Demand Forecasting in a Supply Chain

By Tadesse Gudeta

May 6,2021

1
Learning objectives
• Describe the role of forecasting for both an enterprise
and a supply chain.
• Identify the components of a demand forecast.
• Forecast demand in a supply chain given historical
demand data using time-series methodologies.
• Analyze demand forecasts to estimate forecast error.

2
Over view of Forecasting Methods

3
Forecasting Steps

• What needs to be forecasted?

• Level of detail, units of analysis, & time horizon
required
• What data is available to evaluate?
• Identify needed data & whether it’s available
• Select and test the forecasting model
• Cost, ease of use, & accuracy
• Generate the forecast
• Monitor forecast accuracy over time

4
Types of Forecasting Models

• Qualitative methods:
• Forecasts generated subjectively by the forecaster

• Quantitative methods:
• Forecasts generated through mathematical modeling

5
Qualitative Methods
Type Characteristics Strengths Weaknesses
Executive A group of managers Good for strategic or One person's opinion
opinion meet & come up with new-product can dominate the
a forecast forecasting forecast

Market Uses surveys & Good determinant of It can be difficult to

research interviews to identify customer preferences develop a good
customer preferences questionnaire

Delphi Seeks to develop a Excellent for Time consuming to

method consensus among a forecasting long-term develop
group of experts product demand,
technological
changes, and
scientific advances

6
Quantitative Methods

• Time Series Models:

• Assumes the future will follow same patterns as the
past
• Causal Models:
• Explores cause-and-effect relationships
• Uses leading indicators to predict the future
• E.g. impact of price promotions

7
Time Series Data Composition

• Data = systematic component+ random components

• systematic component consists:
Level (long-term average)
Trend
Seasonality
Cycle
• Random components cannot be predicted

8
Time Series Patterns

9
 Time Series Models
• Naive: Ft 1  At
• The forecast is equal to the actual value observed
during the last period – good for level patterns
• Simple Mean: Ft 1   A t / n
• The average of all available data - good for level
patterns
• Moving Average: Ft 1   A t / n
• The average value over a set time period
(e.g.: the last four weeks)
• Each new forecast drops the oldest data point & adds
a new observation
• More responsive to a trend but still lags behind
actual data
10
Naïve Ft 1  At

• Examples:
• If last week’s demand was 50 units, the naive
forecast for the coming week is 50 units.

11
Simple mean Ft 1   A t / n

• Example 2: Look at these numbers:

• 3, 7, 5, 13, 20, 23, 39, 23, 40, 23, 14, 12, 56, 23, 29
• The sum of these numbers is 330
• There are fifteen numbers.
• The mean is equal to 330 / 15 = 22
• The mean of the above numbers is 22

12
Simple Moving Average:
Example 2.1:
An OB/GYN clinic has the following yearly patient visits, and
would like to predict the volume of business for the next year
for budgeting purposes.

Period (t) Visits

1 15908
2 15504
3 14272
4 13174
5 10022

13
Moving Averages, cont.
Solution:
The three-period moving average (MA3) for period 6 is

F6 = MA3 = (14272+13174+10022) ÷ 3 = 12489.3

Period (t) Visits Forecast
1 15908
2 15504
3 14272
4 13174 15228
5 10022 14317
6 12489
14
 Weighted Moving Average: Ft 1   C t A t

• All weights must add to 100% or 1.00

e.g. Ct .5, Ct-1 .3, Ct-2 .2 (weights add to 1.0)
• Allows emphasizing one period over others; above
indicates more weight on recent data (Ct=.5)
• Differs from the simple moving average that weighs all
periods equally - more responsive to trends

15
Weighted Moving Average cont’d
Example 2.1:
Continuing with Example 2.1; since there is a downward trend in
visits and in period 5 there is a sharp decline, a weight of .5 or
even higher is justified by the healthcare manager to calculate a
weighted average for period 6

Period (t) Visits Weights

1 15908
2 15504
3 14272 0.2
4 13174 0.3
5 10022 0.5
6 16
Weighted Moving Average cont’d

Solution:
In this analysis, a weighted average, for the
OB/GYN clinic for the
period 6 would be:
F6 = 14272*.2+13174*.3+10022*.5
F6 = 11818.

