CTL.

SC1x -Supply Chain & Logistics Fundamentals

Introduction to
Demand Planning
& Forecasting

MIT Center for

Transportation & Logistics

Demand Process Three Key Questions

Demand Planning
n

What should we do to shape and

create demand for our product?

n
n
n

What should we expect demand to

be given the demand plan in place?

Product & Packaging

Promotions
Pricing
Place

Demand Forecasting
n
n
n

Strategic, Tactical, Operational

Considers internal & external factors
Baseline, unbiased, & unconstrained

Demand Management
How do we prepare for and act
on demand when it materializes?

n
n
n

Balances demand & supply

Sales & Operations Planning (S&OP)
Bridges both sides of a firm

Material adapted from Lapide, L. (2006) Course Notes, ESD.260 Logistics Systems.
CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

Forecasting Levels
Level
Strategic

Horizon
Year/Years

Quarterly

Tactical
Months/Weeks

Operational

Days/Hours

Purposes
Business Planning
Capacity Planning
Investment Strategies

Brand Plans
Budgeting
Sales Planning
Manpower Planning

Short-term Capacity Planning

Master Planning
Inventory Planning
Transportation Planning
Production Planning
Inventory Deployment

Material adapted from Lapide, L. (2006) Course Notes, ESD.260 Logistics Systems.
CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

Agenda
Forecasting Truisms
Subjective vs. Objective Approaches
Forecast Quality
Forecasting Metrics

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

Forecasting Truisms 1:
Forecasts are always wrong

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

1. Forecasts are always wrong

Why?
n
n
n

Demand is essentially a continuous variable

Every estimate has an error band
Forecasts are highly disaggregated
w Typically SKU-Location-Time forecasts

Things happen . . .

OK, so what can we do?

n
n
n
n

Dont fixate on the point value

Use range forecasts
Capture error of forecasts
Use buffer capacity or stock

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

Forecasting Truisms 2:
Aggregated forecasts
are more accurate
CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

2. Aggregated forecasts are more accurate

Aggregation by SKU, Time, Location, etc.
Coefficient of Variation (CV)
Definition: Standard Deviation / Mean = /
Provides a relative measure of volatility or uncertainty
CV is non-negative and higher CV indicates higher volatility

n
n
n

Red: =100, =45, CV=0.45

Blue: =100, =1, CV=0.01

200
180

Daily Demand

160
140
120
100
80
60
40
20
2/26/11

3/28/11

CTL.SC1x - Supply Chain and Logistics Fundamentals

4/27/11

5/27/11

6/26/11

Lesson: Demand Forecasting Basics

7/26/11

8/25/11
8

Aggregating by SKU
Coffee Cups and Lids @ the Sandwich Shop
n
n
n

Large
Medium
Small

~N(80, 30)
~N(450, 210)
~N(250, 110)

CV = 0.38
CV = 0.47
CV = 0.44

800
700
600
500

Small

400

Medium

300

Large

200
100
8/14/13

9/13/13

10/13/13

11/12/13

CTL.SC1x - Supply Chain and Logistics Fundamentals

12/12/13

1/11/14

2/10/14

3/12/14

Lesson: Demand Forecasting Basics

4/11/14

5/11/14

6/10/14

Aggregating by SKU
What if I design cups with a common lid?
Common Lid ~N(780, 239) CV = 0.31
n
n

= (80 + 450 + 250) = 780 units/day

= sqrt(302 + 2102 + 1102) = 239 units/day

Large ~N(80, 30) CV=0.38

Med. ~N(450, 210) CV=0.47
Small ~N(250, 110) CV=0.44
Lids

~N(780, 239) CV=0.31

1,600
1,400
1,200
1,000
800
600
400
200
8/14/13

9/13/13

10/13/13

11/12/13

12/12/13

1/11/14

2/10/14

3/12/14

4/11/14

5/11/14

6/10/14

Example of Modularity or Parts Commonality

Reduces the relative variability
Increases forecasting accuracy
Lowers safety stock requirements

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

10

Aggregating by Time
Daily Demand for Lids ~N(780, 239) CV=0.31

1,600
1,200

Forecasts with longer time

buckets have better
forecast accuracy.
The time bucket used
should match the situation.

800
400
8/14/13

9/13/13

10/13/13

11/12/13

12/12/13

1/11/14

2/10/14

3/12/14

4/11/14

5/11/14

6/10/14

Weekly Demand for Lids ~N(5458, 632) CV=0.12

8,000
6,000
4,000
2,000
1

13

17

21

25

29

33

37

41

10

45

49

Monthly Demand for Lids ~N(21840, 1264) CV=0.06

30,000
25,000
20,000
15,000
10,000
5,000
1

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

11

12
11

Aggregating by Locations
Suppose we have three sandwich shops
n

CV reduces as we
aggregate over SKUs,
time, or locations.

