LESSON 8

Introduction to operations management

Demand forecasting
Codex : La gestion des opérations par William, J. Stevenson, Mehran Hojati et
James Cao. Adaptation de Marie-Claude Bolduc, Chenelière Éducation, 2021.
Pages 112-157 (du livre) correspond aux : Pages 1-28 (du codex)
AOT 4200 - Introduction to operations management 1
 Outline of the session

Lesson Topic

8 Demand forecasting methods

Assignment 1
9 Inventory Management - A Strategic Asset

10 Global planning in the company – Theory QUIZ 3

11 Global planning in the company – Practice

12 Scheduling of operations: priority management

13 Operations management in the digital age

14 Material review for final exam QUIZ 4

Assignment 2
15 Final exam

AOT 4200 - Introduction to operations management 2

 Outline

• Summary of the previous session

• Lesson objectives
• Context and basic concepts
• Demand forecasting process
• The different types of forecasting methods
• Time series – methods based on averages
• Error measures

AOT 4200 - Introduction to operations management 3

 Summary from previous lesson
• Quality dimensions: Product vs Service
• Cost of poor quality (iceberg): visible (direct) vs invisible (indirect)
• QC (Quality Control: product focus) vs QA (Quality Assurance: process focus)
• Approaches:
• Six Sigma: reduce variability
• Lean: eliminate non-value-added activities (wastes) / 8 mudas (DOWNTIME)
• DMAIC / PDCA (methodology) – DMAIC = Define, Measure, Analyze, Improve, Control. PDCA = Plan, Do, Check, Act
• Tools:
• 5W2H – Define
• Process mapping + KPI - Measure
• Ishikawa (XM, Cause-Effect diagram, Fishbone diagram) - Analyze
• Pareto (20-80 rule) - Analyze
• 5 whys (root-cause) – Analyze
• 5S (workstation) – Improve
• Poka-Yoke (anti-error mechanism)
• Statistical process control – Monitor
• Sampling – Improve (minimize cost of control)
• Standardization
• ISO (recognized international standards for Quality management)
• Quiz 2: plagiarism reporting

AOT 4200 - Introduction to operations management 4

 Global objective

Describe the strategic role of demand

forecasting in value chain management and
identify the main demand forecasting
methods used in different business contexts

AOT 4200 - Introduction to operations management 5

 Specific objectives

• Explain the role and significance of demand

forecasting;
• Describe the demand forecasting process;
• Distinguish between qualitative and quantitative
forecasting methods;
• Apply some basic methods for demand forecasting;
• Describe the precision measures used to assess the
accuracy of demand forecasting methods.

https://www.clickpost.ai/blog/what-is-demand-management-in-ecommerce

AOT 4200 - Introduction to operations management 6

DEMAND FORECASTING
Context and basic concepts

AOT 4200 - Introduction to operations management 7

 Context
Forecasting demand at Ocean Spray

« One of the challenges faced is the

seasonality of cranberry sales as
approximately 90% of sales occur in the fall,
mostly before Thanksgiving and
Christmas. »

AOT 4200 - Introduction to operations management 8

 SIRTOC3: The Operational System
Resources organization
Sales and Operations
Planning (S&OP)

Goal : Create / add value

Control

Customers
Suppliers

Resources

Inputs Outputs

Constraints

AOT 4200 - Introduction to operations management 9

 Context

https://www.youtube.com/watch?v=kGQ1fNQVbj8
AOT 4200 - Introduction to operations management 10
Context

https://strategok.com/the-lost-link-to-make-strategy-work-sales-operations-planning-sop/

AOT 4200 - Introduction to operations management 11

Context
MRP process flow

https://slideplayer.com/slide/11953423/

AOT 4200 - Introduction to operations management 12

 Introduction
Demand

Independent
demand
Level 0 Table (random) :
End products

Dependent
Level 1 Surface Support demand
(deterministic) :
Raw materials,
Hinge and components
Level 2 Top Box apron Legs
nails

It is important to forecast demand to plan operations and properly

manage inventory.
AOT 4200 - Introduction to operations management 13
 Demand forecast within the organization

Fonction Uses
Accounting • Process/Costs estimates, profit projections, cash
management
Finance • New equipment/Equipment replacement needs, timing
and amount of funding/borrowing needs
Human Resources • Hiring activities, layoff planning, training

Marketing • Pricing • E-business strategies

• Promotion campaign • Competition strategies
Management • New/revised information systems
Information System • Internet services
Operations • Schedules, capacity planning, work assignments and
workloads, inventory planning
Product and Service • Outsourcing, project management
Design • Revision of current features, design of new products
and services
AOT 4200 - Introduction to operations management 14
 Features of demand forecasting methods

• Forecasting methods assume a causal system observed in the past and

influencing demand behavior in the future;
• Forecasts are rarely accurate;
• Forecasts for product families generally perform better than individual
forecasts (Delayed differentiation strategy);
• The accuracy of a forecast decreases as the horizon of the forecast
increases.

AOT 4200 - Introduction to operations management 15

 Characteristics of good forecasts

• Available on time (plan for response time to demand - implementation

of changes)
• Precise (the degree of precision must be indicated - allows users to
account for uncertainty)
• Expressed in significant units (quantity, $, etc.)
• Profitable
• Obtained by a reliable, rigorous, and consistent method
• Obtained by a method that is easy to understand and use

AOT 4200 - Introduction to operations management 16

 Demand forecasting horizon

Long term Medium term Short term

Annual data Monthly data from Daily or weekly data from a
Horizon for data 2 to 5 years several months to 2 years couple of days to a couple
collection of weeks

• Market knowledge • Historical data • Historical data

• Good judgment • Historical analogy • Historical analogy
• Consumer survey • Expert knowledge • Expert knowledge
Require
• Control markets

• Global decisions related • Production planning for • Scheduling

to capacity, facilities and the following year • Daily operations planning
Are used for resources. • Supply

AOT 4200 - Introduction to operations management 17

DEMAND FORECASTING
Demand forecasting process

AOT 4200 - Introduction to operations management 18

 Demand forecasting process

1. Define the purpose of forecasts

2. Setting the forecast horizon
3. Collect and analyze relevant historical data
4. Choose one (or more) forecasting method(s) – depends on
expected precision
5. Perform forecasts and compute error measures
6. Follow the evolution of forecasts (PDCA)
There must be consistency between the steps!

