iAcademy

Demand Management

Operations and Supply Chain Management

Lecture 8

Components of Demand Forecasting

Forecasting Techniques and Accuracy
This lecture and its associated materials have been produced by Dr. Pichawadee Kittipanya-ngam (PhD, Cambridge) of
iAcademy for the purposes of lecturing on the above described subject and the material should be viewed in this
context. The work does not constitute professional advice and no warranties are made regarding the information
presented. The Author and iAcademy do not accept any liability for the consequences of any action taken as a result of
the work or any recommendations made or inferred. Permission to use any of these materials must be first granted by
iAcademy.
iAcademy
 iAcademy

Agenda
• Review of Week 7 Lecture
• Week 8 Lecture Coverage
– Components of Demand Forecasting
– Forecasting Techniques
– Forecasting Accuracy

iAcademy
 iAcademy

Summary: Week 7 Lecture

• Supply Chain is a chain of operations with 5 key
components
• There are 2 product types in Supply Chain
Strategy: functional and innovative products
• Functional product’s supply chain focuses on
COST, resulting in LEAN supply chain
• Innovative product’s supply chain focuses on
SPEED, resulting in RESPONSIVE supply chain

iAcademy
 iAcademy

Any questions?

iAcademy
 iAcademy

Let’s get started with Lecture 8

Demand Management
iAcademy
 iAcademy

Lecture 8 Coverage

• Components of Demand Forecasting

• Forecasting Technique
• Forecasting Accuracy

iAcademy
 iAcademy

Why is “forecasting” important?

• Imagine you are planning to go on a holiday in
the USA. What do you need to know before
packing your clothes? ….. Weather forecast in the USA?
– Temperature in the USA? Whether it is hot or cold?
What kind of jackets do you need to prepare?
– Would it be raining? Do you need an umbrella or a
raincoat?

Forecasting allows you to

better plan for the future

iAcademy
 iAcademy

What if the forecast is incorrect?

• What if the weather forecast you checked was
SUNNY as shown below
• But once you arrived, it is like in the VDO below!
Let’s watch a video of a
weather forecast. Play the
video.
- Reporters weather the storm

Poor forecast could cause a

disaster!!
iAcademy
 iAcademy

1. Components of
Demand Forecasting

iAcademy
 iAcademy

Introduction to Demand Forecasting

• What is “demand forecasting”?
– Demand forecasting is the activity of estimating the
quantity of a product or service that consumers will
purchase in the future.
• For example?
– Demand forecasting of the car purchases in India in
2020
– Demand forecasting of the coffee consumption in
China in 2020

iAcademy
 iAcademy

Example of Demand Forecasting

• How many Big Mac, do you think, are sold at a
McDonald store a day?
– Operators at McDonalds stores need to order raw
materials to meet the demand volume daily e.g.
• How much each store should order beef patties?
• How much each store should order packaging?
• How much each store should order buns?
• Etc….

iAcademy
 iAcademy

Example of Demand Forecasting

• What if the sale forecast of McDonalds stores
are incorrect?
– Suppliers suffer from overproduction?
– McDonalds suffer from
• obsolete inventory in case of over-forecast or
• loss of sales in case of under-forecast.

iAcademy
 iAcademy

Example of Demand Forecasting

• How the sale forecast of the Big Mac impact on
its suppliers’ operations?

• Let’s watch a video of a beef patty supplier of

McDonalds. Play the video.
– Inside the Factory Where McDonalds' Meat Comes From

The supplier needs the demand forecast of

Big Mac to plan its daily operations!

iAcademy
 iAcademy

Introduction to Demand Forecasting

• The demand forecast helps firms to better plan
their operations including the followings
– Production planning and scheduling: without
knowing the demand, firms cannot plan their
production schedule.
• For example, they don’t know what to produce, when to
produce, and at what amount

iAcademy
 iAcademy

Introduction to Demand Forecasting

• The demand forecast helps firms to better plan
their operations including the followings
– Capacity & resource allocation: without knowing the
demand, firms cannot allocate their resources
properly.
• For example, how many workers they need and for how
long (a shift? a day?) to produce the products.

