SCML 200 - Supply Chain Management & Operations

Chapter 3: Forecasting – Part I

1
Learning Objectives of
chapter 3

2
 Budgeting & Cost Control

The Role of Forecasting

 Forecasting is a vital function and impacts every
significant management decision.
New Products Planning

 Finance and accounting use forecasts as the basis

for budgeting and cost control.

 Marketing relies on forecasts to make key decisions

such as new product planning. Decisions about inventory

 Production uses forecasts to select suppliers,

determine capacity requirements, and to drive
decisions about purchasing, staffing, and
inventory.
3
 Dependent demand
Demand management
 The purpose of demand management is
 to coordinate and control all sources of
demand.
Dependent demand

 Dependent demand:
 the demand for a product or service caused by
the demand for other products or services.
 it is demand that must be met.
Independent demand
 Independent demand:
 the demand for a product or service that
cannot be derived directly from that of other
products.

4
Examples:
Independent demand Dependent demand

Bicycle Wheels

Laptop Micro chips

Pizza Flour
5
Types of Forecasting
 Main types of forecasts
 Qualitative
Mathematical model
 Quantitative: Time series analysis
 It is a quantitative approach
 One can construct time series regarding any
data collected over time:
 demand, temperature, population, etc. Computing
 Time series analysis is based on the idea that
 data relating to past demand can be used to
predict future demand.
 In this course we will focus on time series
for demand.
6
Examples of time series
Treasury bill

Dow Jones Index

7
Examples of time series
e-retail [B2C e-commerce]
is retail through Internet

Retail through physical

[brick&mortar] store

Source: https://www.businessinsider.sg/ecommerce-percent-retail-sales-charts-2018-5/ 8
Examples of time series
Apple stock share price over time Apple stock share

Source: https://bit.ly/2kDlZBf 9
Quantitative Demand Forecasting

Source: https://youtu.be/kKm8noEnl5w 10
Characteristics of time series
[decomposition]
Average
demand for
Trend
a period of
time

Cyclical
Seasonality
elements

Random Autocorrelat
variation ion

11
Example 1
Average
demand for
Trend
a period of
time

Cyclical
Seasonality
elements

Random Autocorrelat
variation ion

Source: https://www.yinglinglow.com/blog/2018/08/09/An-Intro-To-TimeSeries-Decomposition 12
Example 2
Observed Data

Average
demand for Trend Seasonality
a period of
time

Cyclical Random Autocorrela

elements variation tion

Source: https://serialmentor.com/dataviz/visualizing-trends.html 13
Trends
 Identification of trend lines is a common starting point when
developing a forecast
 A trend exists when
 there is a long-term increase or decrease in the data. Trend

 Trend does not have to be linear.

 Common trend types include linear, S-curve, asymptotic, and
exponential
Linear Trend S-curve Trend Asymptotic Trend Exponential Trend

14
Seasonal and cyclical
patterns
A seasonal pattern exists when a time series is
• influenced by seasonal factors
• Examples: the quarter of the year, the month, or day of Seasonality
the week.

• Seasonality is always of a fixed and known period.

• Examples: Eid, Christmas, Winter/summer, lunch time
during a day

Cyclical
A cyclic pattern exists when data exhibit elements
• rises and falls that are not of fixed period.

• The duration of these fluctuations is usually of many 15

Autocorrelation and random
variation
Autocorrelation: Situation in which a time series
data is influenced by its own historical values. Autocorrelat
• Example: Number of measles cases on Monday ion
is correlated to the number of measles cases on
Sunday

Random variation:
Random
• Variability of a process
Variation
• caused by many irregular and erratic
fluctuations or chance factors that, in practical
terms,
• cannot be anticipated, detected, identified or 16
Further Reading

Source: http://www.grroups.com/blog/components-of-time-series-data 17
How to perform
decomposition ?  ANSWER:

 Using data mining

software or code
 Ex:
 Python code

 Specialized
functions
 Ex:
seasonal_decomp
ose()
in Pandas Python
library

Source: https://www.cbcity.de/timeseries-decomposition-in-python-with-statsmodels-and-pandas
18
Example: Time series prediction (using
BigML)

Observed data

Prediction

Model with best performance Selected forecast horizon

Source: http://ertekprojects.com/ftp/cet2018/CET_2018_Dubai_Dr_Gurdal_Ertek.pdf 19
Time Series Analysis
 Using the past to predict the future
Short term – forecasting less than three
months
• Used mainly for tactical decisions