AOT 4200 - Introduction to operations management 19

DEMAND FORECASTING
Demand behaviour

AOT 4200 - Introduction to operations management 20

 Demand behaviour

1. The level (or average or constant): value representing all the data

2. The trend: overall evolution of the data (up or down)

Demand

Random variation

The trend Time

AOT 4200 - Introduction to operations management 21

 Demand behaviour
3. The seasonal variation: periodic, repetitive, regular and fairly predictable
fluctuation observable over a short period of time

Demand

Time
Seasonality

AOT 4200 - Introduction to operations management 22

 Demand behaviour
4. The cycle: observable fluctuation over a long period (usually several years
- structural)

Demand

Cycle Time

AOT 4200 - Introduction to operations management 23

 Demand behaviour

5. The irregular variations: come from unusual but explainable situations.

6. The random variations (noise): fluctuations that cannot be anticipated,
avoided or explained.

AOT 4200 - Introduction to operations management 24

 Demand behaviour
Classroom exercise

Demand component Example

Trend

Seasonality

Irregular variations

Random variations

Cycle

AOT 4200 - Introduction to operations management 25

 Demand behaviour
Classroom exercise

Demand component Example

Demographic changes, cultural changes, income

Trend
variations, etc.
Sale of ice cream, bicycles, garden products, air
Seasonality
conditioning, shovel, snowblower, coats, etc.
Strike, power outage, ice, storm, flood, heat
Irregular variations
wave, COVID, etc.

Random variations

Cycle Economic

AOT 4200 - Introduction to operations management 26

 DEMAND FORECASTING
Qualitative and quantitative methods

AOT 4200 - Introduction to operations management 27

 Qualitative methods

• Based on the expertise and opinion of several stakeholders in a given field

(managers, sellers, experts and consumers)

• Use historical analogies of similar products (new product launch)

• Are used when little or no data is available.

AOT 4200 - Introduction to operations management 28

 Quantitative methods

• Time series :
Use historical observations (data points) at regular intervals over a
period of time to develop future forecasts.

• Associative models :
Use one or more causal variables to predict future demand.

AOT 4200 - Introduction to operations management 29

 DEMAND FORECASTING
Quantitative methods based on time series

AOT 4200 - Introduction to operations management 30

 Quantitative methods
Naïve method

• The demand for the current period is used as a demand forecast for the following period.

• Example :
- If the current demand (week 1) is 100 units, then the demand forecast for next
week (week 2) will be 100 units.
- If the actual demand for week 2 turned out to be 90 units, then the demand
forecast for the following week (week 3) will be 90 units and so on.

• The naïve forecast does not take into account the behavior of the demand history.
•

AOT 4200 - Introduction to operations management 31

 Quantitative methods
Simple average

• Consider the following variables :

𝑡 = Forecast period number

𝑃𝑡 = Forecast for period 𝑡
𝑅𝑖 = Demand realization for period 𝑖

The forecast is obtained from the average of all historical data:

t −1

R i
Pt = i =1 [units]
t −1

AOT 4200 - Introduction to operations management 32

 Simple average
Example

• A hardware store has compiled the demand for shovels in recent months:
Month Demand
1 492
2 470
3 493
4 485
5 498
6

a. What will be the forecast for the 6th month?

AOT 4200 - Introduction to operations management 33

AOT 4200 - Introduction to operations management 34
 Simple average
Example

a. What will be the forecast for the 6th month?

t −1

R i
Pt = i =1

t −1

σ5𝑖=1 𝑅𝑖
𝑃6 =
5

AOT 4200 - Introduction to operations management 35

 Simple average
Example

a. What will be the forecast for the 6th month?

σ5𝑖=1 𝑅𝑖
𝑃6 =
5

𝑅1 + 𝑅2 + 𝑅3 + 𝑅4 + 𝑅5
𝑃6 =
5
492 + 470 + 493 + 485 + 498
𝑃6 =
5

𝑃6 = 487,6 units = 488 units

AOT 4200 - Introduction to operations management 36

 Quantitative methods
Moving average
Take the most recent data into account

• The moving average is obtained from the last n historical data:

R t −i
[units]
Pt = i =1

𝑡 = Forecast period number

𝑃𝑡 = Forecast for period 𝑡
𝑅𝑡−𝑖 = Demand realization for period 𝑡 − 𝑖
𝑛 = number of periods in the moving average (n < t )

AOT 4200 - Introduction to operations management 37

 Moving average
Example
• The manager of a convenience store has compiled the number of shopping baskets used
in recent months : Month Demand
1 42
2 40
3 43
4 40
5 41
6

a. Using a 3-period moving average, forecast the number of food baskets needed for the
6th month.
b. If the demand for the 6th month turns out to be 39, does this have an influence on the
forecast for the 7th month (using a 3-period moving average)?
AOT 4200 - Introduction to operations management 38
AOT 4200 - Introduction to operations management 39
 Moving average
Example

a) Using a 3-period moving average, forecast the number of food baskets needed for the 6th
month.
n