• How to workout using excel

17
 Simple Exponential Smoothing (SES):
Ft 1  αA t   1  α  Ft
• Most frequently used time-series method because of
ease of use and minimal amount of data needed
• Need just three pieces of data to start:
• Last period’s forecast (Ft)
• Last periods actual value (At)
• Select value of smoothing coefficient, ,between 0
and 1.0
• If no last period forecast is available, average the last
few periods or use naive method

18
SES……….. Ft 1  αA t  1  α  Ft

• Examples: Given the following data, calculate forecasts for

months 4, 5, 6, and 7 using an exponential smoothing
forecast method.
Month Actual Forecast exponential
smoothing
1 56
2 76
3 58
4 67
5 75
6 76
7
19
Solution Ft 1  αA t   1  α  Ft

• Lets take alpha () 0.4 F2+1=A2+(1-)F2

(will be optimized) F3=0.4(76)+(1-0.4)56
• Initial forecast (F0) is =64
the same as the actual
demand of the same
period (A0) •The same procedure will
be used to determine
• A0=F0 F4,F5,F6,F7
• F1+1=A1+(1-)F1
• F2=A1+(1-)F1 •How to calculate using ex
• F2=0.4(56)+(1-0.4)56 cel
=56
20
 Measuring Forecast Error

• Forecasts are never perfect

• Need to know how much we should rely on our
chosen forecasting method
• Measuring forecast error:

E t  A t  Ft
• Note that over-forecasts = negative errors and
under-forecasts = positive errors

21
Measuring Forecasting Accuracy

• Mean Absolute Deviation

(MAD) MAD 
 actual  forecast
• measures the total error in n
a forecast without regard
to sign

Mean Square Error (MSE)

• The forecast technique giving
the lowest MAD/MSE is
  actual - forecast 
2
preferred MSE 
• MSE magnifies large errors n
through the squaring process

22
Forecasting Accuracy….

• Cumulative Forecast
Error (CFE)
• Measures any bias in
the forecast

CFE
• Tracking Signal TS 
MAD

23
Forecasting Accuracy….

• A way to monitor forecast accuracy is by comparing

a value of TS against predetermined control limits
(usually +/-4 MAD)

24
Tracking signal……

Signal exceeded limit

Tracking signal
Upper control limit = +4MAD
+

0 0 MAD

-
Lower control limit = -4MAD

Time
Over view of Holt’s trend and Holts-
Winter’s models

26
Introduction

• Time series models, any observed demand can be

broken down into a systematic and a random
component:

27
…..continued
• Random component cannot be forecasted
• are sudden changes occurring in a time series which
are unlikely to be repeated
• are components of a time series which cannot be
explained by trends, seasonal or cyclic movements
• It’s among the components of time series

28
…..continued
• Components for Time Series Analysis
• A set of observations ordered with respect to the
successive time periods is a time series.
• The various reasons or the forces which affect the values
of an observation in a time series are the components of a
time series.
• The four categories of the components of time series are,
a) Trend
b) Seasonal Variations
c) Cyclic Variations
d) Random or Irregular movements

29
Trend
The trend shows the general tendency of the data to
increase or decrease during a long period of time.

A B

30
 Seasonal Variations
• Has regular peaks and troughs
– We can predict based on the previous pattern

31
Seasonal V……e.g1

32
Seasonal V……e.g2

33
Cyclic Variations

34
Cyclic Variations-e.g1

35
Cyclic Variations-e.g2

36
Random variation
• Does not have any obvious peaks and troughs
happening for particular reason.

37
Workout

38
 Equation for calculating the systematic
component
• The equation for calculating the systematic
component may take a variety of forms:
• Additive= Level+Trend+Seasonality
• Multiplicative=Level*Trend*Seasonality
• Mixed=(Level+Trend)* Seasonality

39
Cont.d….

• The additive trend model assumes that the level

changes from one period to the next by a fixed
amount.
• In contrast, a multiplicative trend model assumes
that the level changes from one period to the
next by a factor.
• E.g1:additive

40
Cont.d….

• E.g.1: Multiplicative

41
Cont.d….

42
Cont.d….

43
Cont.d….

44
Cont.d….

45
Holt’s trend
• Holt’s model method is appropriate when demand
is assumed to have a level and a trend in the
systematic component, but no seasonality
• In this case, we have systematic component of
demand = level + trend

46
Holt’s trend

• Double exponential • Ft=value at time “t”

smoothing • Lt-1=level at time “t-1”
• Tt-1=trend value at “t-1”
•  and  are smoothing
parameters (01 &
01)
• “h” is forecast horizon
– We will not be using it
for training sample
forecast but for test
sample forecast

47
Holt’s model-e.g

How to use excel for forecasting

48
Winter’s model

• It’s appropriate when the systematic component

of demand has a level, a trend, and a seasonal
factor,

49
Winter’s model…..

50
Winter’s model- e.g.

How to use excel for forecasting

51
Selecting the Right Forecasting Model

• The amount & type of available data

• Some methods require more data than others
• Degree of accuracy required
• Increasing accuracy means more data
• Length of forecast horizon
• Different models for 3 month vs. 10 years
• Presence of data patterns
• Lagging will occur when a forecasting model meant
for a level pattern is applied with a trend

52
The end

53