Weekly lid demand at each ~N(5458, 632) CV=0.12

~N(5458, 632)

~N(5458, 632)

~N(5458, 632)

~N(16374, 1095)

What if demand is pooled at a common Distribution Center?

n

Weekly lid demand at DC ~N(16374, 1095) CV=0.07

CVind =

CTL.SC1x - Supply Chain and Logistics Fundamentals

CVagg

CVind
n

=
=
=
n n
n

Lesson: Demand Forecasting Basics

12

Forecasting Truisms 3:
Shorter horizon forecasts
are more accurate
CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

13

3. Shorter horizon forecasts are

more accurate
?
?
?
?

21

22

23

24

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

10

11

12

14

3. Shorter horizon forecasts are

more accurate
Postponed final customization to
closer time of consumption
Risk pooling of component (e.g.,
ham) increases forecast accuracy.
?

21

22

23

24

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

10

11

12

15

Forecasting Truisms
Forecasts are always wrong
Use ranges & track forecast error
Aggregated forecasts are more accurate
Risk pooling reduces CV
Shorter time horizon forecasts are more
accurate
Postpone customization until as
late as possible

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

16

Subjective & Objective Approaches

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

17

Fundamental Forecasting Approaches

Subjective

Objective

Judgmental
n
n
n

Causal / Relational

Sales force surveys

Jury of experts
Delphi techniques

Experimental
n
n
n

n
n
n

Econometric Models
Leading Indicators
Input-Output Models

Time Series

Customer surveys
Focus group sessions
Test marketing

n
n
n

Black Box Approach

Past predicts the future
Identify patterns

Often times, you will need to use a combination of approaches

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

Forecasting Quality

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

19

Cost of Forecasting vs Inaccuracy

Overly Nave Models

Good Region

Excessive Causal Models

Cost

Total Cost

Cost of Errors
In Forecast

Cost of Forecasting

Forecast Accuracy
CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

How do we determine if a forecast is good?

What metrics should we use?
Example - Which is a better forecast?
n
n

Squares & triangles are different forecasts

Circles are actual values

1100

1000

900

time
CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

Accuracy versus Bias

n
n

Accuracy - Closeness to actual observations

Bias - Persistent tendency to over or under predict

Accurate

Not Accurate
Biased
CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

Not Biased

Forecasting Metrics

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

23

Forecasting Metrics
Mean Deviation
(MD)

Mean Absolute
Deviation (MAD)

MD =
Mean Squared
Error (MSE)

Mean Percent
Error (MPE)

n
2
t

n
n

et
A
MPE = t=1 t
n

Mean Absolute
Percent Error (MAPE)

MAPE =

At = Actual value for obs. t

Ft = Forecasted value for obs. t

CTL.SC1x - Supply Chain and Logistics Fundamentals

t =1

n
n

RMSE =

t =1

Root Mean
Squared
Error (RMSE)

Notation:

MAD =

t =1

MSE =

et = At Ft

2
e
t
t =1

et

A
t=1

et = Error for observation t

n = Number of observations

Lesson: Demand Forecasting Basics

Example: Forecasting Bagels

For the bagel forecast and actual values
shown below, find the:
n
n
n

Mean Absolute Deviation (MAD)

Root Mean Square of Error (RMSE)
Mean Absolute Percent Error (MAPE)

Forecast

Actual

Monday

50

43

Tuesday

50

42

Wednesday

50

66

Thursday

50

38

Friday

75

86

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

MAD =

t =1

n
n

RMSE =

2
e
t
t =1

et

MAPE =

A
t=1

n
25

Example: Forecasting Bagels

Solution:

90

1. Graph it.
2. Extend data table:
w Error: et=At-Ft
w Abs[error] = |et|
w Sqr[error] = e2
w AbsPct[error] = |et/At|
3. Sum and find means

Forecast

Daily Bagel Demand

80

70

60

50

40

Ft

At

et |et|

e2

Monday

50

43

-7

49

16.3%

Tuesday

50

42

-8

64

19.0%

Wednesday

50

66

16

16

256

24.2%

Thursday

50

38 -12

12

144

31.6%

Friday

75

|et/At|

86 11 11 121 12.8%
Sum 0
54 634
104%
Mean 0 10.8 126.8 21%

CTL.SC1x - Supply Chain and Logistics Fundamentals

Actual

30
Monday

Tuesday

Wednesday Thursday

Friday

MAD = 54/5 = 10.8

RMSE = sqrt(126.8) = 11.3
MAPE = 104%/5 = 21%

Lesson: Demand Forecasting Basics

26

Key Points from Lesson

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

27

Key Points
Forecasting is a means not an end
Forecasting Truisms
n
n
n

Forecasts are always wrong

Aggregated forecasts are more accurate
Shorter horizon forecasts are more accurate

Subjective & Objective Approaches

n
n

Judgmental & experimental

Causal & time series

Forecasting metrics
n
n

Capture both bias & accuracy

MAD, RMSE, MAPE

CTL.SC1x - Supply Chain and Logistics Fundamentals

Lesson: Demand Forecasting Basics

CTL.SC1x -Supply Chain & Logistics Fundamentals

Questions, Comments, Suggestions?

Use the Discussion!

Janie
Photo courtesy Yankee Golden
Retriever Rescue (www.ygrr.org)

MIT Center for

Transportation & Logistics

caplice@mit.edu