FORECASTING

Introduction

In simple understanding forecasting means the prediction or estimation of future

events. Production and operating management is the transformation process
designed to convert input into the outputs. This process can work efficiently at that
condition when the operations manager knows the customers need and how much
is the demand. All activities start with the forecast. It influences the long term
planning of facilities, capacities, locations and layout.
Forecasting is the process by which companies consider and prepare for the future.
It involves predicting the future outcome of various business decisions. This
includes the future of the business as a whole, the future of an existing or proposed
product or product line, and the future of the industry in which the business
operates, to name a few.. Prediction is a similar to forecasting, but it is more
general term. Both might refer to formal statistical methods employing time
series, cross-sectional or longitudinal data, or alternatively to less formal
judgmental methods. Usage can differ between areas of application: for example,
in hydrology, the terms "forecast" and "forecasting" are sometimes reserved for
estimates of values at certain specific future times, while the term "prediction" is
used for more general estimates, such as the number of times floods will occur
over a long period.
Forecast is only the assumption of future events. So it can never be perfect.
However, it does not mean that we do not make any forecasts. Operations
managers try to forecast wide range of future events that potentially affect success.
Most often the concern is forecasting customer demand for products or services.
Managers may want long ran estimates of overall demand or short run estimates of
demand for each individual product. Even more details estimates are needed for
specific items or sub-components that go into each product.
 FORECASTING CHAPTER - 20 31

Elements of good forecast

We can distinguish among these different kinds of forecasting needs by

considering how far into the future they focus. Detailed forecasts for individual
items are used to plan the short run use of conversion system. At the other
extreme, overall or aggregate product demand forecasts are used to plan for
capacity, location and layout over much longer time span. Risk and uncertainty are
central to forecasting and prediction; it is generally considered good practice to
indicate the degree of uncertainty attaching to forecasts. In any case, the data must
be up to date in order for the forecast to be as accurate as possible.
The elements of doing business are constantly changing. Interest rates rise and fall,
customer preferences change, suppliers go out of business, and the list goes on and
on. As the factors that affect business change, the company forecasts must also
change. In this time of rapid change, accurate and timely forecasts have become
even more important.
Financial forecasting is important for several reasons. First, it enables management
to change operations at the right time in order to reap the greatest benefit. It also
helps the company prevent losses by making the proper decisions based on
relevant information. Organizations that can create high quality and accurate
forecasts are able to see what interventions are required to meet their business
performance targets.
Forecasting is also important when it comes to developing new products or new
product lines. It helps management decide whether the product or product line
will be successful. Forecasting prevents the company from spending time and
money developing, manufacturing, and marketing a product that will fail.
Stockholder expectations highlight another reason behind the importance of
forecasting. Public companies experience scrutiny and pressure for short-term
performance from investors. Operational results will be examined by investors and
investment analysts, and actual results that differ from forecasts will be bad for the
company and its stock price. This is because both meeting predictions and
exceeding predictions will reduce investor confidence. This will cause investors to
believe that the company does not understand its own business model.
Besides these, the significance or importance of forecasting involves following
points
 It provides relevant and reliable information about the past and present events
and the likely future events. This is necessary for sound planning
 It is the basis for making planning premises
 It gives confidence to the managers for making important decisions
32 / OPERATIONS MANAGEMENT

The process of climate change and increasing energy prices has led to the usage
of Forecasting. The method uses forecasting to reduce the energy needed to heat
the building, thus reducing the emission of greenhouse gases. Forecasting is used
in the practice of Customer Demand Planning in everyday business forecasting for
manufacturing companies. Forecasting has also been used to predict the
development of conflict situations. Experts in forecasting perform research that
use empirical results to gauge the effectiveness of certain forecasting models.
Research has shown that there is little difference between the accuracy of forecasts
performed by experts knowledgeable of the conflict situation of interest and that
performed by individuals who knew much less.
Similarly, experts in some studies argue that role thinking does not contribute to
the accuracy of the forecast. The discipline of demand planning, also sometimes
referred to as supply chain forecasting, embraces both statistical forecasting and a
consensus process. An important, albeit often ignored aspect of forecasting, is the
relationship it holds with planning. Forecasting can be described as predicting
what the future will look like, whereas planning predicts what the
future should look like. There is no single right forecasting method to use.
Forecasting can be used in Supply Chain Management as well to make sure that the
right product is at the right place at the right time. Accurate forecasting will help
retailers reduce excess inventory and therefore increase profit margin. Studies
have shown that extrapolations are the least accurate, while company earnings
forecasts are the most reliable. Accurate forecasting will also help them meet
consumer demand.

Steps in Forecasting process

Following steps are involved in forecasting Procedure:-

 Analysing and understanding the problem The manager must first identify
the real problem for which the forecast is to be made. This will help the
manager to fix the scope of forecasting.
 Developing sound foundation: The management can develop a sound
foundation, for the future after considering available information experience,
type of business, and the rate of development.
 Collecting and analysing data: Data collection is time consuming. Only
relevant data must be kept. Many statistical tools can be used to analyse the
data.
 FORECASTING CHAPTER - 20 33

 Estimating future events: The future events are estimated by using trend
analysis. Trend analysis makes provision for some errors.
 Comparing results: The actual results are compared with the estimated results.
If the actual results tally with the estimated results, there is nothing to worry. In
case of any major difference between the actuals and the estimates, it is
necessary to find out the reasons for poor performance.
 Follow up action: The forecasting process can be continuously improved and
refined on the basis of past experience. Areas of weaknesses can be improved
for the future forecasting. There must be regular feedback on past forecasting.

Forecasting and Operations Sub-System

The aggregate demand forecast is normally obtained by estimating expected values

of sales and production units. The proper resource allocation to each production
units leads the organization to the approx of success. Resource forecasts are used
to plan and control operation sub-systems as shown in figure below:

Information on most recent

demand and production

Demand Forecast
for operation

Planning the system Scheduling the system Controlling the system

(designing)
– Product design – Aggregate production – Producing control
– Process design planning – Inventory control
– Equipment and – Operations scheduling – Labour control
investment replacement – Cost control
– Capacity planning

Output of goods and

services

Feedback

Figure 10.1: Demand forecasting and operations subsystem.

34 / OPERATIONS MANAGEMENT

Above figures shows that there is inter relationship between planning, scheduling
and controlling the system and they are dependent on demand forecast. The
output of goods and services is obtained from planning, scheduling and controlling
system. Also the feedback system provides the opportunity to improve the error in
forecasting.

Approaches to Forecasting

The techniques by which we can forecast future events are called forecasting
techniques and these techniques may be qualitative and quantitative. Qualitative
forecasting techniques are subjective, based on the opinion and judgment of
consumers, experts; appropriate when past data is not available. It is usually
applied to intermediate-long range decisions. Examples of Judgmental and opinion
forecasting methods are informed opinion and judgment, the Delphi
method, market research, historical life-cycle analogy.
Quantitative forecasting models are used to estimate future demands as a function
of past data; appropriate when past data are available. The method is usually
applied to short-intermediate range decisions. Examples of quantitative
forecasting methods are last period demand, simple and weighted moving
averages, simple exponential smoothing, and trend projection method.

(i) Qualitative Techniques of Forecasting: Opinion and Judgmental

It is the subjective techniques of forecasting on which forecasted value may differ
according to the group of forecasters. When the forecasting time horizon is very
large and no numerical description can be formulated then this technique is
generally applied to forecast the future events. Quantitative forecasting techniques
fall into following categories:
 Opinion of jury
 Sales force composite
 Consumer market survey
 Delphi technique
a. Opinion of jury: It is the simplest method of forecasting on which a group of
juries from upper level managers is made. It is often used as a part of long run
planning and new product development process. This exercise usually leads to
a quicker (and often more reliable) result without the use of
elaborate data manipulation and statistical techniques.
 FORECASTING CHAPTER - 20 35

b. Sales force composite: The sales forces are always regarded as the good
source of information due to their direct attachment with consumers. What
consumers want can be viewed by the sales persons only in real sense, because
they are often aware about the choice and future plans of consumers.
Therefore, the view of sales persons is reviewed to make a proper forecast.
c. Consumer market survey: The demand of goods depends upon the taste and
need of consumers. So, the product should be introduced into the market as
per the taste and need of consumers. Every customer is the potential
customers of every product. Therefore, a great deal of case is necessary to
conduct or consumer market survey. After conducting consumer market
survey, the manager can predict what type and size of goods can be sold in
future market.
Survey methodology seeks to identify principles about the design, collection,
processing, and analysis of surveys in connection to the cost and quality of
survey estimates. It focuses on improving quality within cost constraints, or
alternatively, reducing costs for a fixed level of quality. Survey methodology is
both a scientific field and a profession. Part of the task of a survey
methodologist is making a large set of decisions about thousands of individual
features of a survey in order to improve it.
d. Delphi technique: It is the most popular technique particularly for
technological forecasting. It is a structured communication technique,
originally developed as a systematic, interactive forecasting method which
relies on a panel of experts. In the standard version, the experts answer
questionnaires in two or more rounds. After each round, a facilitator provides
an anonymous summary of the experts’ forecasts from the previous round as
well as the reasons they provided for their judgments. Thus, experts are
encouraged to revise their earlier answers in light of the replies of other
members of their panel. It is believed that during this process the range of the
answers will decrease and the group will converge towards the "correct"
answer. Finally, the process is stopped after a pre-defined stop criterion (e.g.
number of rounds, achievement of consensus, and stability of results) and
the mean or median scores of the final rounds determine the results.
The name "Delphi" derives from the Oracle of Delphi. The authors of this
method were not happy with this name, because it implies "something
oracular, something smacking a little of the occult". The Delphi method is
based on the assumption that group judgments are more valid than individual
judgments. The Delphi method was developed at the beginning of the Cold
36 / OPERATIONS MANAGEMENT

