UNIT-II

FORECASTING

Demand forecasting is the process of making estimations about future customer demand over a
defined period, using historical data and other information.

Purposes of forecasting:

Purposes of short-term forecasting

 Appropriate production scheduling.

 Reducing costs of purchasing raw materials.
 Determining appropriate price policy
 Setting sales targets and establishing controls and incentives.
 Evolving a suitable advertising and promotional campaign.
 Forecasting short term financial requirements.

Purposes of long-term forecasting

 Planning of a new unit or expansion of an existing unit.

 Planning long term financial requirements.
 Planning man-power requirements.

Stages in forecasting:

 Setting the Objectives: The purpose for which the demand forecasting is being done
must be clear. Whether it is for short-term or long-term, the market share of the product,
the market share of the organisation, competitors share, etc. By all these aspects, the
objectives for forecasting are framed.

 Determining the Time Perspective: The defined objectives are supported by the period
for which the forecasting is being done. The demand for a commodity varies with the
change in its determinants over the period.
 There is a negligible change in price, income or other factors in the short run. But, the
organisation may notice a considerable difference in these determinants over a long-term,
affecting the demand of a commodity.

 Selecting a Suitable Demand Forecasting Method: Demand forecasting is based on

specific evidence and is determined using a particular technique or method. The method
of prediction must be selected wisely. It is dependant on the information available, the
purpose of predicting and the period it is done for.

 Collecting the Data: Forecasting is based on past experiences and data. This data or
information can be primary or secondary. Primary data comprises of the information
directly collected by the analysts and researchers; whereas secondary data includes the
physical evidence of the past performance, sales trend in the past years, financial reports,
etc.
 Estimating the Results: The data so collected is arranged in a systematic and meaningful
manner. The past performance of a product in the market is analyzed on this basis.
Accordingly, future sales prediction and demand estimation are done. The results so drew
must be in a format which is easy to understand and apply by the management.

Factors effecting demand forecasting:

General Guiding Principles for Forecasting

1. Forecasts are more accurate for larger groups of items.

2. Forecasts are more accurate for shorter periods of time.
3. Every forecast should include an estimate of error.
4. Before applying any forecasting method, the total system should be understood.
5. Before applying any forecasting method, the method should be tested and evaluated.

Types of Forecasts by Time Horizon

1. Short-range forecast
2. Medium-range forecast
3. Long-range forecast
1. Short-range forecast

 It is usually up to 1 year, or generally less than 3 months

 It is required for current purchasing, job scheduling, workforce levels, job assignments,
production levels
 It is concerned with capacity utilization of the firm
 Its accuracy will be more

2.Medium-range forecast

 It is usually one to 3 years

 It is required for Sales and production planning, budgeting
 It is useful for improving or increasing the existing capacity of the plant
 Its accuracy will be good
3.Long-range forecast

 It is usually more than 3 years such as 3 to 5 years.

 It is to consider the changes in the environments such as population, technology,
competition in the market and financial policies of the government
 It is useful for new product planning, facility location, research and development
 Its accuracy will be less

Types of Forecasting Methods

1. Qualitative methods
2. Quantitative methods

Qualitative methods: These types of forecasting methods are based on judgments, opinions,
intuition, emotions, or personal experiences and are subjective in nature. They do not rely on any
rigorous mathematical computations

Quantitative methods: These types of forecasting methods are based on mathematical

(quantitative) models, and are objective in nature. They rely heavily on mathematical
computations.

Qualitative methods Quantitative methods

Based on human judgment
opinions; subjective and Based on mathematics;
1. Characteristics
nonmathematical. quantitative in nature
Consistent and objective;
Can incorporate latest
able to consider much
changes in the environment
2. Strengths information and data at one
and “inside information
time.
Often quantifiable data are
Can bias the forecast and not available. Only as good
3. Weaknesses reduce forecast accuracy as the data on which they
are based.

Qualitative methods of forecasting.

