Forecasting

Yousaf Ali Khan

Department of Management Sciences and Humanities
GIK Institute of Engineering Sciences and Technology

What is Forecasting?
Process of predicting a future event

A statement about the future.

Forecasts more accurate for

groups vs. individuals

Forecast accuracy decreases

as time horizon increases

I see that you will

get an A this semester.

Uses of Forecasts
Accounting

Cost/profit estimates

Finance

Cash flow and funding

Human Resources

Hiring/recruiting/training

Marketing

Pricing, promotion, strategy

MIS

IT/IS systems, services

Operations

Schedules, workloads

Product/service design

New products and services

Forecasting Time Horizons

Short-range forecast
Up to 1 year, generally less than 3 months
Purchasing, job scheduling, workforce
levels, job assignments, production levels

Medium-range forecast
3 months to 3 years
Sales and production planning, budgeting

Long-range forecast
3+ years
New product planning, facility location,
research and development

Elements of a Good Forecast

Timely

Reliable

Accurate

Written

Steps in the 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

Steps in the Forecasting Process

Types of Forecasts
Economic forecasts
Address business cycle inflation
rate, money supply, etc.

Technological forecasts
Predict rate of technological progress
Impacts development of new
products

Demand forecasts
Predict sales of existing product

Determine the type of model to be

used
1. Who will be using the forecast and what
information do they require?
2. How relevant is historical data, and what is
its availability?
3. How accurate does the forecast have to be?
4. What is the time period of the forecast?
5. How much time do we have to develop the
forecast?
6. What is the cost or benefit (value) of this
forecast to our company?

Determine the forecast horizon

Inverse relationship between forecast
accuracy and time horizon.
The longer the time horizon the more
inaccurate the forecast will be.
Time horizon should be at least as long as
time period of strategic plan.
Product life cycles influence length of
forecasts.
Technological product sales would have a
short forecast.
Milk sales would have a long forecast.

Types of Forecasting Models

Judgmental models, which use qualitative
methods, uses subjective inputs
Time series models, which use quantitative
methods, uses historical data assuming
the future will be like the past
Causal models or Associative Model,
uses explanatory variables to predict the future
which use cause-and effect methods.

Forecasting Approaches
Qualitative Methods
Used when situation is vague and
little data exist
New products
New technology

Involves intuition, experience

Forecasting Approaches
Quantitative Methods

Used when situation is stable and

historical data exist
Existing products
Current technology

Involves mathematical techniques

e.g., forecasting sales of color televisions

Judgmental Models
Judgmental models are qualitative and essentially use
estimates based on expert opinion.
Survey of Sales Forces: most appropriate for manufacturing and
wholesale firms.
Surveys of Customers: applicable to all firms. Customers express
preference for new or modified products.
Historical Analogy: most appropriate for firms that have several
outlets. Introduction of new product which has characteristics similar to
previous products.
Market Research: can include surveys, tests, and observations.
Results are statistically extrapolated to develop forecasts of demand
for products.
Delphi Method uses a panel of experts to obtain a consensus of
opinion. Used primarily for unique new products or processes for
which no previous data exists.

Time Series Models

Time series forecasting
models normally use historical
records that are readily available
within the firm or industry to predict
future sales.
For this reason they are often referred to as
internal or intrinsic models.
Assumption in time series forecasting is that past
sales are a fairly accurate predictor of future sales.

