Week 11

Demand Forecasting in
Business Management

1
1
 Why Forecasting?
Forecasting is an important for all strategic
and planning decisions in a business
organization.

Forecasts of product demand, materials,

labor, financing are an important inputs to
scheduling, acquiring resources, and
determining resource requirements.

2
 Why Forecasting?
Decisions Based on Forecasts
Production Personnel
– Aggregate planning, – Workforce planning,
inventory control, hiring, layoff
scheduling
Marketing
– New product
introduction, sales-
force allocation,
promotions
Finance
– Plant/equipment
investment,
budgetary planning
3
Generally speaking the word
forecast…
There is only one forecast, a Sales forecast.

Financial Plan

Sales Forecast
(Unconstrained) Production Plan

Marketing Plan

Most companies have a hard time distinguishing the difference

between the forecasting process and the planning process…
– Dr. John (Tom) Mentzer, University of Tennessee at
Knoxville

4
Everyone wants to touch the
forecast…
But no one wants to be accountable for the results

“Cultural” Issue

5
Law of Universal Forecasting…

The more people who touch the forecast, the

more inaccurate the forecast, and

The more fact-based (information\data

supported) and mathematically derived the
forecast, the more accurate the forecast…

6
 Principles of Forecasting
Forecasts are almost always wrong.
Every forecast should include an estimate
of the forecast error.
The greater the degree of aggregation, the
more accurate the forecast.
Long-term forecasts are usually less
accurate than short-term forecasts.

7
Factors Affecting Forecasting
Methods

Time frame

Demand behavior

8
 Demand Behavior
Trend
– a gradual, long-term up or down movement of
demand
Cycle
– an up-and-down repetitive movement in demand
Seasonal pattern
– an up-and-down repetitive movement in demand
occurring periodically
Random variations
– movements in demand that do not follow a pattern

9
Theory of Forecasting
Methods...
Forecast = Pattern (s) + Randomness

10
 Forecast = Pattern (s) + Randomness

This simple equation is really saying that when

the average pattern of the underlying data has
been identified some deviation will occur
between the forecasting method applied and the
actual occurrence.

That deviation is called “Error”, or unexplainable

variance.
11
 Forecast = Pattern (s) + Randomness

This simple equation is really saying that when

the average pattern of the underlying data has
been identified some deviation will occur
between the forecasting method applied and the
actual occurrence.

That deviation is called “Error”, or unexplainable

variance.
12
 Forecast = Pattern (s) + Randomness
Maximize Minimize

This simple equation is really saying that when

the average pattern of the underlying data has
been identified some deviation will occur
between the forecasting method applied and the
actual occurrence.

That deviation is called “Error”, or unexplainable

variance.
13
 Forecast = Pattern (s) + Randomness

Two Broad mathematical categories

1. Time Series

2. Causal

14
Forecast = Pattern (s) + Randomness

Trend
Seasonality
Cyclical

15
 Forms of Forecast
Movement

Demand
Demand

Random
movement

Time Time
(a) Trend (b) Cycle

Demand
Demand

Time Time
(c) Seasonal pattern (d) Trend with seasonal pattern
16
 Bottom Line…
There is no Best Method
– The best method depends on the data, the
purpose, the organizational environment and
the perspective of the forecaster.

– Your market, products, goals, and constraints

should be considered when selecting the
forecasting tools best for you...

17
 Forecasting Methods

QUALITATIVE
– use management judgment, expertise, and opinion to
predict future demand
QUANTITATIVE
– Time series
statistical techniques that use historical demand data to predict
future demand
– Regression methods
attempt to develop a mathematical relationship between
demand and factors that cause its behavior

18
 QUALITATIVE METHODS

Delphi method
– involves soliciting forecasts about
technological advances from experts

– simply a kind of expert opinion

19
 Forecasting Process
1. Identify the 2. Collect historical 3. Plot data and identify
purpose of forecast data patterns

6. Check forecast 5. Develop/compute 4. Select a forecast

accuracy with one or forecast for period of model that seems
more measures historical data appropriate for data

7.
Is accuracy of No 8b. Select new
forecast forecast model or
acceptable? adjust parameters of
existing model
Yes
9. Adjust forecast based 10. Monitor results
8a. Forecast over
on additional qualitative and measure forecast
planning horizon
information and insight accuracy

20
 QUANTITAIVE METHODS
Time Series
Assume that what has occurred in the past will
continue to occur in the future
Relate the forecast to only one factor - time
Include
– moving average
– exponential smoothing
– Trend (double) exponential smoothening (Holtz model)
– Triple exponential smoothening (Holtz-Winter model)
– Linear/Non-Linear trend line (Regression)
– Deep learning
– Random Forest
– Reinforcement Learning 21
 Moving Average
Naive forecast
– demand the current period is used as next
period’s forecast
Simple moving average
– stable demand with no pronounced behavioral
patterns
Weighted moving average
– weights are assigned to most recent data

