You are on page 1of 28

Demand Management and Forecasting

Ronald S. Lau, Ph.D.


HKUST – ISOM
What is the role of forecast in demand management?

2
Demand management
 Demand management is a process for optimizing the customer demand
with available capacity to maximize a company’s profit
 Revenue: How much to sell and at what price?
 Cost: How much capacity to maintain and at what cost?

 Common approaches and tactics


 Take a passive role to forecast and respond to customer demand
 Take an active role to influence or prioritize customer demand

Influence/
Forecast prioritize
demand demand
Time for
Now actual sales

3
Evolution of demand management

No forecast Forecast Share demand Synchronize Demand


demand using plan to supply demand and collaboration
historical sales organizations supply plan with supply
data with an chain partners
integrated
management
process

4
Forecast-driven planning activities

Collect Plot Select Develop


data data model forecast

Check Adjust Monitor


accuracy forecast forecast

Sales and Operations Planning

Manufacturing Services
Master Workforce Customer
scheduling scheduling scheduling

Master Order
planning scheduling

Source: Wisner and Stanley (2008)


5
Examples of business forecast applications

Accounting Cost/profit estimates and budgets

Finance Cash flow and funding

Human resources Hiring, training, and firing employees

Marketing Pricing, promotion, and strategy

Operations and logistics Scheduling and capacity planning

6
Principles of demand forecasting
 No forecast is perfect
I see that you will…

 Assumes causal relationship to continue

 Forecasts are more accurate for groups


(vs. individuals)

 Forecast accuracy decreases as time


horizon increases

 Good forecasts do not always require


the use of complex forecasting models

7
How to evaluate the accuracy of forecasts?

8
Measuring forecast accuracy

 Forecast error  Mean absolute percent error (MAPE)


 Error = Actual – Forecast  MAPE = ∑|Percent Error| / n
where PE = (Actual – Forecast) / Actual
 Mean absolute deviation (MAD)
 MAD = ∑|Actual – Forecast| / n  Tracking signal
 TS = ∑(Actual – Forecast) / MAD
 Mean squared error (MSE)
 MSE = ∑(Actual – Forecast)2 / n

9
Selecting forecasting model: Example

You have the following sales forecasts ($M) using two different methods.
Which method should you use?

Actual Method 1 Method 2


Month Sales Forecast Forecast

1 100 60 100
2 100 130 100
3 200 200 150
4 200 270 200
5 400 340 250

10
Method 1: Tracking signal

Month Actual Forecast Error ∑|Error| MAD ∑(Error) TS

1 100 60

2 100 130

3 200 200

4 200 270

5 400 340

Error = Actual – Forecast MAD = ∑|Error| / n TS = ∑(Error) / MAD

11
Method 1: Tracking signal

Month Actual Forecast Error ∑|Error| MAD ∑(Error) TS

1 100 60 40

2 100 130 -30

3 200 200 0

4 200 270 -70

5 400 340 60

Error = Actual – Forecast MAD = ∑|Error| / n TS = ∑(Error) / MAD

12
Method 1: Tracking signal

Month Actual Forecast Error ∑|Error| MAD ∑(Error) TS

1 100 60 40 40 40

2 100 130 -30 70 10

3 200 200 0 70 10

4 200 270 -70 140 -60

5 400 340 60 200 0

Error = Actual – Forecast MAD = ∑|Error| / n TS = ∑(Error) / MAD

13
Method 1: Tracking signal

Month Actual Forecast Error ∑|Error| MAD ∑(Error) TS

1 100 60 40 40 40 40

2 100 130 -30 70 35 10

3 200 200 0 70 23 10

4 200 270 -70 140 35 -60

5 400 340 60 200 40 0

Error = Actual – Forecast MAD = ∑|Error| / n TS = ∑(Error) / MAD

14
Method 1: Tracking signal

Month Actual Forecast Error ∑|Error| MAD ∑(Error) TS

1 100 60 40 40 40 40 1.00

2 100 130 -30 70 35 10 0.29

3 200 200 0 70 23 10 0.43

4 200 270 -70 140 35 -60 -1.71

5 400 340 60 200 40 0 0.00

Error = Actual – Forecast MAD = ∑|Error| / n TS = ∑(Error) / MAD

15
Method 2: Tracking signal

Month Actual Forecast Error ∑|Error| MAD ∑(Error) TS

1 100 100 0 0 0 0 0.00

2 100 100 0 0 0 0 0.00

3 200 150 50 50 16.7 50 3.00

4 200 200 0 50 12.5 50 4.00

5 400 250 150 200 40 200 5.00

Error = Actual – Forecast MAD = ∑|Error| / n TS = ∑(Error) / MAD

16
Interpreting tracking signals

1 MAD ≈ 0.8 s.d.


Method 2

Tracking Signal (MAD)