Overview of Quantitative
Approaches
.
1. Naive approach
2. Moving averages
3. Exponential
smoothing
4. Trend projection
5. Linear regression

Time-Series
Models

Associative
Model

Time Series Components

Trend

Cyclical

Seasonal

Random

Time Series Forecasts

Trend - long-term movement in data
Seasonality - short-term regular variations
in data
Cycle wavelike variations of more than
one years duration
Irregular variations - caused by unusual
circumstances
Random variations - caused by chance

Forecast Variations
Irregular
variation

Trend

Cycles
90
89
88
Seasonal variations

Trend Component
Persistent, overall upward or
downward pattern
Changes due to population,
technology, age, culture, etc.
Typically several years
duration

Seasonal Component
Regular pattern of up and
down fluctuations
Due to weather, customs, etc.
Occurs within a single year
Period
Week
Month
Month
Year
Year
Year

Length
Day
Week
Day
Quarter
Month
Week

Number of
Seasons
7
4-4.5
28-31
4
12
52

Cyclical Component
Repeating up and down movements

Affected by business cycle, political,

and economic factors
Multiple years duration
Often causal or
associative
relationships
0

10

15

20

Random Component
Erratic, unsystematic, residual
fluctuations
Due to random variation or
unforeseen events
Short duration and
nonrepeating

Naive Approach
Assumes demand in next
period is the same as
demand in most recent period
e.g., If January sales were 68, then
February sales will be 68
Sometimes cost effective and efficient
Can be good starting point
The forecast for any period equals the
previous periods actual value.

Simple Mean or Average

One of the simplest averaging models is the simple mean or

average. Here the forecast is made by simply taking an average
of all data:

Moving Average Method

MA is a series of arithmetic means
Used if little or no trend

Used often for smoothing

Provides overall impression of data over
time
Moving average =

demand in previous n periods

n

Moving Average Example

Month

January
February
March
April
May
June
July

Actual
Shed Sales

10
12
13
16
19
23
26

3-Month
Moving Average

(10 + 12 + 13)/3 = 11.66

(12 + 13 + 16)/3 = 13.66
(13 + 16 + 19)/3 = 16
(16 + 19 + 23)/3 = 19.33

Weighted Moving Average

Used when trend is present
Older data usually less important

Weights based on experience and

intuition
Weighted
moving average

(weight for period n)

x (demand in period n)
weights

Weights Applied
3
2
1
6

Period
Last month
Two months ago
Three months ago
Sum of weights

Weighted Moving Average

Month

January
February
March
April
May
June
July

Actual
Shed Sales

3-Month Weighted
Moving Average

10
12
13
16
19
23
26

[(3 x 13) + (2 x 12) + (10)]/6 = 121/6

[(3 x 16) + (2 x 13) + (12)]/6 = 141/3
[(3 x 19) + (2 x 16) + (13)]/6 = 17
[(3 x 23) + (2 x 19) + (16)]/6 = 201/2

Moving Average And

Weighted Moving Average
Weighted
moving
average

30
Sales demand

25
Actual
sales

20
15

Moving
average

10
5
|

|
F

|
M

|
A

|
M

|
J

|
J

|
A

|
S

|
O

|
N

|
D

Exponential Smoothing Model

Exponential smoothing model uses a smoothing
constant, alpha (), as an adjustment in determining
the forecast.
A smoothing constant is a value assigned by the
forecaster to adjust the forecast based on the
forecaster assumption of the relationship between
sales in one time period and sales in the next time
period.
Alpha can have any value between 0 and 1; however,
alpha is normally 0.1, 0.2 or 0.3.

Exponential Smoothing
New forecast = last periods forecast
+ a (last periods actual demand
last periods forecast)
Ft = Ft 1 + a(At 1 - Ft 1)
where

Ft =
Ft 1 =
a =

new forecast
previous forecast
smoothing (or weighting)
constant (0 a 1)

Choosing
The objective is to obtain the most
accurate forecast no matter the
technique.
We generally do this by selecting the model
that gives us the lowest forecast error
Forecast error = Actual demand - Forecast value
= At - Ft

Common Measures of Error

Cumulative Sum of Forecast Errors
CFE

e = (actual forecast)

Mean Absolute Deviation (MAD)

|actual - forecast|

MAD =

n
Tracking Signal
(actual forecast)
MAD

TS=

Good tracking signal

has low values

Mean Squared Error

(forecast errors)2
MSE =
n
Mean Absolute Percent Error (MAPE)
n
MAPE =

100

|actual - forecast |/actual

i

i=1

Causal Models
Causal models are also known as external
or exogenous models.
Causal models take into account variables in the general
economy that affect the revenue obtained by a company.
Causal models can be simple or very complex.
Most of them require multiple regression analysis, which is
normally beyond the scope of a small business manager.

