Chapter 8 - Forecasting

Decisions that Need

Forecasts
 Which markets to pursue?
 What products to produce?
 How many people to hire?
 How many units to purchase?
 How many units to produce?
 And so on……
Common Characteristics of
Forecasting
 Forecasts are rarely perfect

 Forecasts are more accurate for

aggregated data than for
individual items

 Forecast are more accurate for

shorter than longer time periods
Forecasting Steps
 What needs to be forecast?
 Level of detail, units of analysis & time
horizon required
 What data is available to evaluate?
 Identify needed data & whether it’s
available
 Select and test the forecasting model
 Cost, ease of use & accuracy
 Generate the forecast
 Monitor forecast accuracy over time
Types of Forecasting
Models
 Qualitative (technological)
methods:
 Forecasts generated subjectively by
the forecaster

 Quantitative (statistical) methods:

 Forecasts generated through
mathematical modeling
 Qualitative Methods
Type Characteristics Strengths Weaknesses
Executive A group of managers Good for strategic or One person's opinion
opinion meet & come up with new-product can dominate the
a forecast forecasting forecast

Market Uses surveys & Good determinant of It can be difficult to

research interviews to identify customer preferences develop a good
customer preferences questionnaire

Delphi Seeks to develop a Excellent for Time consuming to

method consensus among a forecasting long-term develop
group of experts product demand,
technological
changes, and
Statistical Forecasting
 Time Series Models:
 Assumes the future will follow same
patterns as the past

 Causal Models:
 Explores cause-and-effect relationships
 Uses leading indicators to predict the future
 E.g. housing starts and appliance sales
 Composition
of Time Series Data
 Data = historic pattern + random
variation
 Historic pattern may include:
 Level (long-term average)
 Trend
 Seasonality
 Cycle
Time Series Patterns
Methods of Forecasting the
Level
 Naïve Forecasting
 Simple Mean
 Moving Average
 Weighted Moving Average
 Exponential Smoothing
Time Series Problem
Determine forecast for Period Orders
periods 11 1 122
 Naïve forecast 2 91
 Simple average 3 100
 3- and 5-period moving 4 77
average 5 115
6 58
 3-period weighted moving
7 75
average with weights 0.5,
0.3, and 0.2
8 128
9 111
 Exponential smoothing
10 88
with alpha=0.2 and 0.5
11
Time Chart of Orders Data

140

120

100

80

60

40

20

0
1 2 3 4 5 6 7 8 9 10
Naïve Forecasting
Next period forecast = Last
Period’s actual:

Ft 1  At
Simple Average (Mean)
Next period’s forecast = average
of all historical data

At  At  1  At  2  .............
Ft 1 
n
Moving Average
Next period’s forecast = simple
average of the last N periods

At  At  1  .........  At  N 1
Ft 1 
N
The Effect of the
Parameter N
 A smaller N makes the forecast
more
responsive
 A larger N makes the forecast
more
stable
Weighted Moving Average

Ft 1 C1 At  C2 At  1  .........  C N At  N 1
where
C1  C2  .........C N 1
Exponential Smoothing

Ft 1 At  1   Ft
where
0  1
The Effect of the
Parameter 
 A smaller  makes the forecast
more
stable
 A larger  makes the forecast more
responsive
 Time Series Problem
Solution

Simple Simple Weighted Exponential Exponential

Naïve Simple Moving Moving Moving Smoothing Smoothing
Period Orders (A) Forecast Average Average (N=3) Average(N=5) Average (N=3) ( = 0.2) ( = 0.5)
1 122 122 122
2 91 122 122 122 122
3 100 91 107 116 107
4 77 100 104 104 102 113 104
5 115 77 98 89 87 106 91
6 58 115 101 97 101 101 108 103
7 75 58 94 83 88 79 98 81
8 128 75 91 83 85 78 93 78
9 111 128 96 87 91 98 100 103
10 88 111 97 105 97 109 102 107

11 88 97 109 92 103 99 98
Forecast Accuracy
 Forecasts are rarely perfect
 Need to know how much we should rely
on our chosen forecasting method
 Measuring forecast error:

Et  At  Ft
 Note that over-forecasts = negative
errors and under-forecasts = positive
errors
 Tracking Forecast Error
Over Time
 Mean Absolute Deviation
(MAD):
MAD 
 actual  forecast
 A good measure of the actual n
error in a forecast


2
actual - forecast
 Mean Square Error (MSE):MSE 
n
 Penalizes extreme errors

 Tracking Signal  actual - forecast 

TS 
 Exposes bias (positive or MAD
negative)
 Accuracy & Tracking Signal Problem: A company is
comparing the accuracy of two forecasting methods. Forecasts
using both methods are shown below along with the actual values
for January through May. The company also uses a tracking signal
with ±4 limits to decide when a forecast should be reviewed.
Which forecasting method is best?