These methods are used when

1. The time available is not enough for objectively determining the fore casts using quantitative
methods
2. Historical data may not be available
Ex: In the launch of new innovative Product there is no data available from past experience on
the sales
of the Product
3. The country facing economic and political turmoil
4. The available data may become redundant.
Types of Qualitative Forecasting Methods

i. Customer Surveys
ii. Sales Force Composites
iii. Jury of Executive Opinion
iv. The Delphi Method

i. Customer Surveys

 Only a few consumers are selected and their views on the probable demand are collected.
 Thus, it is a miniature form of Complete Enumeration Survey.
 The sample is considered to be a true representation of the entire population.
 This method is simple and cheaper.
 The results of survey can be obtained quickly and results are good.
ii. Sales Force Composites
 In this method sales persons are expected to estimate expected sales in their respective
territories.
 The sales force, which has been selling the product to wholesalers / retailers / consumers
over a period of time, is considered to know the product and the demand pattern very
well.
 These method does not require intricate mathemathical calculations.
 This method 1s based on the first hand knowledge of the salesman.
 It is useful to forecast the sales of new products.

iii. Expert Opinion Method :

 This technique of forecasting demand seeks the views of experts on the likely level of
 demand in the future. They have a rich experience of the behaviour of demand.
 If the forecasting is based on the opinion of several experts, then it is known panel
consensus.
 A specialized form of panel opinion is the Delphi Method. This method seeks the opinion
of a group of experts through mail about the expected level of demand.
 The responses so received are analyzed by an independent body.

iv. The Delphi method

 Choose the experts to participate. There should be a variety of knowledgeable people in

different areas.
 Through a questionnaire (or e-mail), obtain forecasts (and any premises or qualification
captions for the forecasts) from all participants.
 Summarize the results and redistribute them to the participants along with appropriate
new questions.
 Summarize again, refining forecasts and conditions, and again develop new questions.
 Repeat Step 4 if necessary. Distribute the final results to all participants.
Quantitative Forecasting Methods

Demand patterns in forecasting

Time Series: The repeated observations of demand for a service or product in their order of
occurrence.

There are five basic patterns of most time series.

a. Horizontal: The fluctuation of data around a constant mean.
b. Trend: The systematic increase or decrease in the mean of the series over time.
c. Seasonal A repeatable pattem of increases or decreases in demand, depending on the time of
day, week, month, or season.
d. Cyclical: The less predictable gradual increases or decreases over longer periods of time
(years or decades).
1. Stationary (horizontal) demand pattern:

 No growth/decline trend, no seasonal variation.

 This demand pattern exists when the demand in the past do not have any increasing and
decreasing trend i.e only random fluctuations are present around a Particular level of
demand. The mean value does not change over a time
 Ex: Products with stable sales, number of defective items from a stable production
processes
Ex: Rice, vegetables, fruits, on petrol, Gas.
 The simple or weighted moving average method or simple exponential smoothing
methods are suitable.

2. Seasonal demand pattern:

 No growth / decline trend, seasonal variation

 This pattern exists when the series fluctuations according to some seasonal factor the
season may be months, quarters, weeks, etc.
 When there is no increasing or decreasing trend but there are seasonal variations .
(similar crests and troughs in the curve at regular intervals of time).
 Ex: sale of refrigerators, sale of soft drinks, sale of wool items, etcThe simple moving
average is suitable

3. Trend pattern :-

 Linear growth or decline, no seasonal variation.

 This pattern exists when there is no increase or decrease in the value of the demand over
a time i.e increasing or decreasing trend with no seasonal variations.
 Ex: sales of many products, stock prices business and economic indicators
 Holt's double exponential smoothing more appropriate, ,

4. Cyclical pattern

 Linear growth/ decline trend, seasonal variation.

 This pattern exists when the length of a single cycle is longer than a year. The cycle does
not repeat at constant intervals of time.
 i.e where there is an increasing or decreasing trend with seasonal variations
 Ex: prices of some metals, gross national product
Time Series Models

 Try to predict the future based on past data

 Assume that factors influencing the past will continue to influence the future

Naïve method:
– The forecast is equal to the actual value observed during the last period
– good for level patterns

Ft 1  Dt
Moving Average method:
– The average value over a set time period (e.g.: the last three weeks)
– Each new forecast drops the oldest data point & adds a new observation
– More responsive to a trend but still lags behind actual data

Ft1   Dt / n
Weighted Moving Average:

W D t t

M t  t 1n
W
t 1
t

• All weights must add to 100% or 1.00

e.g. Wt=0 .5, Wt-1 =0.3, Wt-2 =0.2 (weights add to 1.0)

• Allows emphasizing one period over others; above indicates more weight on recent data
(Wt=0.5)

• Differs from the simple moving average that weighs all periods equally - more responsive
to trends

• Advantages of Moving Average Method

– Easily understood
– Easily computed
– Provides stable forecasts
 • Disadvantages of Moving Average Method
– Requires saving lots of past data points: at least the N periods used in the moving
average computation
– Lags behind a trend
– Ignores complex relationships in data

Exponential Smoothing method

• Assumes the most recent observations have the highest predictive value
– gives more weight to recent time periods

Ft+1 = Forecast value for time t+1

Dt = Actual value at time t
α = Smoothing constant
 Most frequently used time series method because of ease of use and minimal amount of
data needed
• Need just three pieces of data to start:
– Last period’s forecast (Ft)
– Last periods actual value (Dt)
– Select value of smoothing coefficient’α’ between 0 and 1.0
• If no last period forecast is available, average the last few periods or use naive method
• Higher ‘α’ values (e.g. 0.7 or 0 .8) may place too much weight on last period’s random
variation
Trend Adjusted Exponential Smoothing method

A trend in a time series is a systematic increase or decrease in the average of the series over time.

 Where a significant trend is present, exponential smoothing approaches must be

modified; otherwise, the forecasts tend to be below or above the actual demand.

Trend-adjusted exponential smoothing method for incorporating a trend in an exponentially

smoothed forecast.

 With this approach, the estimates for both the average and the trend are smoothed,
requiring two smoothing constants. For each period, we calculate the average and the
trend.
Regression Model

Simple Linear Regression Model

y = a + bx
where a = intercept
b = slope of the line
x = Independent variable
y = Dependent variable
• Identify dependent (y) and independent (x) variables
• Solve for the slope of the line

b
XY  nXY
 X 2  nX
2

• Solve for the y intercept

a  Y  bX
• Develop the equation for the trend line
Y=a + bX
 FORECASTING QUANTITATIVE TECHNIQUES (PROBLEMS)

Time Series Models

 Try to predict the future based on past data
 Assume that factors influencing the past will continue to influence the future

Naïve method:
– The forecast is equal to the actual value observed during the last period
– good for level patterns

Ft 1  Dt
Moving Average method:
– The average value over a set time period (e.g.: the last three weeks)
– Each new forecast drops the oldest data point & adds a new observation
– More responsive to a trend but still lags behind actual data

Ft1   Dt / n
Simple Moving Average

For Three period Moving Average

F4 = M3

F5 = M4

For Five period Moving Average

F6 = M5

F7 = M6

Where
n = number of periods taken to evaluate the moving average
Dt or Di = Actual demand in that period
 Example 1: Forecast the order for the month of November by
i. Naive Approach
ii. a three period moving average .
iii. a five period moving average.
Month
Jan Feb Mar Apr May Jun Jul Aug Sep Oct
(t)
Order per
month 120 90 100 75 110 50 75 130 110 90
Dt

Solution:

i. Naive Approach
Order per
Month Forecast Ft
S. No. month
(t)
Dt
1 Jan 120
2 Feb 90 120
3 Mar 100 90
4 Apr 75 100
5 May 110 75
6 Jun 50 110
7 Jul 75 50
8 Aug 130 75
9 Sep 110 130
10 Oct 90 110
11 Nov 90

F2 = 120 Ft1  Dt
F3 = 90

F4 = 100
.

F11 = 90

Forecast the order for the month of November = 90

3- Months Moving Average (n=3)
Order per Moving
Month Forecast F Error
S.No. month Average t
(t) (Dt - Ft)
Dt Mt

1 Jan 120 ------ ------ ------

2 Feb 90 ------ ------ ------
3 Mar 100 103 ------ ------
4 Apr 75 88 103 -28
5 May 110 95 88 22
6 Jun 50 78 95 -45
7 Jul 75 78 78 -3
8 Aug 130 85 78 52
9 Sep 110 105 85 25
10 Oct 90 110 105 -15
11 Nov ------ ------ 110 ------