2
1
0
-1
Method 1
-2
-3
0 1 2 3 4 5
Month

17
How to forecast future demand?

18
Forecasting process

“The forecast”

Step 6 Monitor the forecast


Step 5 Prepare the forecast
Step 4 Gather and analyze data
Step 3 Select a forecasting technique
Step 2 Establish a time horizon
Step 1 Determine purpose of forecast

19
Qualitative forecasting

 When to use?  Advantage


 Historical data are scarce or not  Useful when historical data either
available at all are not available or are scarce
 Use expert opinion to predict  Example: Sales of new products,
future events subjectively environment and technology
change over the long term
 Qualitative methods
 Market research  Limitations
 Historical analogy  Subjective
 Panel consensus  Inconsistent
 Delphi method

20
Quantitative forecasting

 When to use?  Advantage


 Enough historical data are available  Objective, data driven
 Assume the patterns to continue  Forecasts can be updated automatically
 Fully reproducible by any forecasters
 Associative methods
 Simple linear regression models  Limitations
Y = a + bX  Need enough historical data to build
 Multiple regression models and verify the forecasting model
Y = f(X1 , X2 , X3 ,…)  Model specifications can be difficult to
define
 Time series methods
 Moving average
 Exponential smoothing

21
Summary of time series components
Classification
Component of Definition Reason for Influence Duration
Component
Trend Systematic Overall or president, long- Changes in Several years
term upward or downward technology, Trend model
pattern of movement population, wealth,
value
Seasonal Systematic Fairly regular periodic Weather conditions, Within 12 months
fluctuations that occur within social customers, (or monthly or Seasonal model
each 12-month period year religious customs quarterly data)
after year

Cyclical Systematic Repeating up-and-down Interactions of Usually 2-10 years,


swings or movements numerous with differing Not often used for
through four phases: from combinations of intensity for a forecasting sales in
peak (prosperity) to factors that influence complete cycle the near term
contraction (recession) to the economy
trough (depression) to
expansion (recovery or
growth)

Irregular Unsystematic The erratic, or “residual,” Random variations in Short duration and
fluctuations in a series that data or due to non-repeating
exist after taking into unforeseen events
account the systematic such as strikes,
effects hurricanes, and
floods
22
Common time series patterns
Purely random error Increasing linear trend
No recognizable pattern
Demand

Demand
Time Time

Seasonal pattern Seasonal pattern plus linear growth


Demand

Demand
Time Time

23
Example: Forecasting with a simple linear regression model
( )
Sales
Sales
quantity
quantity

2015 2016 2017 2018 2019 Year


Advertising
spend ($M)

Y = a + bX Use of linear regression model for


time series forecasting of demand:
Yt = a + b t
24
Example: Simple exponential smoothing
Initialize F2 = A1

Ft = Ft-1 + α (At-1 – Ft–1) (Use α = 0.5)


Actual Forecast
Period Demand Forecast Error
t At Ft |At - Ft|
1 79
2 84 79.00 5.00
3 83 81.50 1.50
4 81 82.25 1.25
5 98 81.63 16.37
6 100 89.81 10.19
7 ? 94.91
F3 = F2 + α (A2 – F2) MAD = 6.86
F3 = 79 + 0.5 (84 – 79)
n
Mean Absolute Deviation (MAD) = ∑ | At − Ft | / n
t
25
Simple exponential smoothing constants

Effect of Alpha ( α = 0.1 vs. α = 0.5)


105
100
Actual
95
Occupancy 90 Forecast
(α = 0.5)
85
Forecast
80 (α = 01
.)
75
0
1 2 3 4 5
Period

26
Some issues to consider when using exponential smoothing
models for sales forecast…
 The choice of smoothing constant
 0≤α≤1
 Optimal value of α is identified by minimizing forecasting errors (MAD or MSE)

 Adaptive smoothing uses variable smoothing constants based on the latest


percent of error
 Ft+1 = (αt+1) At + (1 – αt+1) (Ft)
where αt+1 = | (At – Ft) / At) | = | PEt |

 Exponential smoothing for trend and seasonal factors

27
Example: Exponential smoothing with trend & seasonality
30,000

Sales of Prius in US
25,000

Actual Forecast
20,000

15,000

10,000

5,000

28

You might also like