SALES
MARKET SHARE
24
Market Forecasting
Time Dimension
25
Keys to Success

• Practical precision
• Structured approach
• Multiple methods
• Iterative convergence

26
Demand ForecastingGeneral Principles

• Errors are a certainty
• Aggregate series most stable
• Tendency to over-correct(especially short-run)

27
Demand Forecasting
Market Disaggregation Time Series
Analogies Regression Analysis
Bottom-up Composites
Majority Fallacy
Diffusion Model Intent Translation A-T-R Model
Cannibalization Effect
Value Function Conjoint Analysis Tyranny of 100
28
Demand Forecasting