War to forecast the impact of technology on warfare. In 1944, General Henry

H. Arnold ordered the creation of the report for the U.S. Army Air Corps on the
future technological capabilities that might be used by the military. Different
approaches were tried, but the shortcomings of
traditional forecasting methods, such as theoretical approach, quantitative
models or trend extrapolation, in areas where precise scientific laws have not
been established yet, quickly became apparent. To combat theses
shortcomings, the Delphi method was developed by Project RAND during the
1950-1960s by Olaf Helmer, Norman Dalkey, and Nicholas Rescher. It has been
used ever since, together with various modifications and reformulations, such
as the Imen Delphi procedure.
(i) Quantitative Technique of Forecasting
It is the objective technique of forecasting which numerical description can be
formulated to forecast the future event. There are two models under the
quantitative technique of forecasting.
 Time series model
 Casual model
Time series model:
A time series is an arrangement of statistical data in chronological order i.e. in
accordance with its time of occurrence. Time series data have a natural temporal
ordering. This makes time series analysis distinct from other common data
analysis problems, in which there is no natural ordering of the observations. Time
series analysis is also distinct from spatial data analysis where the observations
typically relate to geographical locations. In addition, time series models will often
make use of the natural one-way ordering of time so that values for a given period
will be expressed as deriving in some way from past values, rather than from
future values. Under this model, time period is considered as the focal point of
forecasting. There are four approaches to forecast future event under time series
model.
 Naïve approach
 Moving average method
 Experiential smoothing
 Trend projection
 FORECASTING CHAPTER - 20 37

1. Naïve Approach
It is the simplest method of forecasting on which forecasted value of one period is
equal to the actual value of most recent period (i.e. just preceding period). It is the
most cost-effective and efficient objective forecasting model, and provide a
benchmark against which more sophisticated models can be compared. In order to
clarify the naïve approach, we take following example.
Example 1
Months January February March April
Sales (Rs 12 10 15 18
000)
The forecasted sales in May or June = Rs 18,000.

2. Moving Average Method

Naïve approach does not trace out all periods data. So, the forecasted value of
naïve approach may not be proper. In fact under the naïve approach only a single
value is used to forecast which is taken from the previous period. But under the
moving method we should calculate the moving average value with chosen number
of periods and the forecasted value is equal to the moving average value. It means
the trend line is found by the smoothing out the fluctuation of the data by means of
moving averages. This method does not require any mathematical skill to find the
trend value. Calculation process is simple but tedious. This method is more
effective when the trend of time series is more irregular. This method is not
applicable for the prediction. This method carries all the merits and demerits of the
arithmetic mean. There are two methods to calculate moving average value. They
are as follows:
 Simple moving average (SMA)
 Weighted moving average (WMA)
Simple moving average (SMA)
In financial applications a simple moving average (SMA) is the un-weighted mean
taken from an equal number of data either side of a central value. This ensures that
variations in the mean are aligned with the variations in the data rather than being
shifted in time. It combines the value of chosen number of periods and takes the
simple average value. It is commonly used with time series data to smooth out
short-term fluctuations and highlight longer-term trends or cycles.
The entrance between short-term and long-term depends on the application, and
the parameters of the moving average will be set accordingly. For example, it is
38 / OPERATIONS MANAGEMENT

often used in technical analysis of financial data, like stock prices, returns or
trading volumes. It is also used in economics to examine gross domestic product,
employment or other macroeconomic time series. Mathematically, a moving
average is a type of convolution and so it can be viewed as an example of a low-
pass filter used in signal processing. When used with non-time series data, a
moving average filters higher frequency components without any specific
connection to time, although typically some kind of ordering is implied. A simple
moving average value can be calculated as:
Sum of value for choosen periods
Simple moving average (SMA) = Choosen number of periods
Example 2 You are given the sales information of Parichaya bakery of
Bhaktapur.
Months Jan. Feb. March Apr. May June
Sales 8 14 10 15 16 14
(000)
Required:
a. Forecast for July using 6 months moving average.
b. Forecast for July using 3 month moving average.
c. Forecast for August using 4 month moving average.
Solution
a. Forecasting for July using 6 months moving average.
Sales of January to June 8 + 14 + 10 + 15 + 16 + 14 83
SMA = 6 = 6 = 6 = 13.83
The forecasted sales of July is Rs 13.83 thousand.
b. Forecasting for July using 3 month moving average.
Sales of April to June 15 + 16 + 14
SMA = 3 = 3 = 15
The forecasted sales of July is Rs 15 thousands.
c. Forecasting for August using 4 month moving average.
Sales of April to July 15 + 16 + 14 + 15
SMA = 4 = 4 = 15
The forecasted sales for August is Rs 15 thousands.

Weighted Moving Average (WMA)

While computing simple moving average value it is assumed that all 'n' periods are
equally important but in practice it may not be so. If one periods value is more
important than that of another's, then that value should be discriminated by
 FORECASTING CHAPTER - 20 39

assigning some weights and the moving average value with respect to the
corresponding weights moving average (WMA).
A weighted average is any average that has multiplying factors to give different
weights to data at different positions in the sample window. Mathematically, the
moving average is the convolution of the datum points with a fixed weighting
function. In technical analysis of financial data, a weighted moving average (WMA)
has the specific meaning of weights that decrease in arithmetical progression. The
moving average value is calculated as:
SDW
WMA = SW

Where,
W = weight or proportion of demand
D = demand of specific periods
[Note: If the probabi
must be one.]
Example 3 The sales of June, July and August are Rs 18,000, 20,000 and 16,000
respectively. There is one third chance of selling in June and one fourth chance of
selling in August. Find the forecasted sales in September.
Solution
Computation table
Month Sales (D) Probability (W) DW
June 18,000 1/3 6,000
July 20,000 5 / 12 (rest) 8,333
August 16,000 1/4 4,000

Now,
SDW 18‚333
WMA = SW = 1 = Rs 18,333
The forecasted sales in September is Rs 18,333.

3. Exponential Smoothing
Exponential smoothing is a technique that can be applied to time series data, either
to produce smoothed data for presentation, or to make forecasts. It is the most
popular statistical forecasting methods for its simplicity and low cost. It requires
less data memory storage and has fast computational speed. Since the late 1950s,
various exponential smoothing models have been developed to cope with various
40 / OPERATIONS MANAGEMENT

types of time series data. For example, time series data with trend, seasonality, and
other underlying patterns.
Exponential smoothing is the revised form of weighted moving average method.
Under this method a single weight called alpha (i.e. α) is used instead of various
weights of respective time periods. It is commonly applied to financial market and
economic data, but it can be used with any discrete set of repeated measurements.
The forecasted value as per exponential smoothing is given by the formula below:
^
yt+1 = ^
yt ^
t - y t)
Where,
yt = actual value for current period t
^
yt = forecasted value for current period t
^
yt+1 = forecasted value for next period t+1.
2
= n+1 (if not given)
n = number of periods in moving average.
In order to clarify the formula of exponential smoothing, we take same specific
periods instead of t and t + 1 as:
^y2005 = ^
y 2004 ^
2004 -y 2004)
^y2006 = ^
y 2005 ^
2005 - y 2005)
^
y2007 = ^
y 2006 -^
y 2006) and so on . . .
2006

Example 4 Demand for a certain item for January, February and March are 1000
units, 1500 units and 800 units respectively. The forecasted demand for January is
2000 units. Forecast the demand in April with smoothing factor 0.20.
Solution
Given that
Months Actual demand Forecasted Smoothing factor
(units) demand (units)
January 1000 2000 0.20
February 1500 – 0.20
March 800 – 0.20
Forecasted demand for April =?
In order to forecast the Aprils demand we must have the forecasted demand of
March. So following calculations should be made before the actual requirement.
 FORECASTING CHAPTER - 20 41

^
yFeb. =^
y Jan. -^
yJan.) = 2000 + 0.20 (1000 – 2000) = 12000 – 200 =
Jan
1800 units
^
yMarch = ^
y Feb -^
yFeb) = 1800 + 0.20 (1500 – 1800) = 1800 – 60 =
Feb.
1740 units
^
yApril = ^
y March -^
y March) = 1740 + 0.20 (800 – 1740) = 1740 – 188
March
= 1552 units
The forecasted demand for April is 1552 units.