Month Demand
R t −i
Pt = i =1

1 42 n
2 40
𝑡 = Forecast period number = 6
3 43
𝑃𝑡 = Forecast for period 𝑡 = 𝑃6
4 40 𝑅𝑡−𝑖 = Demand realization for period 6 − 𝑖
5 41 𝑛 = number of periods in the moving average = 3
6
σ3𝑖=1 𝑅6−𝑖
𝑃6 =
3

AOT 4200 - Introduction to operations management 40

 Moving average
Example

a) Using a 3-period moving average, forecast the number of food baskets needed for the 6th
month.
σ3𝑖=1 𝑅6−𝑖
𝑃6 =
Month Demand 3
1 42
𝑅6−1 + 𝑅6−2 + 𝑅6−3
2 40 𝑃6 =
3
3 43
𝑅5 + 𝑅4 + 𝑅3
4 40 𝑃6 =
3
5 41
41 + 40 + 43
𝑃6 =
3

𝑃6 = 41,33 = 41

AOT 4200 - Introduction to operations management 41

 Moving average
Example
a) If the demand for the 6th month turns out to be 39, what is the demand forecast for the
7th month (using a 3-period moving average)?
By knowing the demand of the 6th month 𝑅6 = 39, we are able to calculate the forecast of the 7th month:

Month Demand
1 42 𝑅6 + 𝑅5 + 𝑅4
𝑃7 =
3
2 40
3 43 39 + 41 + 40
𝑃7 =
4 40 3
5 41 𝑃7 = 40 Baskets.
6 39
7

AOT 4200 - Introduction to operations management 42

 N-period impact on the forecast
Smaller time periods Larger time periods
• The moving average is sensitive (reactive) to • The moving average is not sensitive to the
the most recent data most recent data
• Fast forecast adjustment to changes in • The forecast is smoother than actual data
demand and lag actual data

A 2-period moving average A 6-period moving average

Units Units

Period Period
Legend: Demand Forecast with a 2-period moving average Legend: Demand Forecast with
with aa 6-period
6-period moving
moving average
average
Forecast

AOT 4200 - Introduction to operations management 43

 Quantitative methods
Weighted moving average
• Forecast obtained using the average of the last n historical data whose importance varies
using weights
n
Pt =  wi Rt −i
n

i =1
with w =1
i =1
i
[units]

𝑡 = Forecast period number

𝑃𝑡 = Forecast for period 𝑡
𝑅𝑡−𝑖 = Demand realization for period 𝑡 − 𝑖
𝑤𝑖 = Weight associated with the demand realization of period 𝑡 − 𝑖

• The first weight (w1) affects the importance given to the most recent demand (Rt-1), the
second weight (w2) affects the importance given to the second most recent demand (Rt-2),
and so on.
AOT 4200 - Introduction to operations management 44
 Weighted moving average
Example
The table below presents the demand for raw material over a period of 5 months.
Month Demand
1 38
2 40
3 43
4 40
5 41
6

a) Calculate the forecast for the 6th month using a 4-period weighted moving
n average by
applying the following weighting :  wi = 1 i =1
w1 = 0,40; w2 = 0,30; w3 = 0,20 et w4 = 0,10.
b) If the demand for the 6th month is 39, what would be the forecast of the demand for the
7th month? (using the same parameters)
AOT 4200 - Introduction to operations management 45
AOT 4200 - Introduction to operations management 46
 Weighted moving average
Example

a) Calculate the forecast for the 6th month using a 4-period weighted moving average by
applying the following weighting : w1 = 0,40, w2 = 0,30, w3 = 0,20 et w4 = 0,10.
4
Month Demand 𝑃6 = ෍ 𝑤𝑖 𝑅6−𝑖
1 38 𝑖=1

2 40 𝑃6 = 𝑤1 𝑅6−1 + 𝑤2 𝑅6−2 + 𝑤3 𝑅6−3 + 𝑤4 𝑅6−4

3 43
4 40 𝑃6 = 𝑤1𝑅5 + 𝑤2𝑅4 + 𝑤3𝑅3 + 𝑤4𝑅2
5 41 𝑃6 = 0,4 ∗ 41 + 0,3 ∗ 40 + 0,2 ∗ 43 + (0,1 ∗ 40)
6
𝑃6 = 41 raw material units.

AOT 4200 - Introduction to operations management 47

 Weighted moving average
Example

b) If the demand for the 6th month is 39, what would be the forecast of the demand for the
7th month? (using the same parameters)
4
Month Demand 𝑃7 = ෍ 𝑤𝑖 𝑅7−𝑖
1 38 𝑖=1

2 40 𝑃7 = 𝑤1 𝑅7−1 + 𝑤2 𝑅7−2 + 𝑤3 𝑅7−3 + 𝑤4 𝑅7−4

3 43
4 40 𝑃7 = 𝑤1𝑅6 + 𝑤2𝑅5 + 𝑤3𝑅4 + 𝑤4𝑅3
5 41 𝑃7 = 0,4 ∗ 39 + 0,3 ∗ 41 + 0,2 ∗ 40 + (0,1 ∗ 43)
6 39
7 𝑃7 = 40,2 raw material units.

How to choose historical data weights?