iAcademy
 iAcademy

Introduction to Demand Forecasting

• The demand forecast helps firms to better plan
their operations
– Inventory policy: without knowing the demand,
firms cannot plan when they should order its raw
materials. For example,
• the amount of finished goods firms should produce to
meet the expected demand
• the amount and the time firms should order their raw
materials to be used for the production of the products
– Etc…

iAcademy
 iAcademy

What if the forecast is incorrect?

• What would happen in case of a poor forecast?
– Excess inventory (sunk costs) due to the over-
forecasted demand. The company’s cash would get
stuck in its finished goods inventory.
– Shortage of inventory due to the under-forecasted
demand. The company may lose the potential
customers and revenue plus an opportunity cost.
– Overtime production costs due to the over-
forecasted demand. The company overpays its
workers and its additional resources that eventually
they do not need at all.
iAcademy
 iAcademy

Demand Forecast
• Let’s watch a video clip on introduction to
demand forecast. Play the video.
– Introduction to demand forecast

Demand Forecast
can be wrong and
are often wrong!

iAcademy
 iAcademy

How to achieve a good forecast?

• Unfortunately, just like weather forecast,
demand forecasts are inaccurate more often
that they are correct!!!!
• So, what to do then? Do we still need to
forecast or not?
– yes, we still need the forecast
• The goal is not to have a perfect forecast, but
rather to have a reasonable forecast with
acceptable errors that helps us best plan our
operations
iAcademy
 iAcademy

These are examples of the real

demand1-day
in the market
1-month

3 months

1-year

Verma & Boyer (2010)

iAcademy
 iAcademy

To achieve a good forecast, let’s

begin with the demand
components?

iAcademy
 iAcademy

Components of Demand
• There are six components to create a demand
in the market
1. Average demand by regular customers
2. Trend
3. Seasonal demand
4. Cyclical demand
5. Auto-correlation demand
6. Random variation demand

iAcademy
 iAcademy

1. Average Demand
• The average demand is created by regular
customers, which is highly predictable.
• From the picture below, the average demand is
at 120 units per month.

iAcademy
 iAcademy

2. Trend Demand
• Trend Demand only happens temporarily,
depending on the specific time frame. It could
be disappeared after a period of time.
– For example, the trend of vintage fashion clothes is
growing this year, compared to none last year.

From the picture,

trend demand
increases by 2.5
units per month

iAcademy
 iAcademy

2. Trend Demand
• Trend demand occurs when demand is
increasing or decreasing over time as a result of
– word of mouth (or blogs),
– promotional activities, for example, discounts
– advertising, or
– changes in the population.

iAcademy
 iAcademy

3. Seasonal Demand
• Seasonal demand is usually found in seasonal
products, for example,
– Ice-cream demand increases in summer
– Overcoat demand increases in winter

From the picture,

seasonal demand
increases only during
the summer months
(May-September)

iAcademy
 iAcademy

4. Cyclical Demand
• Cyclical demand is similar to seasonal demand
but it has a much longer time period and are
often harder to identify
– For example, fashion runs in a cyclical pattern over a
number of years or decades.
• Hot pink may be “in fashion” for a year or two, but then
out of fashion for a number of years until a designer
brings it to the forefront again.
• Cyclical factors for demand include things such
as politics, economic conditions, war, and socio-
cultural influences.
iAcademy
 iAcademy

5. Auto-Correlation Demand
• Auto-Correlation demand occurs when the
value for one data point is highly correlated
with the past values. For example,
– When waiting in a long line, the time for the
fifteenth person in line is highly correlated with (an
guaranteed to be longer than) the time for the tenth
person in line.
– The demand of iPhone case is highly correlated with
the demand of iPhone itself

iAcademy
 iAcademy

6. Random Variation Demand

• The random demand can be seen in the small
hills and valleys of the demand curve.
• Random variation is caused by chance events –
after all the other components of demand
(average, trend, seasonal, cyclical, and auto-
correlation demand) have been accounted for.
The remainders are random variation, caused
by chance.