4. Trend Projection Method

Trend projection method is a classical method of business forecasting. This method
is essentially concerned with the study of movement of variable through time. The
use of this method requires a long and reliable time series data. The trend
projection method is used under the assumption that the factors responsible for
the past trends in variables to be projected (e.g. sales and demand) will continue to
play their part in future in the same manner and to the same extend as they did in
the past in determining the magnitude and direction of the variable.
There are different methods of forecasting under trend projection method and
some of the important are as follows:
 Graphic or free hand curve method
 Semi average method
 Least square method
Graphic or free hand curve method
It is the simplest method on which no numerical calculation is required. Under this
method we draw a graph taking time period along with X-axis and variables along
with Y-axis. Then we plot the given data into the graph with dotted points and
finally divide those points by a straight line maintaining the equal number of dots
below and above that straight line. The obtained straight line is taken as trend line
which helps to forecast the future value. But this method provides more accurate
solution at that condition when we use computer because it has no mathematical
basis.
Example 5 Fit a trend line to the following data by graphic method.
Year 2002 2003 2004 2005 2006 2007 2008
Sales ‘000’ 59 58 64 60 66 65 68
42 / OPERATIONS MANAGEMENT

Solution:
The graph showing trend line and original line of profit by graphic method.

68  Trend line

67
66  Observed line

65 
64 
Sales (000)

63
62
61
60 
59 
58 

2002 2003 2004 2005 2006 2007 2008

Year

As per above figure there is increasing trend in sales.

Semi – average method
This method is slightly improved than the freehand curve method. This method is
based on the arithmetic mean, so this method carries bad and good sides of the A.M.
There are very few cases where this method is used for forecasting. following steps
are involved under this method:
 Divide the series into two equal parts
 Find the A.M. of each part separately
 Plot the average of each part against the middle of time period covered by each
part
 Join the plotted points and see results
[Note: If there are odd number of observations, then avoid middle year’s data]
Example 6 Fit the trend line from the following data by semi average method.
Also forecast the value of production at 2012.
Year 2005 200 2007 2008 200 2010
6 9
Production (in 10 17 12 15 20 22
ton)
 FORECASTING CHAPTER - 20 43

Solution:
Computation table
Year Production(in Semi- total Semi- average
ton)
2005 10
2006 17 39 13
2007 12
2008 15
2009 20 57 19
2010 22
Graphical representation of trend and observed line

26 Trend line
24
22
Production (in tons)

20

18
16
14

12
10
8
6

2005 2006 2007 2008 2009 2010 2011 2012

Year

As per above figure there is increasing trend in production. Also the forecasted
value of production in 2012 is 25 tons.
Example 7 Plot the following data and find the trend line by the method of semi-
average.
Year 2003 2004 2005 2006 2007
Sales '000 10 6 14 16 8
Solution
Here, number of years n = 5 odd so we should avoid the sales of middle year 2005
and then we should take an average of the first three and last three sales values.
44 / OPERATIONS MANAGEMENT

Computation table
Year Sales’000’ Semi- total Semi- average
2003 10
16 8
2004 6
2005 14
2006 16
24 12
2007 8
Graph showing the trend line by semi average method.

16 
14 
Trend line
12 
Sales (000)

10  Observed line
8  
6 
4
2

2003 2004 2005 2006 2007

Year

As per above figure there is increasing trend in sales.

Least Square Method

The most widely used method to describe the trend is the method of least square.
It is the best measure of trend for forecasting. The most important application is
in data fitting. The best fit in the least-squares sense minimizes the sum of
squared residuals, a residual being the difference between an observed value and
the fitted value provided by a model. This method is completely based on the
mathematical procedure and on the principle that sum of square of deviation
between the original data and the computed trend values is minimum. This
method gives the straight line to compute the trend values.
This technique fits a trend line to a series of historical data. It is the best method
for forecasting future events which is based on the value of all time periods. Under
this, we draw a linear equation as:
y = a + bx …………… (i)
 FORECASTING CHAPTER - 20 45

Where,
y = dependent variable
a = y-intercept (or constant value)
b = slope of trend or growth rate
x = deviated value of time period
= (T – middle year) → If n is odd
or = 2(T – Middle year) → If n is even
In order to obtain the value of trend parameters 'a' and 'b', we have to solve
following two normal equations.
y = na + bx
xy = ax + bx2
By solving above equations, we get following formula to obtain the value of 'a' and
nSxy - Sx Sy Sy - bSx –
'b' b = nSx2 - (Sx)2 ; a= n = y – b –x

Sy Sxy
a= n ; b = Sx2

[Note: if x = (T-middle year), then b represents the annual growth rate and if x
= 2(T-middle year), then b represents the semi-annual growth]

Example 8 Actual development expenditure on agriculture sector in sixth and

seventh development plans are as below:
Year Expenditure in Agri. Sector (Rs. in
millions)
1980- 81 7
1981 - 82 11
1982 - 83 15
1983 - 84 14
1984 - 85 17
1985 - 86 21
1986 - 87 20
1987 - 88 23
1988 - 89 33
1989 – 90 30
Total 191
46 / OPERATIONS MANAGEMENT

Fit the line of best fit and forecast the expenditure in agricultural sector for the
year 1990 - 91 and 1992 - 93.
Solution:
Computation table
Year T(let) Expenditure X=2(T- x2 Xy
(y) 1984.5)
1980 - 1980 7 -9 81 - 63
81
1981 - 1981 11 -7 49 - 77
82
1982 - 1982 15 -5 25 - 75
83
1983 - 1983 14 -3 9 - 42
84
1984 - 1984 17 -1 1 - 17
85
1985 - 1985 21 1 1 21
86
1986 - 1986 20 3 9 60
87
1987 - 1987 23 5 25 115
88
1988 - 1988 33 7 49 231
89
1989 - 1989 30 9 81 270
90
∑y = 191 ∑x = 0 ∑x2 = ∑xy =
330 423
The trend line that best fits the given data is given by:
y = a + bx ………….……(i)
Since x = 0, so the value of a and b can be obtained as:
∑y 191
a = n = 10 = 19.1 which is Y- intercept.
∑xy 423
b = ∑x2 = 330 = 1.28 which is semi annual growth rate.
From equation (i), we get
^ = 19.1 + 1.28x.
Y
Value for the year 1990 - 91,
We have, T= 1990
 FORECASTING CHAPTER - 20 47

x = 2(1990-1984.5) = 11
^ = 19.1 + 1.28 × 11 = 33.18
 Y
Value for the year 1992 - 93,
We have, T= 1992
x = 2(1992-1984.5) = 15
^ = 19.1 + 1.28 × 15 = 38.3
 Y
Thus the estimated expenditure for 1990 - 91 and 1992 - 93 are 33.18 and 38.3
million rupees respectively.

Casual Model
Casual quantitative model establishes the relationship between the different
variables and their effect on each other. It is the most sophisticated method of
forecasting. Following two methods are involved under the casual model of
forecasting.
 Regression analysis
 Econometric modeling
Regression analysis
Regression analysis can be defined as the statistical tools used to estimate or
predict the value of unknown dependent variable with the given value
independent variable. It shows the relationship between dependent and
independent variables. It also shows the cause and effect analysis between
variables. More specifically, regression analysis helps to understand how the
typical value of the dependent variable changes when any one of the independent
variables is varied, while the other independent variables are held fixed. Most
commonly, regression analysis estimates the conditional expectation of the
dependent variable given the independent variables that is, the average value of
the dependent variable when the independent variables are fixed. In regression
analysis, it is also of interest to characterize the variation of the dependent variable
around the regression function, which can be described by a probability
distribution.
Regression analysis is widely used for prediction and forecasting, where its use has
substantial overlap with the field of machine learning. Regression analysis is also
used to understand which among the independent variables are related to the
dependent variable, and to explore the forms of these relationships. In restricted
circumstances, regression analysis can be used to infer causal
relationships between the independent and dependent variables. However this can
lead to illusions or false relationships, so caution is advisable.
48 / OPERATIONS MANAGEMENT

In order to obtain the estimated (i.e. forecasted) value of dependent variable, we

should draw the linear regression equation which is also known as line of linear
regression. On the basis of number of variables, we can draw two types of linear
regression equation as:

Simple Linear Regression Equation

There are only two variables in the study of simple linear regression. The method
of obtaining regression equations by the method of least square is more time
consuming and tedious. It also involves lot of calculations. A much simpler method
of obtaining regression equations if one in which instead of taking actual values of
x and y, we take their deviations from the mean values. It reduces a lot of
calculations and gives us same result as given by the least square method.
Let x and y be the two variables, then two simple regression equations can be
obtained as:

(x - –x) = bxy (y - –y)  when x depends upon y

(y - –y) = byx (x - –x)  when y depends upon x

Where,
–x = Sx ; –y = Sy
n n
nSxy - Sx Sy
bxy = regression coefficient when x depends upon y = nSy2 - (Sy)2

nSxy - Sx Sy
byx = regression coefficient when y depends upon x = nSx2 - (Sx)2

Example 9 The following table shows the number of motor registration and the
sale of Gorakhali Motor tyres by a whole sale dealer in Kathmandu for the term of
5 years.
Motor Registration in (000 No. of tyres sold (000
Year
nos) nos)
1 60 125
2 63 110
3 72 130
4 75 135
5 80 150
i) Estimate the sale of tyres when expected motor registration in next year is
90,000.
 FORECASTING CHAPTER - 20 49

ii) Also calculate correlation coefficient and interpret.