By taking into account the precision of the method.
AOT 4200 - Introduction to operations management 48
 Quantitative methods
Exponential smoothing

• Method also based on the principle of the weighted moving average but more
sophisticated than this method.
• Uses forecast error to adjust the calculation of forecasts from one period to the next.
• To calculate the forecast error 𝒆𝒕 for a given period 𝑡, we must know the actual demand
Rt and the forecast of this demand Pt. Previous example:
P6 = 41, R6 = 39 [units]
et = Rt – Pt e6 = R6 - P6 = 39 – 41 = -2

• The forecast is obtained with the formula:

Pt =  Rt −1 + (1 −  ) Pt −1 = Pt −1 +  ( Rt −1 − Pt −1 ) = Pt −1 +  et −1 [units]

where: 𝛼 = Smoothing constant (0 < 𝛼 < 1) The smoothing constant determines how quickly the forecast is
adjusted (near 0 = little importance to forecast errors)
𝑒𝑡 = forecast error for period 𝑡 When 𝛼 = 0, Pt = Pt-1

AOT 4200 - Introduction to operations management 49

 Exponential smoothing
Example
• The table below shows the demand for chairs from a furniture store.
•

Period (t) 1 2 3 4 5 6 7 8 9 10 11

Demand (R) 42 40 43 40 41 39 46 44 45 38 40

• Calculate the forecast of the 12th period using simple exponential smoothing with two
different alpha values: α = 0,10 et α = 0,40.

AOT 4200 - Introduction to operations management 50

 Exponential smoothing
Example

To compute 𝑃12 , I need 𝑃11 ∶

𝑃12 = 𝛼𝑅11 + 1 − 𝛼 𝑃11

But to get 𝑃11 , I need 𝑃10 :

𝑃11 = 𝛼𝑅10 + 1 − 𝛼 𝑃10

And so on…

For exponential smoothing, it is important to initialize the model by setting 𝑃2 = 𝑅1 .

Initialization must be done with the oldest data.

AOT 4200 - Introduction to operations management 51

AOT 4200 - Introduction to operations management 52
 Exponential smoothing Given (from table on page
Imposed (page 50)
Example 50)

a. α = 0,1 α = 0,1
Period Demand Forecast Forecast error
(t) (Rt) (Pt) (et)
We must start by setting: 1 42
P2 = R1. 2 40

3 43
Add column Forecast Add column Forecast error
(Pt) (et)
4 40

5 41
6 39
7 46
8 44
9 45
10 38
11 40
12

AOT 4200 - Introduction to operations management 53

 Exponential smoothing Given (from table on page
Imposed (page 50)
Example 50)

a. α = 0,1 α = 0,1
Period Demand Forecast Forecast error
(t) (Rt) (Pt) (et)
We must start by setting: 1 42
P2 = R1. 2 40 P2 = R1 = 42

3 43

4 40

5 41
6 39
7 46
8 44
9 45
10 38
11 40
12

AOT 4200 - Introduction to operations management 54

 Exponential smoothing Given (from table on page
Imposed (page 50)
Example 50)

a. α = 0,1 α = 0,1
Period Demand Forecast Forecast error
(t) (Rt) (Pt) (et)
We must start by setting: 1 42
P2 = R1. 2 40 42 e2 = R2 - P2
e2 = 40-42
= -2
3 43

4 40

5 41
6 39
7 46
8 44
9 45
10 38
11 40
12

AOT 4200 - Introduction to operations management 55

 Exponential smoothing Given (from table on page
Imposed (page 50)
Example 50)

a. α = 0,1 α = 0,1
Period Demand Forecast Forecast error
(t) (Rt) (Pt) (et)
We must start by setting: 1 42
P2 = R1. 2 40 42 e2 = R2 - P2
e2 = 40-42
= -2
3 43 P3 = P2 + αe2
= 42 + 0,1(-2)
= 41,8
4 40

5 41
6 39
7 46
8 44
9 45
10 38
11 40
12

AOT 4200 - Introduction to operations management 56

 Exponential smoothing Given (from table on page
Imposed (page 50)
Example 50)

a. α = 0,1 α = 0,1
Period Demand Forecast Forecast error
(t) (Rt) (Pt) (et)
We must start by setting: 1 42
P2 = R1. 2 40 42 e2 = R2 - P2
e2 = 40-42
= -2
3 43 P3 = P2 + αe2 e3 = R3 - P3
= 42 + 0,1(-2) e3 = 43- 41,8
= 41,8 = 1,2
4 40

5 41
6 39
7 46
8 44
9 45
10 38
11 40
12

AOT 4200 - Introduction to operations management 57

 Exponential smoothing Given (from table on page
Imposed (page 50)
Example 50)

a. α = 0,1 α = 0,1
Period Demand Forecast Forecast error
(t) (Rt) (Pt) (et)
We must start by setting: 1 42
P2 = R1. 2 40 42 e2 = R2 - P2
e2 = 40-42
= -2
3 43 P3 = P2 + αe2 e3 = R3 - P3
= 42 + 0,1(-2) e3 = 43- 41,8
= 41,8 = 1,2
4 40 P4 = 41,8 +0,1*(1,2)
= 41,92
5 41
6 39
7 46
8 44
9 45
10 38
11 40
12

AOT 4200 - Introduction to operations management 58

 Exponential smoothing Given (from table on page
Imposed (page 50)
Example 50)

a. α = 0,1 α = 0,1
Period Demand Forecast Forecast error
(t) (Rt) (Pt) (et)
We must start by setting: 1 42
P2 = R1. 2 40 42 e2 = R2 - P2
e2 = 40-42
= -2
3 43 P3 = P2 + αe2 e3 = R3 - P3
= 42 + 0,1(-2) e3 = 43- 41,8
= 41,8 = 1,2
4 40 P4 = 41,8 +0,1*(1,2) e4 = 40 - 41,92
= 41,92 = -1,92
5 41
6 39
7 46
8 44
9 45
10 38
11 40
12