iAcademy
 iAcademy

Any questions?

iAcademy
 iAcademy

2. Forecasting
Techniques – Time
Series Methods

iAcademy
 iAcademy

Forecasting Techniques
• There are two basic forecasting techniques:
quantitative and qualitative techniques.
– Quantitative methods: rely on the existing data for
demand forecast and use mathematical formulas of
varying complexity to accommodate different types
of demand
• Time-series analysis: utilises past demand data to predict
future demand
• Causal relationship: identify a connection between two
factors e.g. advertisement on sales
– Qualitative methods: based on subjective factors.
Tend to use for new product introduction
iAcademy
 iAcademy

Forecasting Techniques
• This lecture only focuses on the key forecasting
technique that is widely used, which is “Time
Series Analysis”
• Time-Series Analysis is based on historical data
and the assumption that past patterns will
continue in the future.
– The goal is to identify the underlying patterns of
demand and develop model to predict these
patterns in the future.

iAcademy
 iAcademy

Time-Series Analysis
• There are five basic time-series techniques as
follow:
1. Naive forecasting
2. Moving average forecasting
3. Exponential smoothing forecasting
4. Trend-adjusted exponential smoothing
5. Seasonal pattern forecasting
• However, in this lecture, only the first 2
techniques, which are widely used, will be
discussed.
iAcademy
 iAcademy

1. Naive Forecasting
• A naive forecast assumes that the demand for
the current period equals to the forecast for the
next period
– If demand for Friday at a coffee shop is for 142
lattes, the forecast for Saturday is 142 lattes.
However, if the actual demand for Saturday is 150
lattes, the forecast for Sunday will be 150 lattes.
• The naive forecast is very simple and low cost to
use. It works best when demand, trends, and
seasonal patterns are stable and there is
relatively little random variation.
iAcademy
 iAcademy

2. Moving Average Forecasting

• There are two basic techniques of moving
average forecasting
1. Moving average forecast
2. Weighted moving average forecast
• These techniques include 2 components of
demand
– An average demand
– Random variation
• There may be or may not be a trend, seasonal,
cyclical, or autocorrelation component.
iAcademy
 iAcademy

2.1 Moving Averages (MA)

• MA is a technique for estimating the average of
a demand series and filtering out the effects of
random variation
• Its formula is shown below

Verma & Boyer (2010)

iAcademy
 iAcademy

2.1 Moving Averages (MA)

• Example 1
– Please calculate the 4-day moving average demand
forecast of Day 5 using the actual demand data in
the table below

D1
D2
D3
D4

– Given demand forecast of Day 5 = F5 = F4+1 and

n = t = 4 periods (D1, D2, D3, D4)
iAcademy
 iAcademy

2.1 Moving Averages (MA)

• Find F5
F5 = F4+1
= [D1 + D2 +D3 + D4]/4
= [228+228+225+188]/4
= 217.25

• This technique can use as many periods of past

demand as desired. Using more periods result in
a more stable forecasts, while using fewer
periods results in a more responsive forecast.
iAcademy
 iAcademy

2.1 Moving Averages (MA)

• For example, 2-day MA would jump around more often
when demand changes and it reacts more quickly to
the changes in demand than that of 4-day MA as
shown below

Verma & Boyer (2010)

iAcademy
 iAcademy

2.1 Moving Averages (MA)

• What to be careful of MA
– When the demand is truly random, the longer MA is
less likely to overreact.
– If the demand fluctuation turns out to be a trend,
the shorter MA will react more quickly.
• Disadvantage
– This technique places equal importance/weight on
each data item whereas in reality, the recent data
would be more accurate than the old ones.