Solution:
Let, motor registration = x and Number of tyres sold = y
Computation table
X Y x 2 xy y2
60 125 3600 7500 15625
63 110 3969 6930 12100
72 130 5184 9360 16900
75 135 5625 10125 18225
80 150 6400 12000 22500
x = 350  = 650 x2 = 24778 xy = 45915 y2 = 85350
(i) Here y is the dependent variable. So, the regression equation of y on x that best
fits the given data is given by:
y - ȳ = bxy (x - ̄) ……………….(i)
Where,
∑x 350 ∑y 650
̄ = n = 5 = 70 ȳ = n = 5 = 130
n∑ xy - ∑x. ∑y 5 × 45915 - 350 × 650 229575 - 227500 2075
byx = n∑x2 - (∑x)2 = 5 × 24778 - 350 × 350 = 123890 - 122500 = 1390 =
1.4928
From equation (i), we get
y – 130 = 1.4928 (x - 70)
Or, y – 130 = 1.4928 x - 104.50
Or, y = 1.4928 x + 25.5 which is required estimated regression of y on x.
Again,
x = 90,000 y=?
y = 1.4928 × 90 + 25.5 = 134.352 + 25.5 = 159.852
Therefore, the estimated number of tyres is 159852, when expected motor
registration is 90,000
(ii) Calculation of correlation coefficient
n∑xy - (∑x) (∑y)
Correlation Coefficient (r) =
{n∑x2 - (∑x)2} {n∑y2 - (∑y)2}
5 × 45915 - 350 × 650
=
{5 × 247.78 - (350)2} {5 × 85350 - (650)2}
229575 - 227500 2075
= = 2430.535 = 0.85
(123890 -122500) (426750 - 422500)
Thus, three is high degree of positive correlation between the no. of motor
registration and no. of tyres sold.
50 / OPERATIONS MANAGEMENT

Multiple Linear Regression Equation

Multiple regression analysis consists of the measurement of the relationship
between one dependent variable and two or more independent variables. Multiple
regression equation is the algebraic relationship between one dependent and two
are more independent variables. The procedure for studying multiple regressions
is similar to the one we have for simple regression. Let y is the dependent variable
and x1 and x2 be the two independent variables, then the multiple regression
equation of y on x1 and x2 order least square method can be obtained as
y = a + b1x1 + b2x2 ……….… (i)
Here, a, b1 and b2 are regression parameters and to obtain the value of them, we
need to solve following three normal equations.
y = na + b1x1 + b2x2
yx1 = ax1 + b1x12 + b2x1x2
yx2 = ax2 + b1x1x2 + b2x22
Example 10 For the following data on sales of a product, the advertising
expenses and the price of the product, estimate the line that best describe the data.
Y X1 X2
Year Sales Advertising exp. Price
('00 units sold) ('00 Rs.) ('00 Rs./unit)
1 2 1 2
2 3 2 2
3 5 3 2
4 4 5 3
5 1 7 4
6 2 6 4
7 5 4 5
8 3 5 6
9 2 3 6
10 3 4 6
Total 30 40 40
Also, obtain the coefficient of determination and coefficient of multiple
correlations.
Solution
We have, Sales=Y ('00 units sold)
Advertising exp. = X1 ('00 Rs.)
Price = X2 ('00 Rs./unit)
 FORECASTING CHAPTER - 20 51

Computation table
Y X1 X2 YX1 YX2 X1X2 X12 X22 Y2
2 1 2 2 4 2 1 4 4
3 2 2 6 6 4 4 4 9
5 3 2 15 10 6 9 4 25
4 5 3 20 12 15 25 9 16
1 7 4 7 4 28 49 16 1
2 6 4 12 8 24 36 16 4
5 4 5 12 15 20 16 25 9
3 5 6 10 12 30 25 36 25
2 3 6 15 30 18 9 36 9
3 4 6 12 18 24 16 36 4
Y = X1 = 40 X2 = 40 YX1 = YX2 = X1X2 = X12 = X22 = Y2 =
30 111 119 171 190 186 106
Here Y is the dependent variable. So, the regression line of Y on X1 and X2 is
given by:
Y = a + b1X1 + b2X2………………………………….……..(i)
Here, a. b1, & b2 are estimated by solving the following 3 normal equations:
Σy = na + b1 ΣX1 + b2 ΣX2
ΣX1y = a ΣX1 + b1 ΣX21 + b2 ΣX1X2 ………………….….(ii)
ΣX2y = a ΣX2 + b1 ΣX1X2 + b2 ΣX22
Now substituting corresponding sum values in normal equations, we get;
30 = l0a + 40b1 + 40b2 ……………………….…….………(iii)
111 = 40a +190b1 + l71b2 ……………………….…………(iv)
119 = 40a +171b1 + 186b2 ………………………….…….…(v)
Multiplying equation (iii) by 4 and solving with (v), we get;
120 = 40a + 160b1 + 160b2
111 = 40a + 190b1 + 171b2
- - - -
9 = - 30b1 - 11b2
or, 9 = -30b1 -11b2 …………………….………………….…….(vi)
On solving equation (iv) and (v), we get;
111 = 40a +190b1 + l71b2
119 = 40a +171b1 + 186b2
- - - -
- 8 = 18b1 - 15b2
52 / OPERATIONS MANAGEMENT

or, - 8 = 9b1 - 15b2 ……………………………………….…… (vii)

Again, multiplying equation (vi) by 19, equation (vii) by 30 and then solving,
we get;
171 = - 570b1 - 209b2
240 = 570b1 - 450b2
- - +
- 69 = -659b2
 b2 = 0.105
From equation (vii), we get; From equation (iii), we get;
-8 = 9b1 - 15 × 0.105 30 = l0a + 40 × (-0.338) + 40 ×
 b1 = -0.338 (0.105)
 a =3.932
Hence, from equation (i), we get;
^
Y = 3.932 - 0.338X1 + 0.105X2
Which required multiple regression equation that best fits the given data.
Now,
Multiple correlation coefficient;
a  Y  b1 X 1Y  b 2  X 2 Y  n 3 .9 3 2  3 0  0 .3 3 8  1 1 1  0 .1 0  1 1 8  1 0  3
2

i.e. RY..X1X2 = =
X  Y 10  10  3
2 2 2 2

1 1 7 .9 6  3 7 .5 1 8  1 2 .4 9 5  9 0 2 .9 3 7
= = = 0.428
16 16

And, Coefficient of multiple determinations;

R2Y.X1X2 = (0.428)2 = 0.183
which shows that there is 18.3% variation in dependent variable due to two
independent variables X1 and X2.

Econometric Modeling
Some forecasting methods use the assumption that it is possible to identify the
underlying factors that might influence the variable that is being forecast. For
example, including information about weather conditions might improve the
ability of a model to predict umbrella and raincoat sales. This is a model of
seasonality which shows a regular pattern of up and down fluctuations. In addition
to weather, seasonality can also be due to holidays and customs such as predicting
that sales in college football apparel will be higher during football season as
opposed to the off season.
It is also subject to the discretion of the forecaster. There are several informal
methods which do not have strict algorithms, but rather modest and unstructured
 FORECASTING CHAPTER - 20 31

guidance. This is quite different from the aforementioned model of seasonality

whose graph would more closely resemble a sine or cosine wave. The most
important factor when performing this operation is using concrete and
substantiated data. Forecasting off of another forecast produces inconclusive and
possibly erroneous results.