AOT 4200 - Introduction to operations management 59

 Exponential smoothing Given (from table on page
Imposed (page 50)
Example 50)

a. α = 0,1 α = 0,1
Period Demand Forecast Forecast error
(t) (Rt) (Pt) (et)
We must start by setting: 1 42
P2 = R1. 2 40 P2 = R1 = 42 e2 = R2 - P2
e2 = 40-42
= -2
3 43 P3 = P2 + αe2 e3 = R3 - P3
= 42 + 0,1(-2) e3 = 43- 41,8
= 41,8 = 1,2
4 40 P4 = 41,8 +0,1*(1,2) e4 = 40 - 41,92
= 41,92 = -1,92
5 41 41,728 -0,728
6 39 41,6552 -2,6552
7 46 41,3897 4,6103
8 44 41,8507 2,1493
9 45 42,0656 2,9344
After calculations, we 10 38 42,3591 -4,3591
41 - 41.728
11 40 41,9232 -1,9232
obtain P12 = 41,73 chairs. 12 41,7309 Keep the decimals

41.92 + 0.1 * (-1.92)

AOT 4200 - Introduction to operations management Keep the decimals 60

 Exponential smoothing
Example
α = 0,4
b. α = 0,4 Period Demand Forecast Forecast error
(t) (Rt) (Pt) (et)
1 42
2 40

3 43

4 40

5 41
6 39
7 46
8 44
9 45
10 38
11 40
12

AOT 4200 - Introduction to operations management 61

 Exponential smoothing
Example
α = 0,4
b. α = 0,4 Period Demand Forecast Forecast error
(t) (Rt) (Pt) (et)
1 42
We must start by setting : 2 40 P2 = R1 = 42 e2 = R2 - P2

P2 = R1. e2 = 40-42
= -2
3 43 P3 = P2 + αe2 e3 = R3 - P3
= 42 + 0,4(-2) e3 = 43- 41,2
= 41,2 = 1,8
4 40 P4 = 41,2 +0,4*(1,8) e4 = 40 - 41,92
= 41,92 = -1,92
5 41 41,152 -0,152
6 39 41,0912 -2,0912
7 46 40,2547 5,7453
8 44 42,5528 1,4472
9 45 43,1317 1,8683
10 38 43,8790 -5,8790
After calculations, we 11 40 41,5274 -1,5274

obtain P12 = 40,92 chairs. 12 40,9164

AOT 4200 - Introduction to operations management 62

 Impact of the smoothing constant α

Low values of α High values of α

• The errors impact is low since they are of • The errors impact is important since they
little importance have are very important
• The forecast is stable since it reacts slowly to • The forecast is reactive since it reacts
changes in demand. promptly to changes in demand.
Exponential smoothing forecast with α = 0.15 Exponential smoothing forecast with α = 0.8

Units

Period
Legend: Forecast Period
Demand AOT 4200 - Introduction to operations 63
Legend: managementDemand Forecast
 Impact of the smoothing constant α

The choice of the constant in practice:

• It is therefore necessary to choose a compromise value between the advantages of
smoothing random variations and the advantages of responsiveness to real changes.
• The most commonly used values vary between α = 0,05 et α = 0,5.

AOT 4200 - Introduction to operations management 64

 Realities of average-based methods

• The application of the formulas must be consistent with the presence or

absence of seasonality;
• Methods based on averages are generally less accurate in the presence of
a trend;
• Methods based on means require the calibration of several parameters;
• Parameter calibration has an influence on the errors made.

AOT 4200 - Introduction to operations management 65

 Average based methods
Graphical illustrations
300

250

Demand
200

Moving average

150
Weighted moving average
(40%-30%-20%-10%)

100 Exponential smoothing 0,1

Exponential smoothing 0,7

50

0
0 2 4 6 8 10 12

AOT 4200 - Introduction to operations management 66

 Average based methods
Graphical illustrations
160

140

Demand
120

Moving average

100
Weighted moving average
(40%-30%-20%-10%)

80 Exponential smoothing 0,1

Exponential smoothing 0,7

60

40
0 2 4 6 8 10 12

AOT 4200 - Introduction to operations management 67

DEMAND FORECASTING
Error measures

AOT 4200 - Introduction to operations management 68

 Forecast error measures

• Measures the difference between forecast and actual demand

deviation = demand - forecast

• Four types of forecast error measurements :

- Mean deviation (MD)
- Mean absolute deviation (MAD)
- Mean squared error (MSE)
- Mean absolute percent forecast error (MAPE)

AOT 4200 - Introduction to operations management 69

 Forecast error measures
Mean deviation (MD)
• Represents the overall deviation of each forecast from its actual demand.

σ𝑛𝑡=1 𝑒𝑡
𝑀𝐷 =
𝑛

Mean absolute deviation (MAD)

• Takes into account the magnitude of the difference between actual demand and its
forecast without taking into account the negative or positive nature of the error.

et = Rt − Pt Forecast error
σ𝑛𝑡=1 𝑒𝑡
𝑀𝐴𝐷 = n = Number of periods
𝑛

AOT 4200 - Introduction to operations management 70

 Forecast error measures
Mean squared error (MSE)
• Gives the extent of the errors made.

σ𝑛𝑡=1 𝑒𝑡2
𝑀𝑆𝐸 =
𝑛

Mean absolute percent forecast error (MAPE)

• Allows you to assess the proportion of the error made.

𝑒𝑡
σ𝑛𝑡=1
𝑅𝑡
𝑀𝐴𝑃𝐸 = × 100
𝑛

AOT 4200 - Introduction to operations management 71

 Forecast error measures
Example
The table below presents the demand for one product per period as well as the forecast
obtained for the same period. Calculate the four error measures.