iAcademy
 iAcademy

2.2 Weighted Moving Average (WMA)

• It is a technique that allows periods to have
different weights, with the total weight
equaling 1.0
– From the example 1, a 3-period weighted moving
average might assign the most recent period a
weight of 0.6, the 2nd most recent period 0.25, and
the 3rd most recent period 0.15. Thus, the forecast
period t+1 would be
Ft+1 = 0.6Dt + 0.25Dt-1 + 0.15Dt-2
F4 = 0.6D3 + 0.25D2 + 0.15D1
= (0.6)225 + (0.25)228 + 0.15(228) = 226.2
iAcademy
 iAcademy

2.2 Weighted Moving Average (WMA)

• The benefit of a WMA is that it allows a greater
emphasis on the most recent demand than on
past demand. This has the effect of making the
forecast more responsive to changes.
• In fact, seasonal and trend effects can be
captured if high weights are used, but this
method will still lag behind trend.
• If the weights for WMA are equal, then the
forecasts will be the same as for MA.

iAcademy
 iAcademy

Any questions?

iAcademy
 iAcademy

3. Forecasting
Accuracy

iAcademy
 iAcademy

Forecasting Accuracy
• All forecasts are usually wrong to some extent,
so how do we choose the “best” forecast?
– The best forecast is the one with the least forecast
error!!!
• Forecast Error is the difference between the
forecast and the actual demand for a given
period.
– Forecast errors can be classified as either bias errors
or random errors.

iAcademy
 iAcademy

Forecasting Accuracy
• Bias errors occur when the forecast is
consistently over or under the actual demand.
These errors often occur when a key demand
component is neglected. For example,
– When an exponential smoothing model is used to
forecast ice cream sales, which have a distinct
seasonal component.
• Random errors result from unpredictable
factors and do not exhibit a distinct pattern.
These errors are difficult to eliminate.
iAcademy
 iAcademy

Forecast Error Measurement

• Calculating forecast error is fairly
straightforward. It is simply the difference
between the actual demand and the forecast.
• The formula for the errors are shown below.
Single forecast errors

Cumulative sum of Forecast Errors

iAcademy
 iAcademy

Example of Forecast Error Measures

• Example 3
– The table provides Week Demand Forecast
demand for the past 6 1 200 190
weeks and the 2 220 198
corresponding forecast for 3 190 210
4 245 195
total cabinet sales at
5 225 217
Quality Cabinets. Please 6 210 235
calculate the errors and 7 218
CFE for the set of
forecasts.

iAcademy
 iAcademy

Example of Forecast Error Measures

• From the formula
– For week 1; E1 = D1 – F1 = 200 – 190 = 10
– For week 2; E2 = D2 – F2 = 220 – 198 = 22
– For week 3; E3 = D3 – F3 = 190 – 210 = -20
– For week 4; E4 = D4 – F4 = 245 – 195 = 50
– For week 5; E5 = D5 – F5 = 225 – 217 = 8
– For week 6; E6 = D6 – F6 = 210 – 235 = -25

iAcademy
 iAcademy

Example of Forecast Error Measures

• Next, we calculate cumulative forecast errors
(CFE)
• From the formula
Hence, CFE = E1 + E2 + E3 + E4 + E5 + E6
= 10 + 22 + (-20) + 50 + 8 + (-25)
= 45

• Next we will learn about the percentage of

errors
iAcademy
 iAcademy

Any questions?

iAcademy
 iAcademy

Summary: Week 8 Lecture

• Demand forecasting is the activity of estimating
the quantity of a product or service that
consumers will purchase in the future.
• The demand forecast helps firms to better plan
their operations
• There are six components to create a demand
• There are five basic time-series forecasting
techniques
• The best forecast is the one with the least forecast
error!!!
iAcademy
 iAcademy

What to Expect: Week 8 Tutorial

• 1 Activity in this tutorial
– To solve problems of demand forecasting and its
errors using different forecasting techniques.

iAcademy
 iAcademy

iAcademy