22
Example:

Moving Average:
Naïve Approach
ORDERS
MONTH PER MONTH FORECAST

Jan 120 -
Feb 90 120
Mar 100 90
Apr 75 100
May 110 75
June 50 110
July 75 50
Aug 130 75
Sept 110 130
Oct 90 110
Nov - 90
23
 Simple Moving Average
n

i=1
Di where
MAn = n = number of periods in
n the moving average
Include n most recent observations Di = demand in period i
Weight equally
Ignore older observations

1/n

n ... 3 2 1
today 24
Example:
3-month Simple Moving Average
3
ORDERS MOVING  Di
i=1
MONTH PER MONTH AVERAGE MA3 =
Jan 120 – 3
Feb 90 –
Mar 100 – 90 + 110 + 130
= 3
Apr 75 103.3
May 110 88.3
June 50 95.0
= 110 orders
July 75 78.3
Aug 130 78.3
for Nov
Sept 110 85.0
Oct 90 105.0
Nov - 110.0

25
Example:
5-month Simple Moving Average

ORDERS MOVING
5
MONTH PER MONTH AVERAGE
Jan 120 –

i=1
Di

Feb 90 – MA5 =
Mar 100 –
5
Apr 75 –
90 + 110 + 130+75+50
May 110 – =
June 50 99.0
5
July 75 85.0
Aug 130 82.0 = 91 orders
Sept 110 88.0 for Nov
Oct 90 95.0
Nov - 91.0

26
 Smoothing Effects
150 –

125 – 5-month

100 –
Orders

75 –

50 – 3-month

Actual
25 –

0– | | | | | | | | | | |
Jan Feb Mar Apr May June July Aug Sept Oct Nov
Month
27
 Weighted Moving Average

▪ Adjusts WMAn =  Wi Di
i=1
moving
where
average
method to Wi = the weight for period i,
between 0 and 100 percent
more closely
reflect data  W = 1.00
i
fluctuations

28
Example:
Weighted Moving Average
Example
MONTH WEIGHT DATA
August 17% 130
September 33% 110
October 50% 90
3
November Forecast WMA3 = 
i=1
Wi Di

= (0.50)(90) + (0.33)(110) + (0.17)(130)

= 103.4 orders

29
Let’s GO TO EXCEL for
Moving Average Method

30
 Exponential Smoothing

▪ Averaging method

▪ Weights most recent data more strongly

▪ Reacts more to recent changes

▪ Widely used, accurate method

31
 Include all past observations
Weight recent observations much more heavily than very old
observations:

weight
Decreasing weight given
to older observations

today

32
33
34
35
36
Exponential Smoothing: Analytics

37
New forecast = weighted sum of last period
actual value and last period forecast
– : Smoothing constant
– Ft : Forecast for period t
– Ft-1: Last period forecast
– Dt-1: Last period actual value
38
 Effect of Smoothing Constant
0.0    1.0
If  = 0.20, then Ft +1 = 0.20 Dt + 0.80 Ft

If  = 0, then Ft +1 = 0 Dt + 1 Ft = Ft
Forecast does not reflect recent data

If  = 1, then Ft +1 = 1 Dt + 0 Ft = Dt
Forecast based only on most recent data
39
Example:
Exponential Smoothing (α=0.30)

PERIOD MONTH DEMAND Sample calculations for 2nd period

1 Jan 37 F2 = D1 + (1 - )F1
2 Feb 40 = (0.30)(37) + (0.70)(37)
3 Mar 41
= 37
4 Apr 37 F3 = D2 + (1 - )F2
5 May 45 = (0.30)(40) + (0.70)(37)
6 Jun 50
= 37.9
7 Jul 43
8 Aug 47
F13 = D12 + (1 - )F12
9 Sep 56
10 Oct 52 = (0.30)(54) + (0.70)(50.84)
11 Nov 55 = 51.79
12 Dec 54

40
Exponential Smoothing
(cont.)
FORECAST, Ft + 1
PERIOD MONTH DEMAND ( = 0.3) ( = 0.5)
1 Jan 37 – –
2 Feb 40 37.00 37.00
3 Mar 41 37.90 38.50
4 Apr 37 38.83 39.75
5 May 45 38.28 38.37
6 Jun 50 40.29 41.68
7 Jul 43 43.20 45.84
8 Aug 47 43.14 44.42
9 Sep 56 44.30 45.71
10 Oct 52 47.81 50.85
11 Nov 55 49.06 51.42
12 Dec 54 50.84 53.21
13 Jan – 51.79 53.61
41
 Exponential Smoothing (cont.)
70 –