Period (t) Demand (Rt) Forecast (Pt)

1 217 215
2 213 216
3 216 215
4 210 214
5 213 211
6 219 214
7 216 217
8 212 216

AOT 4200 - Introduction to operations management 72

AOT 4200 - Introduction to operations management 73
 Forecast error measures
Example

Absolute
Error (𝒆𝒕 ) Absolute Squared percent error
Period (t) Demand (𝑹𝒕 ) Forecast (𝑷𝒕 )
et = Rt – Pt error( 𝒆𝒕 ) error (𝒆𝟐𝒕 )
( 𝒆𝒕 ൗ𝑹𝒕)

1 217 215
2 213 216
3 216 215
4 210 214
5 213 211
6 219 214
7 216 217
8 212 216
Average

AOT 4200 - Introduction to operations management 74

 Forecast error measures
Example

Absolute
Error (𝒆𝒕 ) Absolute Squared percent error
Period (t) Demand (𝑹𝒕 ) Forecast (𝑷𝒕 )
et = Rt – Pt error( 𝒆𝒕 ) error (𝒆𝟐𝒕 )
( 𝒆𝒕 ൗ𝑹𝒕)

1 217 215 2
2 213 216 -3
3 216 215 1
4 210 214 -4
5 213 211 2
6 219 214 5
7 216 217 -1
8 212 216 -4
Average

AOT 4200 - Introduction to operations management 75

 Forecast error measures
Example – Mean Deviation (MD)

Absolute
Error (𝒆𝒕 ) Absolute Squared percent error
Period (t) Demand (𝑹𝒕 ) Forecast (𝑷𝒕 )
et = Rt – Pt error( 𝒆𝒕 ) error (𝒆𝟐𝒕 )
( 𝒆𝒕 ൗ𝑹𝒕)

1 217 215 2
2 213 216 -3
3 216 215 1
4 210 214 -4
5 213 211 2
6 219 214 5
7 216 217 -1
8 212 216 -4
Average -0,25

AOT 4200 - Introduction to operations management 76

 Forecast error measures
Example – Mean Absolute Deviation (MAD)

Absolute
Error (𝒆𝒕 ) Absolute Squared percent error
Period (t) Demand (𝑹𝒕 ) Forecast (𝑷𝒕 )
et = Rt – Pt error( 𝒆𝒕 ) error (𝒆𝟐𝒕 )
( 𝒆𝒕 ൗ𝑹𝒕)

1 217 215 2 2
2 213 216 -3 3
3 216 215 1 1
4 210 214 -4 4
5 213 211 2 2
6 219 214 5 5
7 216 217 -1 1
8 212 216 -4 4
Average -0,25 2,75

AOT 4200 - Introduction to operations management 77

 Forecast error measures
Example - Mean squared error (MSE)

Absolute
Error (𝒆𝒕 ) Absolute Squared percent error
Period (t) Demand (𝑹𝒕 ) Forecast (𝑷𝒕 )
et = Rt – Pt error( 𝒆𝒕 ) error (𝒆𝟐𝒕 )
( 𝒆𝒕 ൗ𝑹𝒕)

1 217 215 2 2 4
2 213 216 -3 3 9
3 216 215 1 1 1
4 210 214 -4 4 16
5 213 211 2 2 4
6 219 214 5 5 25
7 216 217 -1 1 1
8 212 216 -4 4 16
Average -0,25 2,75 9,5

AOT 4200 - Introduction to operations management 78

 Forecast error measures
Example - Mean absolute percent forecast error (MAPE)

Absolute
Error (𝒆𝒕 ) Absolute Squared percent error
Period (t) Demand (𝑹𝒕 ) Forecast (𝑷𝒕 )
et = Rt – Pt error( 𝒆𝒕 ) error (𝒆𝟐𝒕 )
( 𝒆𝒕 ൗ𝑹𝒕)

1 217 215 2 2 4 0,0092

2 213 216 -3 3 9 0,0141
3 216 215 1 1 1 0,0046
4 210 214 -4 4 16 0,0190
5 213 211 2 2 4 0,0094
6 219 214 5 5 25 0,0228
7 216 217 -1 1 1 0,0046
8 212 216 -4 4 16 0,0189
Average -0,25 2,75 9,5 0,0128
Forecast accuracy = 100% - 1,28% = 98.72
AOT 4200 - Introduction to operations management 79
 Forecast error measures
Example - Interpretation
• The negative MD means that the forecast is on
average too high relative to actual demand. The
forecast exceeds the actual demand by an
average of 0.25 units. MD -0,25
• The MAD reports without regard to the sign of
the error that there is a difference of 2.75 units MAD 2,75
on average between the forecast and the actual
demand. MSE 9,5

• The MSE indicates that the magnitude of the MAPE 1,28 %

error between the forecast and the actual
demand is 9.5 squared units.
• The MAPE gives an average error of 1.28%. The
forecast accuracy is :
(100 % - 1,28%) = 98,72%.
AOT 4200 - Introduction to operations management 80
 Forecast error measurements

• MD and MAD weight all errors equally.

• MSE adjusts the errors according to their quadratic values. High errors have a higher
weight.

• MAPE highlights the proportion of the error made.

• For the same time series, the error measurements allow to choose the most accurate
forecasting method.