One of the best known causal method is

regression model.

When Is Causal Forecasting Used?

Know or believe something caused demand to act a certain
way
Demand or sales patterns that vary drastically with
planned or unplanned events
Examples of a variable that causes demand to act a certain
way:
Sales of ice cream increase when temperature is high
New home starts increase when interest rates are low
Number of employees increase when demand increases
An planned event could be a sale or and advertising
promotion.
An unplanned event could be a snow storm or severe
weather, strike, or materials shortages.

Linear Regression Model

Linear regression uses a statistical method known as

least squared method, this method minimizes the sum of
the squared errors (Deviations)
Four areas of variation:
Seasonal variation is caused by the predictable
shopping habits of our customers.
Trend variation is variation caused by growth or decline
in demand for our product or service over time.
Cyclical variation is caused by general economic factors
that affect our industry.
Noise is random variation in our data that is not
explained by the preceding factors.

Linear Regression Model (Cont)

Linear regression is used to determine two factors:

the slope of the regression line
the intercept of the regression line
Basic formula for the regression line
y = a + bx
y is the dependent variable. A dependent variable is one
that relies on other variables for its value.
x is the independent variable. An independent variable is one
that does not depend on other variables for its value. In
forecasting models, x is often a time period.
a is the y intercept. The y intercept is the value of y when x
equals 0.
b is the slope of the regression line. Slope is defined as rise
over run.

Least Squares Method

Equations to calculate the regression variables
^

y = a + bx
b=

Sxy - nxy

Sx2 - nx2

a = y - bx

Least Squares Example

Year

1999
2000
2001
2002
2003
2004
2005

Time
Period (x)

Electrical Power
Demand

1
2
3
4
5
6
7
x = 28
x=4
b=

74
79
80
90
105
142
122
y = 692
y = 98.86
xy - nxy
x2 - nx2

x2

xy

1
4
9
16
25
36
49
x2 = 140

74
158
240
360
525
852
854
xy = 3,063

3,063 - (7)(4)(98.86)
=
= 10.54
140 - (7)(42)

a = y - bx = 98.86 - 10.54(4) = 56.70

Least Squares Example

Year

Time
Period (x)

Electrical Power
Demand

1999
2000
2001The
2002
2003
2004
2005

1
2
trend
3 line is
4
^
y =5 56.70 +
6
7

Sx = 28
x=4
b=

xy

1
4
9
16
25
36
49

74
158
240
360
525
852
854

Sx2 = 140

Sxy = 3,063

74
79
80
90
105
10.54x
142
122

Sy = 692
y = 98.86
Sxy - nxy
Sx2 - nx2

x2

3,063 - (7)(4)(98.86)
=
140 - (7)(42)

a = y - bx = 98.86 - 10.54(4) = 56.70

= 10.54

Power demand

Least Squares Example

160
150
140
130
120
110
100
90
80
70
60
50

Trend line,
^y = 56.70 + 10.54x

|
1999

|
2000

|
2001

|
2002

|
2003
Year

|
2004

|
2005

|
2006

|
2007

Values of Dependent Variable

Least Squares Method

Actual observation
(y value)

Deviation7

Deviation5

Deviation6

Deviation3
Deviation4
Deviation1
Deviation2

Trend line, y =^ a + bx

Time period

Correlation
Measures the strength of the relationship
between the dependent and independent
variable
Once a model has been developed and tested,
the effectiveness of the model needs to be
determined. Correlation is used to determine
how strong the relationship between the
dependent and independent variables are.