Method A Method B
Month Actu F’cas Error Cum Trackin F’cas Error Cum. Tracking
al t . g t Error Signal
sales Signal
Error
Jan. 30 28 2 2 2 28 2 2 1

Feb. 26 25 1 3 3 25 1 3 1.5
Marc 32 32 0 3 3 29 3 6 3
h
April 29 30 -1 2 2 27 2 8 4
May 31 30 1 3 3 29 2 10 5

MAD 1 2
MSE 1.4 4.4
Forecasting Trends
 Trend-adjusted exponential
smoothing
 Three step process:
St At  (1  
Smooth  1  Tt  1of
)( Stlevel
the ) the series:

Tt   ( St  Stthe
Smooth 1   )Tt  1
 1 )  (trend:

FITt 1 S t  Tt the forecast including trend:

 Calculate
 Forecasting trend problem: a company uses exponential smoothing with
trend to forecast usage of its lawn care products. At the end of July the
company wishes to forecast sales for August. July demand was 62. The
trend through June has been 15 additional gallons of product sold per
month. Average sales have been 57 gallons per month. The company uses
alpha+0.2 and beta +0.10. Forecast for August.

 Smooth the level of the series:

S July αA t  (1  α)(S t  1  Tt  1 ) 0.262  0.857  15 70

 Smooth the trend:

TJuly β(S t  S t  1 )  (1  β)Tt  1 0.170  57   0.915 14.8

 Forecast including trend:

FITAugust S t  Tt 70  14.8 84.8 gallons
Adjusting for Seasonality
 Calculate the average demand per season
 E.g.: average quarterly demand
 Calculate a seasonal index for each season
of each year:
 Divide the actual demand of each season by
the average demand per season for that year
 Average the indexes by season
 E.g.: take the average of all Spring indexes,
then of all Summer indexes, ...
Adjusting for Seasonality
 Forecast demand for the next year &
divide by the number of seasons
 Use regular forecasting method & divide by
four for average quarterly demand
 Multiply next year’s average seasonal
demand by each average seasonal
index
 Result is a forecast of demand for each
season of next year
 Seasonality problem: a university wants to develop
forecasts for the next year’s quarterly enrollments. It has
collected quarterly enrollments for the past two years. It
has also forecast total enrollment for next year to be
90,000 students. What is the forecast for each quarter of
next year?

Quarter Year 1 Seasonal Year 2 Seasonal Avg. Year3

Index Index Index
Fall 24000 26000
Winter 23000 22000
Spring 19000 19000
Summer 14000 17000
Total 90000
Average
 Seasonality Problem: Solution

Quarter Year 1 Seasonal Year 2 Seasonal Avg. Year3

Index Index Index
Fall 24000 1.20 26000 1.24 1.22 27450
Winter 23000 1.15 22000 1.05 1.10 24750
Spring 19000 0.95 19000 0.90 0.93 20925
Summer 14000 0.70 17000 0.81 0.76 17100
Total 80000 4.00 84000 4.00 4.01 90000
Average 20000 21000 22500
Casual Models
 Often, leading indicators hint can
help predict changes in demand
 Causal models build on these
cause-and-effect relationships
 A common tool of causal modeling
is linear regression:
Y a  bx
Linear Regression

Identify dependent (y)
and independent (x)
b
 XY  X  Y  variables
 X 2  X  X   Solve for the slope of the

lineb  XY  n X Y
2
X 2
 nX

 Solve for the y intercept

a Y  b X
 Develop your equation for
the trend line
Y=a + bX
 Linear Regression Problem: A maker of golf shirts has
been tracking the relationship between sales and
advertising dollars. Use linear regression to find out what
sales might be if the company invested $53,000 in
advertising next year.

b
 XY  n XY
Sales $ Adv.$ XY X^2 Y^2 2
(Y) (X)  X  nX2

1 130 48 4240 2304 16,90 30282  451.25147.25

0 b  3.58
10533  451.25
2

2 151 52 7852 2704 22,80 a Y  bX 147.25  3.5851.25

1
a -36.20
3 150 50 7500 2500 22,50 Y a  bX -36.20  3.58x
0  
Y5 -36.20  3.58 53 153.54
4 158 55 8690 3025 2496
4
5 153.8 53
5
Tot 589 205 3028 1053 8716
 How Good is the Fit?
 Correlation coefficient (r) measures the direction and strength of
the linear relationship between two variables. The closer the r value
is to 1.0 the better the regression line fits the data points.

n  XY 
 X  Y 
r
n  X   X  * n  Y   Y 
2 2
2 2

(4)30,282 (205)589
r 2
.888
4(10,533) - (205) * 487,165 589
2

r 2 .982 .788
2

2
 Coefficient of determination r ( ) measures the amount of variation
in the dependent
2
variable about its mean that is explained by the
regressionr line. Values of ( ) close to 1.0 are desirable.
Factors for Selecting a
Forecasting Model
 The amount & type of available
data
 Degree of accuracy required
 Length of forecast horizon
 Presence of data patterns