• Error measures can lead to conflicting results. The manager will need to select the
forecasting method that meets his business objectives.
AOT 4200 - Introduction to operations management 81
 Exercise

We use the same data as before. Period (t) Demand (Rt)

1. Perform demand forecasts for period 4 to 8 1 217
using: 2 213
1. Moving average of order 3 3 216
2. Exponential smoothing with 𝛼 = 0,3 4 210
2. Calculate the different error metrics for each 5 213
method. 6 219
3. Suggest choosing a method. 7 216
8 212

AOT 4200 - Introduction to operations management 82

AOT 4200 - Introduction to operations management 83
 Exercise
1 - Demand forecasts – Moving average

Moving average of Exponential

Period (t) Demand (Rt)
order 3 smoothing 𝜶 = 𝟎, 𝟑
1 217
2 213
3 216
4 210
5 213
6 219
7 216
8 212

AOT 4200 - Introduction to operations management 85

 Exercise
1 - Demand forecasts – Moving average

Moving average of Exponential

Period (t) Demand (Rt)
order 3 smoothing 𝜶 = 𝟎, 𝟑
1 217
2 213
3 216
4 210 215,33
5 213
6 219
7 216
8 212

AOT 4200 - Introduction to operations management 86

 Exercise
1 - Demand forecasts – Moving average

Moving average of Exponential

Period (t) Demand (Rt)
order 3 smoothing 𝜶 = 𝟎, 𝟑
1 217
2 213
3 216
4 210 215,33
5 213 213
6 219
7 216
8 212

AOT 4200 - Introduction to operations management 87

 Exercise
1 - Demand forecasts – Moving average

Moving average of Exponential

Period (t) Demand (Rt)
order 3 smoothing 𝜶 = 𝟎, 𝟑
1 217
2 213
3 216
4 210 215,33
5 213 213
6 219 213
7 216
8 212

AOT 4200 - Introduction to operations management 88

 Exercise
1 - Demand forecasts – Moving average

Moving average of Exponential

Period (t) Demand (Rt)
order 3 smoothing 𝜶 = 𝟎, 𝟑
1 217
2 213
3 216
4 210 215,33
5 213 213
6 219 213
7 216 214
8 212 216

AOT 4200 - Introduction to operations management 89

 Exercise
2 – Errors moving average

Absolute
Error (𝒆𝒕 ) Absolute Squared percent error
Period (t) Demand (𝑹𝒕 ) Forecast (𝑷𝒕 )
et = Rt – Pt error ( 𝒆𝒕 ) error (𝒆𝟐𝒕 )
( 𝒆𝒕 ൗ𝑹𝒕)

1 217
2 213
3 216
4 210 215,33
5 213 213
6 219 213
7 216 214
8 212 216
Average

AOT 4200 - Introduction to operations management 90

CODEX – LA GESTION DES OPÉRATIONS P.17 - P.19
 Exercise
2 – Errors moving average

Absolute
Error (𝒆𝒕 ) Absolute Squared percent error
Period (t) Demand (𝑹𝒕 ) Forecast (𝑷𝒕 )
et = Rt – Pt error ( 𝒆𝒕 ) error (𝒆𝟐𝒕 )
( 𝒆𝒕 ൗ𝑹𝒕)

1 217
2 213
3 216
4 210 215,33 -5,33
5 213 213 0
6 219 213 6
7 216 214 2
8 212 216 -4
Average -0,27

AOT 4200 - Introduction to operations management 91

CODEX – LA GESTION DES OPÉRATIONS P.17 - P.19
 Exercise
2 – Errors moving average

Absolute
Error (𝒆𝒕 ) Absolute Squared percent error
Period (t) Demand (𝑹𝒕 ) Forecast (𝑷𝒕 )
et = Rt – Pt error ( 𝒆𝒕 ) error (𝒆𝟐𝒕 )
( 𝒆𝒕 ൗ𝑹𝒕)

1 217
2 213
3 216
4 210 215,33 -5,33 5,33 28,44 0,02540
5 213 213 0 0 0 0
6 219 213 6 6 36 0,02740
7 216 214 2 2 4 0,0093
8 212 216 -4 4 16 0,0189
Average -0,27 3,47 16,89 0,02

AOT 4200 - Introduction to operations management 92

CODEX – LA GESTION DES OPÉRATIONS P.17 - P.19
AOT 4200 - Introduction to operations management 93
 Exercise
1 - Demand forecasts – Exponential smoothing with 𝛼=0,3

Moving average of Exponential

Period (t) Demand (Rt)
order 3 smoothing 𝜶 = 𝟎, 𝟑
1 217
2 213
3 216
4 210
5 213
6 219
7 216
8 212

AOT 4200 - Introduction to operations management 94

 Exercise
1 - Demand forecasts - Exponential smoothing with 𝛼=0,3

Exponential
Moving average Error (𝒆𝒕 )
Period (t) Demand (Rt) smoothing 𝜶 =
of order 3 et = Rt – Pt
𝟎, 𝟑
1 217
2 213
3 216
4 210
5 213
6 219
7 216
8 212

AOT 4200 - Introduction to operations management 95

 Exercise
1 - Demand forecasts - Exponential smoothing with 𝛼=0,3

Exponential
Moving average Error (𝒆𝒕 )
Period (t) Demand (Rt) smoothing 𝜶 =
of order 3 et = Rt – Pt
𝟎, 𝟑
1 217
2 213
3 216
4 210
5 213
6 219
7 216
8 212

AOT 4200 - Introduction to operations management 96

 Exercise
1 - Demand forecasts - Exponential smoothing with 𝛼=0,3

Exponential
Moving average Error (𝒆𝒕 )
Period (t) Demand (Rt) smoothing 𝜶 =
of order 3 et = Rt – Pt
𝟎, 𝟑
1 217
2 213 217
3 216
4 210
5 213
6 219
7 216
8 212

AOT 4200 - Introduction to operations management 97

 Exercise
1 - Demand forecasts - Exponential smoothing with 𝛼=0,3

231 - 217

Exponential
Moving average Error (𝒆𝒕 )
Period (t) Demand (Rt) smoothing 𝜶 =
of order 3 et = Rt – Pt
𝟎, 𝟑
1 217
2 213 217 -4
3 216
4 210
5 213
6 219
7 216
8 212