Correlation Coefficient Formula

r=

nSxy - SxSy
[nSx2 - (Sx)2][nSy2 - (Sy)2]

r = correlation coefficient
n = number of periods
x = the independent variable
y = the dependent variable

y
Correlation

(a) Perfect positive

correlation:
r = +1

(b) Positive
correlation:
0<r<1

(c) No correlation:
r=0

(d) Perfect negative

correlation:
r = -1

Coefficient of Determination
Coefficient of Determination, r2, measures the
percent of change in y predicted by the change in x
Values range from 0 to 1
A high value of r2, say .80 or more, would indicate
that the independent variable is a good predictor of
values of the dependent variable.
A low value say .25 or less, would indicate a poor
predictor
A value Between .25 and .80 would indicate a
moderate predictor.

Seasonal Variations In Data

The multiplicative seasonal model can modify
trend data to accommodate seasonal
variations in demand
1.

Find average historical demand for each season

2.

Compute the average demand over all seasons

3.

Compute a seasonal index for each season

4.

Estimate next years total demand

5.

Divide this estimate of total demand by the number of

seasons, then multiply it by the seasonal index for that
season

Seasonal Index Example

Month

Demand
2003 2004 2005

Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sept
Oct
Nov
Dec

80
70
80
90
113
110
100
88
85
77
75
82

85
85
93
95
125
115
102
102
90
78
72
78

105
85
82
115
131
120
113
110
95
85
83
80

Average
2003-2005

Average
Monthly

90
80
85
100
123
115
105
100
90
80
80
80

94
94
94
94
94
94
94
94
94
94
94
94

Seasonal
Index

Seasonal Index Example

Month

Demand
2003 2004 2005

Average
2003-2005

Average
Monthly

Jan
80
85 105
90
94
Feb
70
85
85
80
94
Mar
80
93 average
82 2003-2005
85monthly demand
94
Seasonal index
Apr
90 = 95 115average monthly
100 demand 94
May
113 125 131
123
94
= 90/94 = .957
Jun
110 115 120
115
94
Jul
100 102 113
105
94
Aug
88 102 110
100
94
Sept
85
90
95
90
94
Oct
77
78
85
80
94
Nov
75
72
83
80
94
Dec
82
78
80
80
94

Seasonal
Index
0.957

Seasonal Index Example

Month

Demand
2003 2004 2005

Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sept
Oct
Nov
Dec

80
70
80
90
113
110
100
88
85
77
75
82

85
85
93
95
125
115
102
102
90
78
72
78

105
85
82
115
131
120
113
110
95
85
83
80

Average
2003-2005

Average
Monthly

Seasonal
Index

90
80
85
100
123
115
105
100
90
80
80
80

94
94
94
94
94
94
94
94
94
94
94
94

0.957
0.851
0.904
1.064
1.309
1.223
1.117
1.064
0.957
0.851
0.851
0.851

Seasonal Index Example

Month

Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sept
Oct
Nov
Dec

Demand
2003 2004 2005

Average
2003-2005

80
85 105
90
for 2006 80
70
85 Forecast
85
80
93
82
85
90Expected
95 annual
115 demand = 1,200
100
113 125 131
123
110 115 120 1,200
115
Jan
x
.957 = 96
100 102 113 12
105
88 102 110 1,200
100
Feb
x .851
85
90
95 12
90 = 85
77
78
85
80
75
72
83
80
82
78
80
80

Average
Monthly

94
94
94
94
94
94
94
94
94
94
94
94

Seasonal
Index

0.957
0.851
0.904
1.064
1.309
1.223
1.117
1.064
0.957
0.851
0.851
0.851

Seasonal Index Example

2006 Forecast
2005 Demand
2004 Demand
2003 Demand

140
130
Demand

120
110

100
90
80

70
|
J

|
F

|
M

|
A

|
M

|
|
J
J
Time

|
A

|
S

|
O

|
N

|
D