Medium term – forecasting three months to

two years
• Used to develop a strategy which will be implemented
over the next six to eighteen months (e.g. meeting
demand)
Long term – forecasting greater than two
years
• Useful for detecting general trends and identifying
major turning points
20
Forecasting Method Selection Guide
Forecasting Amount of Forecast
Data Pattern
Method Historical Data Horizon
6 to 12 months; Stationary
Simple moving
weekly data are often (i.e. no trend or Short
average
used seasonality)
Weighted moving
average & simple 5 to 10 observations
Stationary Short
exponential needed to start
smoothing
5 to 10 observations
Exponential Stationary or
needed to start Short
smoothing with trend data with trend

Stationary,
Short to
Linear regression 10 to 20 observations trend, and
Medium
seasonality 21
Terminology and notation
 Terminology and notation
t period index

 realized value (actual value) in a period t

 Example:
 On Tuesday, the temperature forecast for Wednesday
is 25C.
 On Wednesday, the temperature turns out to be 27C.

 In this case, F
wed=25C, but Awed=27C.

22
Forecasting methods for
stationary data
 A data is said stationary when demand is
not growing or declining rapidly and there is
no seasonality or trend
 Most commonly used methods with
stationary data are:
 Simple moving average
 Weighted moving average
 Exponential smoothing Stationary Demand

demand

time
23
Simple Moving Average
 Forecast in future period (t) is the average of
the realized values in the past n periods (periods
)

 Notation
 : period index
 forecasted value (predicted value) for period t

 realized value (actual value) in period t

 : number of periods in the past to be averaged

24
Simple Moving Average –
Example
Demand
Week (realized)
1 800
2 1400
3 1000

 3 weeks moving average :

𝐴1 + 𝐴2 + 𝐴3
𝐹 4=
3
 The demand forecast in week 4 (=4) is the
average of the realized demand for the past 3
weeks
𝐴 (weeks
+ 𝐴 + 3,2,1)
𝐴
1 2 3
𝐹 4=
3
25
Simple Moving Average –
Example
 Suppose that in week 4, it turns out that the real demand was 1500,
so =1500.
Demand Forecasted
Week (realized) Demand
1 800
2 1400
3 1000
4 1500 1067

 In week 4, we forecast the demand for week 5, by considering

the past 3 weeks (weeks 4,3,2) :
𝐴 2+ 𝐴 3+ 𝐴 4
𝐹 5=
3

 When we are in week 5, the real demand turns out to be 1500, hence
=1500. 26
Simple Moving Average –
Example
 Continuing with simple moving averages with n=3, and
registering the demand as we advance in time, we obtain:
Demand
Week (realized) Forecasted
1 800
2 1400
3 1000
4 1500 1067
5 1500 1300
6 1300 1333
7 1800 1433
8 1700 1533
9 1300 1600
10 1700 1600
11 1700 1567
12 1500 1567
13 2300 1633
14 2300 1833 27
Simple Moving Average –
Exercise
Assume: We start with the data for the first 9 weeks and forecast
the demand
for week 10 with a simple moving average with 9 periods (weeks)
 Check whether you obtain theForecasted
values in the table below (rounded to
Week
the nearest Demand (realized)
integer) demand
1 800
2 1400
3 1000
4 1500
5 1500
6 1300
7 1800
8 1700
9 1300
10 1700 1367
11 1700 1467
12 1500 1500
13 2300 1556
14 2300 1644 28
Selecting the length of the period
 Longer periods • 3-week forecast follows better • Choosing short periods alway
 provide smoother forecasts
the changes in demand, is not advisable:
 Shorter periods • 9-week forecast is smoother. • One may react too strongly
 react to trends more quickly
to temporary variations.

29
Weighted Moving Average
 Disadvantage of simple moving
average:
 All periods are equally important

 A solution: Weighted moving

average:
 Allows giving higher weights to
more important periods.
 Usually, the most recent periods
have higher weights.
30
Weighted Moving Average
Notation:
forecasted value for period
realized value in period
: number of periods in the past

𝑤𝑛
that are considered in the forecast

is the weight for ,

is the weight for

Remark:
weights

31
Weighted Moving Average
Example:

5-weeks weighted moving average with weights:

If =6 then

• the two most recent weeks have higher weights of 0.25

• week -5 (week 1) has the lowest weight of 0.1
32
Selecting Weights
 Experience and/or trial-and-error are
the simplest approaches.

 The recent past is often the best indicator

of the future, so weights are generally
higher for more recent data.

33
34
35
Homework Part 1

 8,
 11,
 13 (only a and b),
 15 (only a, b and c)

 Pages 82-84 of the textbook

36
Thank you.
Your
Questions?

37