AOT 4200 - Introduction to operations management 98

 Exercise
1 - Demand forecasts - Exponential smoothing with 𝛼=0,3

Exponential
Moving average Error (𝒆𝒕 )
Period (t) Demand (Rt) smoothing 𝜶 =
of order 3 et = Rt – Pt
𝟎, 𝟑
1 217
2 213 217 -4
3 216 215.8
4 210
5 213
6 219
7 216
8 212
217 + 0.3 * (-4)

AOT 4200 - Introduction to operations management 99

 Exercise
1 - Demand forecasts - Exponential smoothing with 𝛼=0,3

216 – 215.8

Exponential
Moving average Error (𝒆𝒕 )
Period (t) Demand (Rt) smoothing 𝜶 =
of order 3 et = Rt – Pt
𝟎, 𝟑
1 217
2 213 217 -4
3 216 215,8 0.2
4 210
5 213
6 219
7 216
8 212

AOT 4200 - Introduction to operations management 100

 Exercise
1 - Demand forecasts - Exponential smoothing with 𝛼=0,3

Exponential
Moving average Error (𝒆𝒕 )
Period (t) Demand (Rt) smoothing 𝜶 =
of order 3 et = Rt – Pt
𝟎, 𝟑
1 217
2 213 217 -4
3 216 215,8 0.2
4 210 215,86
5 213
6 219
7 216 215.8 + 0.3 * (0.2)

8 212

AOT 4200 - Introduction to operations management 101

 Exercise
1 - Demand forecasts - Exponential smoothing with 𝛼=0,3
210 – 215.86

Exponential
Moving average Error (𝒆𝒕 )
Period (t) Demand (Rt) smoothing 𝜶 =
of order 3 et = Rt – Pt
𝟎, 𝟑
1 217
2 213 217 -4
3 216 215,8 0.2
4 210 215,86 -5.86
5 213
6 219
7 216
8 212

AOT 4200 - Introduction to operations management 102

 Exercise
1 - Demand forecasts - Exponential smoothing with 𝛼=0,3

Exponential
Moving average Error (𝒆𝒕 )
Period (t) Demand (Rt) smoothing 𝜶 =
of order 3 et = Rt – Pt
𝟎, 𝟑
1 217
2 213 217 -4
3 216 215,8 0.2
4 210 215,86 -5.86
5 213 214,1
6 219
7 216
8 212 215.86 + 0.3 * (-5.86)

AOT 4200 - Introduction to operations management 103

 Exercise
1 - Demand forecasts - Exponential smoothing with 𝛼=0,3

Exponential
Moving average Error (𝒆𝒕 )
Period (t) Demand (Rt) smoothing 𝜶 =
of order 3 et = Rt – Pt
𝟎, 𝟑
1 217
2 213 217 -4
3 216 215,8 0.2
4 210 215,86 -5,86
5 213 214,1 -1,10
6 219 213,77 5,23
7 216 215,34 0,66
8 212 215,54 -3,54

AOT 4200 - Introduction to operations management 104

 Exercise
2 – Errors for exponential smoothing

Absolute
Error (𝒆𝒕 ) Absolute Squared percent error
Period (t) Demand (𝑹𝒕 ) Forecast (𝑷𝒕 )
et = Rt – Pt error ( 𝒆𝒕 ) error (𝒆𝟐𝒕 )
( 𝒆𝒕 ൗ𝑹𝒕)

1 217
2 213 217
3 216 215,8
4 210 215,86 -5,86
5 213 214,1 -1,10
6 219 213,77 5,23
7 216 215,34 0,66
8 212 215,54 -3,54
Average -0,92

AOT 4200 - Introduction to operations management 105

CODEX – LA GESTION DES OPÉRATIONS P.17 - P.19
 Exercise
2 – Errors for exponential smoothing

Absolute
Error (𝒆𝒕 ) Absolute Squared percent error
Period (t) Demand (𝑹𝒕 ) Forecast (𝑷𝒕 )
et = Rt – Pt error ( 𝒆𝒕 ) error (𝒆𝟐𝒕 )
( 𝒆𝒕 ൗ𝑹𝒕)

1 217
2 213 217
3 216 215,8
4 210 215,86 -5,86 5,86 34,34 0,0279
5 213 214,1 -1,10 1,10 1,21 0,0052
6 219 213,77 5,23 5,23 27,34 0,0239
7 216 215,34 0,66 0,66 0,44 0,0031
8 212 215,54 -3,54 3,54 12,52 0,0167
Average -0,92 3,28 15,17 0,02

AOT 4200 - Introduction to operations management 106

CODEX – LA GESTION DES OPÉRATIONS P.17 - P.19
 Exercise
3 – Choice of method

Important: we compare errors over the same periods! Forecast is on average

too high relative to
actual demand
Exponential
Moving average
Error smoothing with
order 3
𝜶 = 𝟎, 𝟑
Difference on average
between the forecast
MD -0,27 -0,92 and the actual demand

MAD 3,47 3,28

Magnitude of the error
between the forecast
MSE 16,89 15,17 and the actual demand

MAPE 2% 2%
Average error (2%).
Forecast accuracy =
Use of exponential smoothing 98%

AOT 4200 - Introduction to operations management 107

 Keywords

• Qualitative methods • Weighted moving average

• Quantitative methods • Weighting
• Tendency • Exponential smoothing
• Seasonality • Smoothing constant
• Cycle • Error measurements
• Time series • The average error
• Naïve method • The absolute mean error
• Simple average • Mean quadratic error
• Moving average • The average absolute error
percentage

AOT 4200 - Introduction to operations management 108

 Next lesson

• Inventory Management - A Strategic Asset

Book pages 330 to 380

AOT 4200 - Introduction to operations management 109