M3 = (D1+D2+D3) /3
= (120+90+100) /3
= 103

F4 = M3 = 103

M4 = 88

.F5 = M4 = 88
.
.
.
.M10 = 110

F11 = M10 = 110

Forecast the order for the month of November = 110

5- Months Moving Average (n=5)

Order per Moving

Month Forecast F Error
S.No. month Average t
(t) (Dt - Ft)
Dt Mt
1 Jan 120 ------ ------ ------
2 Feb 90 ------ ------ ------
3 Mar 100 ------ ------ ------
4 Apr 75 ------ ------ ------
5 May 110 99 ------ ------
6 Jun 50 85 99 -49
7 Jul 75 82 85 -10
8 Aug 130 88 82 48
9 Sep 110 95 88 22
10 Oct 90 91 95 -5
11 Nov ------ ----- 91 ------

M5 = (D1+ D2+D3+D4+D5) /5
= (120+90+100+75+110)/5
= 99

F6 = M5 = 99

M6 = 85

F7 = M6 = 85
.
.
.
M10 = (50+75+130+110+90)/5
= 91
F11 = M10 = 91

Forecast the order for the month of November = 91

 Example 2 : During the past ten weeks, sales of cases of Comfort brand headache medicine at
Robert's Drugs have been as follows. Forecast the sales for period 11 using

i) a three period moving average .

ii) a five period moving average.
Week 1 2 3 4 5 6 7 8 9 10

Sales 110 115 125 120 125 120 130 115 110 130

Solution:
i) 3 Period Moving Average (n=3)

Actual Moving Forecast

Week Error
Demand Average F
(t) (Dt - Ft)
Dt Mt t

1 110 ------ ------ ------

2 115 ------ ------ ------
3 125 117 ------ ------
4 120 120 117 3
5 125 123 120 5
6 120 122 123 -3
7 130 125 122 8
8 115 122 125 -10
9 110 118 122 -12
10 130 118 118 12
11 ------ ------ 118 ------

M3 = (D1+D2+D3) /3
= (110+115+125) /3
= 117
F4 = M3 = 117
.
.
M10 = 118
F11 = M10 = 118
Sales for the week 11 = 118
ii) 5 Period Moving Average (n=5)

Actual Moving
Week Forecast F Error
Demand Average t
(t) (Dt - Ft)
Dt Mt
1 110 ------ ------ ------
2 115 ------ ------ ------
3 125 ------ ------ ------
4 120 ------ ------ ------
5 125 124 ------ ------
6 120 122 124 -4
7 130 120 122 8
8 115 121 120 -5
9 110 121 121 -11
10 130 118 121 9
11 ------ ------ 118 -----

M5 = (D1+ D2+D3+D4+D5) /5
= (110+115+125+120+125)/5
= 124
F6 = M5 = 124
.

M10 = 118
F11 = M10 = 118
Sales for the week 11 = 118
Weighted Moving Average:

W D t t
Mt  t 1
n

W
t 1
t

Three period Weighted moving average

W 1D1  W 2 D 2  W 3D3
M3 
W1W 2 W 3
F4 = M3

F5 = M4

Example 3: Use a 3 period weighted moving average to forecast the sales for week 11 giving a
weight of 0.6 to the most recent period, 0.3 to the second most recent period, and 0.1 to the third
most recent period

Week 1 2 3 4 5 6 7 8 9 10
Sales 110 115 125 120 125 120 130 115 110 130

Solution:

Weighted
Actual Forecast
Week Moving Error
Demand F
(t) Average (Dt - Ft)
Dt t
Mt
1 110 ------ ------ ------
2 115 ------ ------ ------
3 125 121 ------ ------
4 120 121 121 -1
5 125 124 121 4
6 120 122 124 -4
7 130 127 122 8
8 115 120 127 -12
9 110 114 120 -10
10 130 123 114 17
11 ------ ------ 123 -----
W1 =0.1, W2= 0.3 and W3=0.6

W 1D1 W 2 D2 W 3D3
M3
W 1 W 2 W 3

0.1x110  0.3x115  0.6x125

M 3 
0.1 0.3  0.6

= 121

F4 = M3 = 121

F5 = M4 = 121

F11 = M10 = 123

Sales for the week 11 = 123

Exponential Smoothing method

• Assumes the most recent observations have the highest predictive value
– gives more weight to recent time periods

Ft+1 = Forecast value for time t+1

Dt = Actual value at time t
α = Smoothing constant
Example 4 : Using the exponential smoothing technique, compute the forecasts from the
following data (time series) under the situations when α = 0.1 and 0.6. Compute the forecast for
the 10th period?