60 – Actual  = 0.50

50 –

40 –
Orders

 = 0.30
30 –

20 –

10 –

0– | | | | | | | | | | | | |
1 2 3 4 5 6 7 8 9 10 11 12 13
Month
42
Simple Exponential Smoothing
Properties of Simple Exponential Smoothing
– Widely used and successful model

– Requires very little data

– Larger , more responsive forecast; Smaller , smoother

forecast

– “best”  can be found by Solver

– Suitable for relatively stable time series

43
 Let’s GO TO EXCEL for
Exponential Smoothening
Method

44
 Complicating Factors
Simple Exponential Smoothing works well with
data that is “stable time series”

Must be adapted for data series which exhibit a

definite trend

Must be further adapted for data series which

exhibit seasonal patterns

45
 Holt’s Method:
Double Exponential Smoothing
Ideas behind smoothing with trend:
– ``De-trend'' time-series by separating base from trend
effects
– Smooth base in usual manner using 
– Smooth trend forecasts in usual manner using 

46
 AFt +1 = Lt +1 + Tt +1 or

AFt = Lt + Tt
Forecast k periods into future Ft+k with base and trend

where
T = an exponentially smoothed trend factor

Tt +1 = (Lt +1 - Lt) + (1 - ) Tt or

Tt = (Lt – Lt-1) + (1 - ) Tt-1

where
AF = Adjusted forecast
Tt = the last period trend factor
 = a smoothing constant for trend
 = a smoothing constant for level
47
Example:
Adjusted Exponential
Smoothing (β=0.30)
PERIOD MONTH DEMAND
Sample calculations for 3rd period
1 Jan 37 T3 = (L3 - L2) + (1 - ) T2
2 Feb 40 = (0.30)(38.5 - 37.0) + (0.70)(0)
3 Mar 41
4 Apr 37 = 0.45
AF3 = L3 + T3 = 38.5 + 0.45
5 May 45
= 38.95
6 Jun 50
7 Jul 43 T13 = (L13 - L12) + (1 - ) T12
8 Aug 47 = (0.30)(53.61 - 53.21) + (0.70)(1.77)
9 Sep 56 = 1.36
10 Oct 52
11 Nov 55 AF13 = L13 + T13 = 53.61 + 1.36 = 54.96
12 Dec 54

48
 Adjusted Exponential Smoothing:
Example
FORECAST TREND ADJUSTED
PERIOD MONTH DEMAND Lt +1 Tt +1 FORECAST AFt +1

1 Jan 37 37.00 – –
2 Feb 40 37.00 0.00 37.00
3 Mar 41 38.50 0.45 38.95
4 Apr 37 39.75 0.69 40.44
5 May 45 38.37 0.07 38.44
6 Jun 50 38.37 0.07 38.44
7 Jul 43 45.84 1.97 47.82
8 Aug 47 44.42 0.95 45.37
9 Sep 56 45.71 1.05 46.76
10 Oct 52 50.85 2.28 58.13
11 Nov 55 51.42 1.76 53.19
12 Dec 54 53.21 1.77 54.98
13 Jan – 53.61 1.36 54.96
49
 Adjusted Exponential Smoothing
Forecasts
70 –

Adjusted forecast ( = 0.30)

60 –
Actual
50 –

40 –
Demand

30 – Forecast ( = 0.50)

20 –

10 –

0– | | | | | | | | | | | | |
1 2 3 4 5 6 7 8 9 10 11 12 13
Period
50
 Holt-Winter’s Method:
Exponential Smoothing w/ Trend & Seasonality

Ideas behind smoothing with trend and seasonality:

– “De-trend” and “de-seasonalize” time-series by separating
base from trend and seasonality effects
– Smooth base in usual manner using 
– Smooth trend forecasts in usual manner using 
– Smooth seasonality forecasts using 

Assume m seasons in a cycle

– 12 months in a year
– 4 quarters in a year
– 3 months in a quarter
– et cetera
51
Smooth the base forecast Lt

Smooth the trend forecast Tt

Smooth the seasonality forecast St

Adjusted Forecast Ft with trend and seasonality

Forecast k periods into future Ft+k with trend & seasonality

52
There are following two ways to adjust seasonality:

1. Multiplicative seasonality: (Level + Trend) *

Seasonality

2. Additive Seasonality: Level + Trend + Seasonality

This is a
multiplicative one

53
Let’s GO TO EXCEL for Holt-
Winter (Triple) Exponential
Smoothening Method

54
Example Using Holt-Winter Model

DEMAND (1000’S PER QUARTER)

YEAR 1 2 3 4
2002 12.6 8.6 6.3 17.5
2003 14.1 10.3 7.5 18.2
2004 15.3 10.6 8.1 19.6
2005 ?