Seasonal Variations
The fluctuation in time series due to seasonal factors like weather, climate,
festivals and occasions is called seasonal variation. Seasonal variations are
generally related with less than one year time period. Most of business and
economic data expressed weekly, monthly or quarterly show the seasonal
variation. If the data is observed annually, in this situation we draw trend line and
compute trend values. If the data is observed seasonally (or in a part of year), in
this situation we can measure seasonal variation (effect of season). In order to
measure the seasonal variations in the time series, we need to calculate seasonal
index and the seasonal index can be observed in two forms. They are as follows:
 Simple average seasonal index
 Ratio to moving average seasonal index
a. Multiplicative model
b. Additive model

Simple Average Seasonal Index

Simple average method is the simplest method for measuring seasonal index on
which ratio between average of individual seasons and average of average is
considered as seasonal index (SI). This method involves following steps:
Step I: Calculate the seasonal total
Step II: Calculate the seasonal average
Step III: Obtain the average of average (i.e. grand average) in working note
Sum of seasonal averages
Grand average = Number of averages
Step IV: Calculate the seasonal index (SI) for each seasons
Average of individual season
Seasonal index (SI) = Grand average × 100

Example 12 The Reliable Home Builders has collected data on the homes
it has started during the last 5 years.
32 / OPERATIONS MANAGEMENT

Year Spring Summer Fall Winter

1988 8 10 7 5
1989 9 10 7 6
1990 10 11 7 6
1991 10 12 8 7
1992 11 13 9 8
(a) Calculate the seasonal index for each quarter.
(b) If the company's working capital needs are related directly to the number of
starts, by how much should his working capital need change between summer
and winter?
Solution:
(a) Computation of Seasonal Indices
Year Spring Summer Fall Winter
1988 8 10 7 5
1989 9 10 7 6
1990 10 11 7 6
1991 10 12 8 7
1992 11 13 9 8
Total 48 56 38 32
Average 9.6 11.2 7.6 6.4
S.I. 110.34 128.74 87.36 73.56
S.I. represents seasonal indices:
Sum of average 9.6+11.2+7.6+6.4
Grand average = No. of averages = 4 = 8.7

Average for the spring 9.6

S.I. for the spring = Grand average × 100 = 8.7 × 100 = 110.34 and so
on……
(b) In order to know that by how much the working capital should be changed
from summer to winter, we need to calculate the %change in S.I. from summer
to winter.
S.I. for the summer = 128.74
S.I. for the winter = 73.56
 Change in the S.I. between summer and winter = 128.74 - 73.56 = 55.18
55.18
and Change in the S.I. between summer and winter = 128.74 × 100% = –42.86% =
42.86% (Decreased)
It means the company's working capital should be decreased by 42.86% from
summer to winter.
 FORECASTING CHAPTER - 20 33

Ratio to moving average seasonal index

Ratio to moving average method is more complex and tedious method for
measuring seasonal index but even it is more satisfactory and widely used method.
It is applied under either multiplicative model or additive model. It consists
following steps:
 Arrange the years and quarters vertically along with the given values (i.e. Y)
 Calculate 4 quarterly moving total
 Calculate 4 quarterly moving average
 Calculate centered moving average and denote it by Ye
 Calculate ratio to moving average (RTMA) value as:
a. If multiplicative model
Y
RTMA   100
Ye

b. If additive model
RTMA = Y - Ye
Construct seasonal index (SI) table
Note:
The sum of seasonal Indices (∑SI) must be 400 for multiplicative model and zero
for additive model. But, if it is not so, then the adjusted SI should be calculated by
using correction factor as below:
Required SSI
a. Correction factor for multiplicative model = Actual SSI

Actual SSI
b. Correction factor for additive model = 4

Example 13 Calculation seasonal indices by the ratio to moving average method

from the following data:
Years 1972 1973 1974 1975
Quarters
Q1 75 86 90 100
Q2 60 65 72 78
Q3 54 63 66 72
Q4 59 80 85 93
Solution: Computation of trend.
34 / OPERATIONS MANAGEMENT

4- RTMA
4- Central
Year & quarterly Central value =
Values (y) quarterly moving
quarter moving moving col. (2)
(2) moving average
(1) average total (5) col. (6) ×
total (3) (6)
(4) 100
Q1 75

Q2 60
 248 62
1972
Q3 54 126.75 63.375 85.21
 259 64.75
Q4 59 130.75 65.375 90.25
 264 66
Q1 86 134.25 67.125 128.12
 273 68.25
Q2 65 141.75 70.875 91.71
 294 73.50
1973
Q3 63 148 74 85.14
 298 74.50
Q4 80 150.75 75.375 106.14
 305 76.25
Q1 90 153.25 76.625 117.46
 308 77
Q2 72 155.25 77.625 92.75
 313 78.25
1974
Q3 66 159 79.5 83.02
 323 80.75
Q4 85 163 81.5 104.29
 329 82.25
Q1 100 166 83 120.48
 335 83.25
Q2 78 169.5 84.75 92.04
1975  343 85.75
Q3 72

Q4 93
Assuming multiplicative model of the time series the trend (moving average)
values are eliminated on expressing the given value (y) as a percentage of the
trend values.
 FORECASTING CHAPTER - 20 35

Example 14 Assuming additive model, calculate the seasonal indices by the

method of moving average of following data:
Year Q1 Q2 Q3 Q4
2006 20 70 62 110
2007 22 90 80 162
2008 40 142 88 218
2009 55 160 122 226
2010 65 190 135 300
Solution
Computation of trend.
4-
4- Central RTMA
Year & quarterly Central
Values (y) quarterly moving value =
quarter moving moving
(2) moving average Col. 2 –
(1) average total (5)
total (3) (6) Col. 6
(4)
Q1 20

Q2 70
 262 65.5
2006
Q3 62 131.5 65.75 –3.75
 264 64.66
Q4 110 130.137 68.50 41.50
 284 71
Q1 22 146.5 73.25 –51.25
 302 68.75.5
Q2 90 164 82 8
 354 73.88.5
2007
Q3 80 181.5 90.75 –10.75
 372 74.93
Q4 162 199 99.5 71.25
 424 76.106
Q1 40 214 107 –67
 432 108
Q2 142 230 115 27
 488 122
2008
Q3 88 247.75 123.875 –35.875
 503 125.75
Q4 218 256 128 90
 521 130.25
36 / OPERATIONS MANAGEMENT

Q1 55 269 134.50 –79.50

 555 138.75
Q2 160 255.75 127.875 32.125
 468 117
2009
Q3 122 260.25 130.125 –8.125
 573 143.25
Q4 226 294 147 79
 603 150.75
Q1 65 304.75 152.375 –87.375
 616 154
Q2 190 326 163.25 26.75
2010  690 172.5
Q3 135

Q4 300
Now, RTMA value i.e. short term fluctuations are used to obtain the seasonal
indices because we have used the additive model of the time series.
Computation of Seasonal Indices
Short term fluctuation
Year
Q1 Q2 Q3 Q4
2006 – – –3.75 41.50
2007 –51.25 8.00 –10.75 71.25
2008 –67.00 27.00 –35.875 90
2009 –79.50 32.125 –8.125 79
2010 –87.375 26.75 – –
Total –285.125 93.875 –58.5 281.75
Average seasonal
–171.28 23.469 -14.625 70.4375
indices
Adjusted seasonal
- 73.28 21.47 -16.630 68.44
indices
Sum of the seasonal indices = – 71.28 + 23.469 – 14.625 + 70.4375 =
Hence an adjustments to the seasonal indices are to be made by subtracting each of
the seasonal index by the correction factor k where k = 8.0015/4 = 2.000375
Adjusted seasonal index for Q1 = –71.20 – 2.000375 = – 73.28
Adjusted seasonal index for Q2 = 23.469 – 2.000375 = 21.47 and so on.
 FORECASTING CHAPTER - 20 37

Accuracy and control of forecast / Forecast Errors

Forecasts are predictions of future values and are bound to have some errors in
them. Error is the difference between the actual value and forecasted value. If we
want to measure the degree of relationship between dependent and independent
variables, than there may occur some error because we generally avoid many
independent variables. For example, demand depends upon many factors but we
often show the relationship of demand with price. Therefore, it is necessary to
measure the accuracy of forecasted values. It enables the managers to choose the
best method of forecasting.
The measurement of accuracy of forecasting depends upon the measuring the
forecasting errors and there are various methods to measure forecasting errors.
Some of them are as follows:
 Error or residual (E)
 Standard error estimate (SE)
 Mean squared error (MSE)
 Mean absolute deviation (MAD)
 Mean absolute percent error (MAPE)
(i) Error or residual (E)
Error is simply the forecasting error which the gap between actual value and
forecasted value. In simple cases, a forecast is compared with an outcome at a
single time-point and a summary of forecast errors is constructed over a collection
of such time-points. Here the forecast may be assessed using the difference or
using a proportional error. By convention, the error is defined using the value of
the outcome minus the value of the forecast. Mathematically, it can be expressed
as:

Error or residual (E) = y - ^

y

Where, y = actual value and ^

y = forecasted value
(ii) Standard error of estimate (SE or Syx)
Standard error of estimate can be defined as the statistical tools used to measure
the degree of variability in the expected value of dependent variable around the
estimations. It also shows the reliability of data fitted for regression equation. It
can be obtained as:

E2 S(y - ^
y)2
SE or Syx = n = n
38 / OPERATIONS MANAGEMENT

(iii) Mean squared error (MSE)

The mean squared error of an estimator is one of many ways to quantify the
difference between values implied by an estimator and the true values of the
quantity being estimated. MSE is a risk function, corresponding to the expected
value of the squared error loss or quadratic loss. MSE measures the average of the
squares of the "errors." The error is the amount by which the value implied by the
estimator differs from the quantity to be estimated. The difference occurs because
of randomness or because the estimator doesn't account for information that could
produce a more accurate estimate.
The MSE is the second moment (about the origin) of the error, and thus
incorporates both the variance of the estimator and its bias. For an unbiased
estimator, the MSE is the variance of the estimator. It can be calculated as:

S(y - ^
y)2
MSE = (SE)2 = n
(iv) Mean absolute deviation (MAD)
It represents the average value of non negative deviation of variable taken from its
mean value. It is obtained by taking absolute deviation between the actual value
and forecasted value. The equation for mean absolute deviation is as follows:

S |E| S |y - ^
y|
MAD = n = n
(v) Mean absolute percent error (MAPE)
The mean absolute percentage error (MAPE), also known as mean absolute
percentage deviation (MAPD), is a measure of accuracy of a method for
constructing fitted time series values in statistics, specifically in trend estimation.
It helps to relate the forecast error to the level of demand. It usually expresses
accuracy as a percentage, and is defined by the formula:

S|y - ^
y|
S(|E|/y) × 100 y × 100
MAPE = n = n
Example 11 The number of cases of Merlot wine sold by a Paso Robles
winery in an 8 year period as follows:
Year 1991 1992 1993 1994 1995 1996 1997 1998
Cases of wine 270 356 398 456 358 500 410 376
Required:
 FORECASTING CHAPTER - 20 39

a. Exponential smoothing with a weight or smoothing constant of 0.4 will be used

to forecast wine sales. What will be the forecast for year 1999?
b. Find standard error of estimate (Syx), mean absolute deviation (MAD) and mean
absolute percent error (MAPE). [PU 2005, Fall]
Solution
a. We have,
Smoothing factor () = 0.4
Forecasted value at (1991) = 270 (assumed)
Calculation of forecasted value for each year
^
y 1992 = ^
y1991 + (y1991 - ^
y1991) = 270 + 0.4(270 – 270) = 270
^
y 1993 = ^
y1992 + (y1992 - ^
y1992) = 270 + 0.40 (356 – 270) = 304.4
^
y 1994 = ^
y1993 + (y1993 - ^
y1993) = 304.4 + 0.4(398 – 304.4) = 341.84
^ ^ ^
y 1995 = y1994 + (y1994 - y1994) = 341.84 + 0.4 (456 – 341.84) = 387.5
^
y 1996 = ^
y1995 + (y1995 - ^
y1995) = 387.5 + 0.4(358 – 387.5) = 375.5
^
y 1997 = ^
y1996 + (y1996 - ^
y1996) = 375.5 + 0.4 (500 – 375.5) = 425.3
^
y 1998 = ^
y1997 + (y1997 - ^
y1997) = 425.3 + 0.4 (376 – 425.3) = 419.18
^
y 1999 = ^
y1998 + (y1998 - ^
y1998) = 419.18 + 0.4(376 – 419.18) = 401.90
b. Computation of Syx, MAD and MAPE
Year Actual value Forecast (^ y) |y - ^
y| (y - ^
y)2 (y - ^
y)
(y) y × 100
1991 270 270 0 0 0
1992 356 304.4 51.6 2662.56 14.49
1993 398 341.84 56.16 3153.94 85.90
1994 456 387.5 68.50 4692.25 15.02
1995 358 375.5 17.50 306.25 4.89
1996 500 425.3 74.7 5580.09 14.94
1997 440 419.18 9.18 84.27 2.24
1998 376 401.90 25.9 670.81 6.89

17150.17

(y - ^
y)2 17150.17
Syx = n = 8 = 213.77 = 46.30

S|y - ^
y|
× 100
S|y - ^
y| 303.54 y 144.37
MAD = n = 8 = 37.94 MAPE = n = 8 = 18.05
40 / OPERATIONS MANAGEMENT

Choosing Forecast Techniques

Forecast error measures provide important information for choosing the best forecasting
method for a service or product. They also guide managers in selecting the best values
for the parameters needed for the method: n for the moving average method, the weights
for the weighted moving average method, a for the exponential smoothing method, and
when regression data begins for the trend projection with regression method. The
criteria to use in making forecast method and parameter choices include (1) minimizing
bias (CFE); (2) minimizing MAPE, MAD, or MSE; (3) maximizing r 2; (4) meeting
managerial expectations of changes in the components of demand; and (5) minimizing
the forecast errors in recent periods. The first three criteria relate to statistical measures
based on historical performance, the fourth reflects expectations of the future that may
not be rooted in the past, and the fifth is a way to use whatever method seems to be
working best at the time a forecast must be made.
Using Statistical Criteria Statistical performance measures can be used in the selection
of which forecasting method to use. The following guidelines will help when searching
for the best time series models:
1. For projections of more stable demand patterns, use lower a values or larger n values
to emphasize historical experience.
2. For projections of more dynamic demand patterns using the models covered in this
chapter, try higher a values or smaller n values. When historical demand patterns are
changing, recent history should be emphasized.
Often, the forecaster must make trade-offs between bias (CFE) and the measures of
forecast error dispersion (MAPE, MAD, and MSE). Managers also must recognize that
the best technique in explaining the past data is not necessarily the best technique to
predict the future, and that “overfitting” past data can be deceptive. A forecasting
method may have small errors relative to the history file, but may generate high errors
for future time periods.
Tracking Signals
A tracking signal is a measure that indicates whether a method of forecasting is
accurately predicting actual changes in demand. The tracking signal measures the
number of MADs represented by the cumulative sum of forecast errors, the CFE. The
CFE tends to be close to 0 when a correct forecasting system is being used. At any time,
however, random errors can cause the CFE to be a nonzero number. The tracking signal
formula is
Tracking signal = =

Each period, the CFE and MAD are updated to reflect current error, and the tracking
signal is compared to some predetermined limits (±4).
 FORECASTING CHAPTER - 20 41

Competitive priorities and capabilities

Often companies must prepare forecasts for hundreds or even thousands of services or
products repeatedly. For example, a large network of health care facilities must
calculate demand forecasts for each of its services for every department. This
undertaking involves voluminous data that must be manipulated frequently. However,
software system can ease the burden of making these forecasts and coordinating the
forecasts between customers and suppliers. Many forecasting software packages are
available, including Manugistics, Forecast Pro, and SAS. The forecasting routines in
OM Explorer and POM for Windows give some hint of their capabilities. Forecasting is
not just a set of techniques, but instead a process that must be designed and managed.
There is no one process that works for everyone..

WORKED OUT EXAMPLES

Example 15 In a manufacturing company, the operations manager records
the demand of product for 5 days in following order.
Days 1 2 3 4 5
Demand 1500 2000 1500 4000 6000
(a) Forecast the demand for 6th day using four days model for simple moving
average.
(b) Forecast the demand for 6th day using three days model with the demand in
most recent period weighted twice as heavily as each of twice of previous 2
periods.
Solution
(a) Forecasting the value of 6th day using four month simple moving average.
Demand of (2nd + 3rd + 4th + 5th) day
SMA = 4
2000 + 1500 + 4000 + 6000 13500
= 4 = 4 = 3375
 The forecasted demand for 6h day is 3375.
(b) Forecasting the value of 6th day using three month weighted moving
average.
Days Demand Weight (W) DW
(D)
3 1500 0.25 375
4 4000 0.25 1000
5 6000 0.50 3000
W = 1 DW = 4375
42 / OPERATIONS MANAGEMENT

DW 4375
WMA = = 1 = 4375
W
 The forecasted value for 6th day is 4375.

Example 16 Taking the smoothing factor of 0.25, fill in the blank of

following table.
Month Sales in Forecast (units)
units
Baishakh 10,000 65000
Jestha 25,400 ?
Ashad 18,750 ?
Shrawan 22,300 ?
Bhadra – ?
Solution
From exponential smoothing, we can forecast t he value of next period by using
following formula:
^
yt+1 = ^
yt t-^y t)
Then,
^
yJestha = ^
y Baishakh + 0.25 (yBaishakh – ^yBaishakh)= 6500 + 0.25(10000 – 6500) =
7375
^
yAshad = ^ y Jestha + 0.25 (yJestha – ^y Jestha)= 7375 + 0.25 (25400 – 7375) =
11881.25
^
yShrawan =^y Ashadh + 0.25(yAshadh – ^ yAshadh) = 11881.25 + 0.25 (18750 –
11881.25) = 13598.44
^
yBhadra = ^y Shrawan + 0.25(yShrawan – ^ y Shrawan) = 13598.44 + 0.25 (22300 –
13598.44) = 15773.83
Now, the filled table can be obtained as
Month Baishakh Jestha Ashad Shrawan Bhadra
Sales in 10,00 25,400 187,500 22,300 –
units
Forecast 6,500 7,375 11881.25 13598.44 15,733.83
(units)
 FORECASTING CHAPTER - 20 43