Month 1 2 3 4 5 6 7 8 9
Demand 820 775 680 655 750 802 798 689 775

Solution:

Assume F1=D1

Forecast Forecast
Week Demand a=0.1 a=0.6
1 820 820.00 820.00
2 775 820.00 820.00
3 680 815.50 793.00
4 655 801.95 725.20
5 750 787.26 683.08
6 802 783.53 723.23
7 798 785.38 770.49
8 689 786.64 787.00
9 775 776.88 728.20
10 776.69 756.28

Ft+1 = Ft + a(D t - Ft)

For a=0.1
F2 = F1+ a(D1 - F1)
=820+0.1(820-820)
=820.00
 F3 = F2+ a (D2 – F2)
=820+0.1 (775-820)
= 815.5
.
.
F10 = F9 + a (D9 – F9)
= 776.88+0.1 (775-776.88)
= 776.69
Forecast for the 10th period when a is 0.1 =776.69

For a=0.6

F2 = F1+ a(D1 - F1)

=820+0.6(820-820)
=820.00
F3 = F2+ a (D2 – F2)
=820+0.6 (775-820)
= 793
.
.
F10 = F9 + a (D9 – F9)
= 728+0.6 (775-728)
= 756.28
Forecast for the 10th period when a is 0.6 =756.28
Example 5 : Using the exponential smoothing technique, Compute the forecasts from the
following data (time series) under the situations when α = 0.3 and 0.6. Compute the forecast for
the month of july ? Assume F1=100.00

Month January February March April May June

Demand 120 90 101 91 115 83

Solution:

Forecast Forecast
S.No. Month Demand
a = 0.3 a = 0.6
1 January 120 100.00 100.00
2 February 90 106.00 112.00
3 March 101 101.20 98.80
4 April 91 101.14 100.12
5 May 115 98.10 94.65
6 June 83 103.17 106.86
7 July 97.12 92.54

Ft+1 = Ft + a(D t - Ft)

For a = 0.3 and Assuming F1=100.00
F2 = F1+ a(D1 - F1)
=100+0.3(120-100)
=106.00
F3 = F2+ a (D2 – F2)
=106+0.3(90-106)
=101.2
.
.
F7 = F6 + a (D6 – F6)
= 103.17+0.3(83-103.17)
= 97.12
Forecast for the month of july when a is 0.3 =97.12
For a = 0.6 and Assuming F1=100.00
F2 = F1+ a(D1 - F1)
=100+0.6(120-100)
=112.00
F3 = F2+ a (D2 – F2)
=112+0.6(90-112)
= 98.8
.
.
F7 = F6 + a (D6 – F6)
= 106.86+0.6(83-106.86)
= 92.54
Forecast for the month of july when a is 0.3 = 92.54
Linear Regression
A time series technique that computes a forecast with trend by drawing a straight line through a
set of data using this formula:
Y = a + bX where
Y = Dependent variable (forecast for X)
X = Independent variable
a = value of y at X = 0 (Y intercept)
b = slope of the line
• Solve for the slope of the line

b
XY  nXY
 X2  nX
2

• Solve for the y intercept

a  Y  bX
• Develop the equation for the trend line Y=a + bX

Example 6: 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.