55
Example:
DEMAND (1000’S PER QUARTER)
YEAR 1 2 3 4 Total
2002 12.6 8.6 6.3 17.5 45.0
2003 14.1 10.3 7.5 18.2 50.1
2004 15.3 10.6 8.1 19.6 53.6
Total 42.0 29.5 21.9 55.3 148.7

Di
Seasonal factor = Si = D
D1 42.0 D3 21.9
S1 = = = 0.28 S3 = = = 0.15
D 148.7 D 148.7
D 29.5 D 55.3
S2 = 2 = = 0.20 S4 = 4 = = 0.37
D 148.7 D 148.7
56
 Seasonal Adjustment (cont.)

For 2005

y = 40.97 + 4.30x = 40.97 + 4.30(4) = 58.17

SF1 = (S1) (F5) = (0.28)(58.17) = 16.28

SF2 = (S2) (F5) = (0.20)(58.17) = 11.63
SF3 = (S3) (F5) = (0.15)(58.17) = 8.73
SF4 = (S4) (F5) = (0.37)(58.17) = 21.53

57
 Comparison of
Holts & Winter Model

58
Regression: Linear Trend Line

 xy - nxy
y = a + bx b =
 x2 - nx2
where a = y-bx
a = intercept
b = slope of the line where
x = time period n = number of periods
y = forecast for
demand for period x x
x = = mean of the x values
n
y
y = n = mean of the y values

59
 Least Squares Example
x(PERIOD) y(DEMAND) xy x2
1 37 37 1
2 40 80 4
3 41 123 9
4 37 148 16
5 45 225 25
6 50 300 36
7 43 301 49
8 47 376 64
9 56 504 81
10 52 520 100
11 55 605 121
12 54 648 144
78 557 3867 650

60
 Least Squares
Example (cont.)
78
x = = 6.5
12
557
y = = 46.42
12
xy - nxy 3867 - (12)(6.5)(46.42)
b = = =1.72
x - nx
2 2
650 - 12(6.5) 2

a = y - bx
= 46.42 - (1.72)(6.5) = 35.2

61
 Linear trend line y = 35.2 + 1.72x
Forecast for period 13 y = 35.2 + 1.72(13) = 57.56 units

70 –

60 –
Actual

50 –
Demand

40 –
Linear trend line
30 –

20 –

10 – | | | | | | | | | | | | |
1 2 3 4 5 6 7 8 9 10 11 12 13
0– Period

62
Let’s GO TO EXCEL for
Regression Analysis

63
 Forecasting Accuracy or
Performance
How good is the forecast?
Mean Forecast Error (MFE or Bias): Measures
average deviation of forecast from actuals.

Mean Absolute Deviation (MAD): Measures

average absolute deviation of forecast from
actuals.

Mean Absolute Percentage Error (MAPE):

Measures absolute error as a percentage of the
forecast.

Standard Squared Error (MSE): Measures

variance of forecast error 64
Forecasting Performance Measures

65
 Mean Forecast Error
(MFE or Bias)

Want MFE to be as close to zero as possible -- minimum

bias
A large positive (negative) MFE means that the forecast
is undershooting (overshooting) the actual observations
Note that zero MFE does not imply that forecasts are
perfect (no error) -- only that mean is “on target”
Also called forecast BIAS

66
Mean Absolute Deviation (MAD)

Measures absolute error

Positive and negative errors thus do not cancel out (as
with MFE)
Want MAD to be as small as possible
No way to know if MAD error is large or small in relation
to the actual data

67
Mean Absolute Percentage Error
(MAPE)

Same as MAD, except ...

Measures deviation as a percentage of actual data

68
Mean Squared Error (MSE)

Measures squared forecast error -- error variance

Recognizes that large errors are disproportionately more
“expensive” than small errors
But is not as easily interpreted as MAD, MAPE -- not as
intuitive

69
 Comparison of
Forecast

FORECAST MAD MAPD E (E)

Exponential smoothing ( = 0.30) 4.85 9.6% 49.31 4.48
Exponential smoothing ( = 0.50) 4.04 8.5% 33.21 3.02
Adjusted exponential smoothing 3.81 7.5% 21.14 1.92
( = 0.50,  = 0.30)
Linear trend line 2.29 4.9% – –

70
 Tracking Signal Plot

3 –
Tracking signal (MAD)

2 –
Exponential smoothing ( = 0.30)
1 –

0 –

-1 –

-2 – Linear trend line

-3 –
| | | | | | | | | | | | |
0 1 2 3 4 5 6 7 8 9 10 11 12
Period

71
72