Example 17 Calculate the MAD and MAPE from the following data:
Year 1990 1991 1992 1993 1994 1995
Actual demand (000) 95 107 110 96 109 105
Forecast (000) 100 100 100 100 100 100
Solution
Computation of MAD and MAPE
Year Actual demand Forecast |y - ^
y| (y - ^
y)/y × 100
(y) ^
(y)
1990 95 100 5 5.26
1991 107 100 7 6.54
1992 110 100 10 9.09
1993 96 100 4 4.17
1994 109 100 9 8.26
1995 105 100 5 4.76
-^y| = -^y)/y × 100 =
40 38.08
We know,
|y - ^
y| 40
MAD = n = 6 = 6.67

|y - ^
y|
y × 100 38.08
MAPE = n = 6 = 6.35

Example 18 The number of cases of merlot wine sold by a Paso Robles

Winery in an 8-year period follows:
Year 1991 1992 1993 1994 1995 1996 1997 1998
Cases of wine 270 356 398 456 358 500 419 376
(y)
a. Exponential smoothing with a weight or smoothing constant of 0.4 will be used
to forecast wine sales. What will be the forecast for year 1999?
b. Find MAD, and standard error of estimate (SYX) for the exponentially
smoothed forecasts in (a).
[PU 2005, Fall]
Solution
a. Since forecasted demand for the year 1991 is not given. We assume actual
demand is equal to forecasted demand for this year. That is forecasted demand
for 1991 (^
y1991) = 270 units.
Now,
 44 / OPERATIONS MANAGEMENT

^
y1992 = ^
y1991 + (y1991 – ^
y 1991) = 270 + 0.4(270 – 270) = 270
^
y1993 = ^
y1992 + (y1992 – ^
y 1992) = 270 + 0.4(356 – 270) = 304.4
^
y1994 = ^
y1993 + (y1993 – ^
y 1993) = 304.4 + 0.4(398 – 304.4) = 341.84
^
y1995 = ^
y1994 + (y1994 – ^
y 1994) = 341.84 + 0.4(456 – 341.84) = 387.5
^
y1996 = ^
y1995 + (y1995 – ^
y 1995) = 387.5 + 0.4(358 – 387.5) = 375.5
^
y1997 = ^
y1996 + (y1996 – ^
y 1996) = 375.5 + 0.4(500 – 375.5) = 425.3
^
y1998 = ^
y1997 + (y1997 – ^
y 1997) = 425.3 + 0.4(410 – 425.3) = 419.18
^
y1999 = ^
y1998 + (y1998 – ^
y 1998) = 419.18 + 0.3(376 – 419.18) = 401.90
b. Computation table
Year Case of wine ^
y |y – ^
y| (y – ^
y)2
1991 270 270 0 0
1992 356 304.4 51.6 2662.5
1993 398 341.84 56.16 3153.9
1994 456 387.5 68.5 4692.25
1995 358 375.5 17.5 306.25
1996 500 425.3 74.7 5580.09
1997 410 419.18 9.18 84.27
1998 376 401.90 25.9 670.81

–^
y| = 303.5 –^
y)2 =
17150.17
Now,
S|y – ^
y| 303.54
MAD = n = 8 = 37.94

S(y – ^
y)2 17150.17
Standard error of estimate (SYX) = n = 8 =

2143.77 = 46.30
Example 19 The following data represent the short-term interest date (90
day commercial paper) in the United States from 1991 to 2000.
Year 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Short-term interest 5.87 3.75 3.22 4.67 5.93 5.41 5.59 5.38 5.23 6.34
rate (in %)
a. Fit a 3-year moving average to the data.
 FORECASTING CHAPTER - 20 45

b. Using a smoothing coefficient of W = 0.05, exponentially smooth the series and

plot your result on your chart.
c. What is your exponentially smoothed forecast for 2001? [PU 2006, Spring]
Solution
a. Computation table
Year Short-term interest rate (y) 3-year moving average
1991 5.87 –
1992 3.75 –
1993 3.22 –
1994 4.67 4.28
1995 5.93 3.88
1996 5.41 4.61
1997 5.59 5.33
1998 5.38 5.64
1999 5.23 5.46
2000 6.34 5.4
b. Assume ^ y = 5.87
^
y1992 = ^
y 1991 + (y1991 – ^
y 1991) = 5.87 + 0.05(5.87 – 5.87) = 5.87
^
y1993 = ^
y 1992 + (y1992 – ^
y 1992) = 5.87 + 0.05(3.75 – 5.87) = 5.764
^
y1994 = ^
y 1993 + (y1993 – ^
y 1993) = 5.764 + 0.05(3.22 – 5.704) = 5.128
^
y1995 = ^
y 1994 + (y1994 – ^
y 1994) = 5.128 + 0.05(4.67 – 5.128) = 5.105
^
y1996 = ^
y 1995 + (y1995 – ^
y 1995) = 5.105 + 0.05(5.93 – 5.105) =5.15
^
y1997 = ^
y 1996 + (y1996 – ^
y 1996) = 5.15 + 0.05(5.41 – 5.15) = 5.16
^
y1998 = ^
y 1997 + (y1997 – ^
y 1997) = 5.16 + 0.05(5.59 – 5.16) = 5.18
^
y1999 = ^
y 1998 + (y1998 – ^
y 1998) = 5.18 + 0.05(5.38 – 5.18) = 5.19
^
y2000 = ^
y 1999 + (y1999 – ^
y 1999) = 5.19 + 0.05(5.23 – 5.19) = 5.192
Graphical representation.

6
 Forecasted value
5.8

Sales (000)

5.6
5.4
5.2     
  
5
4.8
4.6
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Year

c. ^
y2001 = ^
y2000 2000 –^
y 2000) = 5.192 + 0.05(6.34 – 5.192) = 5.25
46 / OPERATIONS MANAGEMENT

Example 20 Data on sugar production for 20 periods are tabulated below.

data. The first 12 values can be used for initialization of the model.
Sugar production (in Sugar production (in
Month tons) Month tons)
(y) (y)
1 30.50 11 25.70
2 28.80 12 30.90
3 31.50 13 31.50
4 29.90 14 28.10
5 31.40 15 30.80
6 33.50 16 29.50
7 25.70 17 29.80
8 32.10 18 30.00
9 29.10 19 29.90
10 30.80 20 31.50
Also calculate the mean absolute deviation, and mean squared error.[PU 2007, Fall]
Solution
a. Computation table:
Sugar prodn. Forecasted Sugar
Year |y – ^
y| (y - ^
y)2
(y) prodn. (^
y)
1 30.50 30.50 0 0
2 28.80 30.50 1.7 2.89
3 31.50 30.33 1.17 1.37
4 29.90 30.44 0.55 2.40
5 31.40 30.39 1.01 1.02
6 33.50 30.49 3.01 9.06
7 25.70 30.79 5.09 25.90
8 32.10 30.28 1.82 3.31
9 29.10 30.46 1.36 1.85
10 30.80 30.32 0.48 0.23
11 25.70 30.37 4.67 21.81
12 30.90 29.90 1.00 1.00

|y – ^
y| = (y - ^
y)2 =
2186 93.61
 FORECASTING CHAPTER - 20 47

Now,
^
y2 = ^
y1 + (y1 – ^
y1) = 30.50 + 0.1(30.50 – 3.50) = 30.50
^
y3 = ^
y2 + (y2 – ^
y2) = 30.50 + 0.1(28.80 – 30.50) = 30.33
^ ^ ^
y4 = y3 + (y3 – y3) = 30.33 + 0.1(31.50 – 30.33) = 30.447
^
y5 = ^
y4 + (y4 – ^
y4) = 30.447 + 0.1(29.90 – 30.447) = 30.3923
^
y6 = ^
y5 + (y5 – ^
y5) = 30.3923 + 0.1(31.40 – 30.3923) = 30.49307
^
y7 = ^
y6 + (y6 – ^
y6) = 30.49307 + 0.1(33.50 – 30.49307) = 30.79
^
y8 = ^
y7 + (y7 – ^
y7) = 30.79 + 0.1(25.70 – 30.79) = 30.28
^
y9 = ^
y8 + (y8 – ^
y8) = 30.28 + 0.1(32.10 – 30.28) = 30.46
^
y10 = ^
y9 + (y9 – ^
y9) = 30.46 + 0.1(29.10 – 30.46) = 30.32
^
y11 = ^
y10 + (y10 – ^
y10) = 30.32 + 0.1(30.80 – 30.32) = 30.37
^
y12 = ^
y11 + (y11 – ^
y11) = 30.32 + 0.1(25.70 – 30.37) = 29.90
S|y – ^
y| 21.86 S(y - ^
y)2 93.61
MAD = n = 12 = 1.82 MSE = n = 12 = 7.80

Example 21 The quarterly sales of a department store chain were recorded for the
past four years. These figures are shown in the accompanying table. Apply the
exponential smoothing technique with exponential constant 0.2 and graph the results.
Year Time period Quarter Quantity of petrol sold (y)
1 1 39
2 2 37
1995
3 3 61
4 4 58
5 1 18
6 2 56
1996
7 3 82
8 4 27
9 1 41
10 2 69
1997
11 3 49
12 4 66
13 1 54
14 2 42
1998
15 3 90
16 4 66
48 / OPERATIONS MANAGEMENT