Sales $ 130 151 150 158

Adv.$ 32 52 50 55
Solution:

Adv.$ Sales $
S.No. XY X^2
(X) (Y)
1 32 130 4160 2304
2 52 151 7852 2704
3 50 150 7500 2500
4 55 158 8690 3025
5 53
Tot 189 589 28202 9253
 189
X  47.25
4
589
Y  147.25
4

b
XY  nXY
 X2  nX
2

28202  447.25147.25
b  1.15
9253  447.25
2

a  Y  bX  147.25 1.1547.25
a  92.9

Y  a  bX
 92.9 1.15X
Y  92.9 1.1553
 153.85
Correlation Coefficient:

• 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 2    X * n  Y 2    Y 
2 2

428,202189589
r  .982
487,165  589 
2 2
4(9253) - (189) *
r 2  .982 2  .964

Coefficient of determination ( r2 ) measures the amount of variation in the dependent variable

about its mean that is explained by the regression line. Values of ( r2 ) close to 1.0 are desirable
Example 7: A firm believes that its annual profit depends on its expenditures for research. The
information for the preceding six years is given below. Estimate the profit when expenditure is 6
units.

Year 1 2 3 4 5 6
Expenditure
2 3 5 4 11 5
for research
Annual profit 20 25 34 30 40 31

Solution:

Expenditure for Annual

2
Year research profit XY X
X Y
1 2 20 40 4
2 3 25 75 9
3 5 34 170 25
4 4 30 120 16
5 11 40 440 121
6 5 31 155 25
7 6 ?
Total 30 180 1000 200

b
XY  nXY
 X 2  nX
2

30
X 5 1000  6 x 5 x 30
6 b  2
200  6 x 5
2

180
Y  30 a  Y  bX  30  2 x 5
6
a  20
Y  a  bX
 20  2X
Y  20  2 x 6
 32
Measuring Forecast Error

• Forecasts are never perfect

• Need to know how much we should rely on our chosen forecasting method

Measuring Forecast Error

E t  Dt  Ft
• Note that over-forecasts = negative errors and under-forecasts = positive errors

Measuring Forecasting Accuracy

 n
i) Mean Absolute Deviation (MAD)
 D -F
t=1
t t

Measures the total error in a forecast without regard to sign MAD =

n
ii) Cumulative Forecast Error (CFE)

Measures any bias in the forecast

CFE  actual  forecast 

iii) Mean Square Error (MSE)
 D
n
t - F 2
Penalizes larger errors t
t=1
MSE =
n
iv) Root Mean Squared Error (RMSE)

RMSE  MSE
v) Mean Absolute Percentage Error (MAPE)
n

 D -F
t=1
t t
MAPE = X100
Dt
vi) Mean forecast error (MFE)
n

MFE =
 (D  F )
t=1
t t

n
vii) Tracking Signal

Measures if your model is working

CFE
TS 
MAD
Example 8: Determine the Mean Absolute Deviation (MAD), Mean Square Error (MSE), Mean
Absolute Percentage Error (MAPE), Mean forecast Error (MFE) for the following Data

Period 1 2 3 4 5
Demand 150 160 165 175 180
Forecast 165 165 165 165 165

Solution:

Absolute Squared Percentage Absolute

Demand Forecast Error Error
Period deviation Error Percentage Error
D F (D -F ) |D – F |
2
(D -F ) *100/D |D – F | *100/|D |
t t t t
t t
(D – F ) t t t t t t
t t

1 150 165 -15 15 225 -10 10

2 160 165 -5 5 25 -3.125 3.125

3 165 165 0 0 0 0 0

4 175 165 10 10 100 5.71 5.71

5 180 165 15 15 225 8.33 8.33

TOTAL 5 45 575 27.165

D t - Ft

45
9
t=1
MAD =
n 5

 D t - Ft 2

575
 115
t=1
MSE =
n 5

1 n 27.165
Dt - Ft   5.433%
n 
MAPE  X100
t 1 Dt 5

 (D  F ) t t 5
MFE = t=1
n
  1
5
Example 9: Determine the MAD, MSE and RMSE values for the given forecast values in the
table below?

Month 1 2 3 4 5
Sales 220 250 210 300 325
Forecast N/A 255 205 320 315
Solution:

Absolute Squared
Month Sales Forecast Error deviation Error
|D – F | 2
D F (D -F ) t t (D – F )
t t t t
t t

1 220 N/A --- --- ----

2 250 255 -5 5 25

3 210 205 5 5 25

4 300 320 -20 20 400

5 325 315 10 10 100

TOTAL 40 550

D
n
t - Ft 40
t=1
  10
MAD = 4
n
n

 D t - Ft 
2
550
MSE = t=1   137.5
n 4

RMSE  MSE  137.5  11.73

18