Solution
Time Quantity of petrol ^
Year Quarter y
period (y)
1 1 39 39
2 2 37 39
1995
3 3 61 38.6
4 4 58 43.08
5 1 18 46.064
6 2 56 40.4512
1996
7 3 82 43.56
8 4 27 51.23
9 1 41 46.38
10 2 69 45.304
1997
11 3 49 50.04
12 4 66 49.03
13 1 54 53.06
14 2 42 53.25
1998
15 3 90 51.00
16 4 66 58.8
Now,
^
y2 = ^
y1 + (y1 – ^
y1) = 39 + 0.2(39 – 39) = 39
^
y3 = ^
y2 + (y2 – ^
y2) = 39 + 0.2(37 – 39) = 38.6
^
y4 = ^
y3 + (y3 – ^
y3) = 38.6 + 0.2(61 – 38.6) = 43.08
^
y5 = ^
y4 + (y4 – ^
y4) = 43.08 + 0.2(58 – 43.08) = 46.064
^
y6 = ^
y5 + (y5 – ^
y5) = 46.064 + 0.2(18 – 46.064) = 40.4512
^
y7 = ^
y6 + (y6 – ^
y6) = 40.4512 + 0.2(56 – 40.45) = 43.56
^
y8 = ^
y7 + (y7 – ^
y7) = 3.56 + 0.2(82 – 43.50) = 51.23
^
y9 = ^
y8 + (y8 – ^
y8) = 51.23 + 0.2(27 – 51.23) = 46.38
^
y10 = ^
y9 + (y9 – ^
y9) = 46.38 + 0.2(41 – 46.38) = 45.304
^
y11 = ^
y10 + (y10 – ^
y10) = 45.304 + 0.2(69 – 45.304) = 50.04
^ ^ ^
y12 = y11 + (y11 – y11) = 50.04 + 0.2(49 – 50.04) = 49.83
^
y13 = ^
y12 + (y12 – ^
y12) = 49.83 + 0.2(66 – 49.83) = 53.06
^ ^ ^
y14 = y13 + (y13 – y13) = 53.06 + 0.2(54 – 53.06) = 53.25
^
y15 = ^
y14 + (y14 – ^
y14) = 53.25 + 0.20(42 – 53.25) = 51
^ ^ ^
y16 = y15 + (y15 – y15) = 51 + 0.20(90 – 51) = 58.8
 FORECASTING CHAPTER - 20 49

Graphical representation

70 Forecasted value
60
Forecasted value


   
50  
   
40  
 
30
20
10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Year

Example 22 A manufacturing company has monthly demand for one of its

products as follows:
Month February March April May June July August Sept.
Demand 520 490 550 580 600 420 510 610
Develop a three-period average forecast and a three-period weighted moving
average forecast with weights of 0.50, 0.30, and 0.20 for the most recent demand
values, in that order. Calculate MAD for each forecast, and indicate which would
seem to be most accurate:
Solution:
Compute the 3-month simple moving average
520 + 490 + 550
For May: SMA1 = 3 = 520 and so on………….
Compute the 3-month weighted moving average
For May: WMA1 = (0.50) (520) + (0.30) (490) + (0.20) (550) = 526.00 and so
on………….
Computation table
Month Demand (Y) SMA (^
y) WMA (^ y') |y - ^
y| |y - ^
y'|
February 520 – –
– –
March 490 – –
– –
April 550 – –
– –
May 580 520.00 526.00
60 54
June 600 540.00 553.00
60 47
July 420 576.67 584.00
156.67 164
August 510 533.33 506.00
23.33 4
September 610 510.00 501.00
100 109
October – 513.33 542.00
Total 400.00 378
50 / OPERATIONS MANAGEMENT

Compute the MAD value for both forecasts:

For SMA;
|y – ^
y| 400
MAD = n = 5 = 80

|y – ^
y'| 378
For WMA; MAD' = n = 5 = 75.6
From the above, there is not much difference in accuracy between the two
forecasts, although the weighted moving average is slightly better.

EXERCISE – 20

Theoretical Questions
1. What is forecasting? Explain its importance to the operations managers.
2. Differentiate quantities and qualitative methods of forecasting.
3. Describe the Delphi technique of forecasting? When would you use this
technique?
4. What do you mean by time series analysis? Explain its components.
5. What kind of forecasting techniques are used for long range strategic
planning?
6. Explain the weighted moving average method of forecasting. How it is differ
from the simple moving average methods?
7. What is exponential smoothing? How do you decide the smoothing factor?
8. What is linear regression? Explain the difference between dependent and
independent variables under the regression model.
9. What are the different measures of forecasting accuracy? Which one is the
superior?
10. What factors do you consider to select the forecasting technique?

Numerical Problems
1. The following is the historical demand of a product.
Month Jan. Feb. Mar. Apr. May Jun.
Demand 15 11 12 12 16 17
Forecast the demand for July by using:
a. 3 month simple moving average.
 FORECASTING CHAPTER - 20 51

b. 4 month simple moving average.

c. 6 month simple moving average.
2. You are given the information of manufacturing company:
Month 2001 2002 2003 2004 2005
Demand (000) 15 18 12 10 20
Forecast the demand for 2006 by using:
a. 4 years simple moving average method.
b. 3 years weighted moving average method taking weight in most recent
period as double as in two previous periods.
3. A company sales 10,000, 15,000 and 12,000 units of product in Baishakh,
Jestha and Ashadh respectively. The probability of goods sold in Baishakh is
0.40 and rest chance for being sales in Jestha and Ashadh both. Forecast the
sales in Shrawan by using weighted moving average methods.
4. Assuming the exponential constant as 0.40, fill in the blank of following table.
Year Actual sales (units) Forecast (units)
1 1000 450
2 1250 ?
3 1400 ?
4 1000 ?
5 – ?

5. The demand for a certain commodity was 200 units in June, 50 in July and 150
in August. The forecasted demand in June was 100 units with the smoothing
factor 0.20, forecast the demand of September.
6. The following tale shows the number of tourists in Pokhara over a 10 years
time period. Forecast the number of tourists in 2013 by using exponential
smoothing model with smoothing constant w = 0.10.
Year 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Number of tourist 28 27 33 25 34 33 35 30 33 35
(100)

7.
0.30.
52 / OPERATIONS MANAGEMENT

Year Actual demand

2007 150
2008 200
2009 100
8. Calculate SYX, MAD and MAPE from the following information.
Months 1 2 3 4 5
Actual value 1100 1050 1230 1400 1140
Forecasted value 1200 1200 1200 1200 1200

9. The following table shows the demand for five years.

Months 2060 2061 2062 2063 2064
Demand (units) 40 35 25 20 40
Given that the forecasted demand for 2060 is 50 units, obtain the forecasted
value for rest y
standard error of estimate and mean absolute deviation.
10. Sales of wave air conditioners have grown steadily during the past five years.
The sales manager had predicted in 1999 that 2000 sales would be 410 air
conditioners. Using exponential smoothing with a weight of a 0.30, develop
forecasts for 2001 through 2005.
Year Sales Forecast
2000 450 410
2001 495
2002 518
2003 563
2004 584
2005 ?
11. Civil Service Hospital has used a 9-month moving-average forecasting method
to predict drug and surgical dressing inventory requirements. The actual
demand for one item is as shown in the accompanying table. Using the
previous moving – average data, convert to an exponential smoothing forecast
for month 21.
Month 12 12 14 15 16 17 18 19 20
Actual Demand 78 65 90 71 80 101 84 60 73
 FORECASTING CHAPTER - 20 53

12. A computer software firm has experienced the following demand for its
"Personal Finance" software package.
Period 1 2 3 4 5 6 7 8
Units 56 61 55 70 66 65 72 75
Develop an exponential smoothing forecast using  = 0.40. Compute the
accuracy of the forecast using MAD and Bias (i.e. Error of estimate).
13. Suppose you are the capital budgeting officer of a small corporation whose
financing requirements over the last few years have been.
Year 1997 1998 1999 2000 2001 2002 2003
Millions of rupees 2.2 2.1 2.4 2.6 2.7 2.9 2.8
required
a. What is the trend equation which best describes the above data?
b. Calculate the percent of trend for the above data.
c. Calculate the relative cyclical residual for the above data.
d. In which year does the largest fluctuation from trend occur, and is it the
same for both methods? [PU 2005, Spring]
14. A computer firm specializing in software engineering has compiled the
following revenue records for the year 1989 to 1995.
Year 1989 1990 1991 1992 1993 1994 1995
Revenue (in million) 1.1 1.5 1.9 2.1 2.4 2.9 3.8
The second degree equation that best describe e the secular trend for these
data is
^
y = 2.119 + 0.375x + 0.020x2, where 1992 = 0 and x units = 1 year.
a. Calculate the percent of trend for these data.
b. Calculate the relative cyclical residual for these data.
c. In which years does the largest fluctuation from trend occur and is it same
for both methods?
d. Calculate MAD and MSE for these data.
15. The following table shows the assets of Nepal Bank limited in crors of rupees.
Year 2048/49 2049/50 2050/51 2051/52 2052/53 2053/54
Assets 83 92 71 90 169 91
Fit a straight line least square trend and forecast the figures for 2055/56.
54 / OPERATIONS MANAGEMENT

16. Mr. x has a car dealership for Indica in Kathmandu. The number of cars sold for
first 8 months are given below.
Months Feb Mar April May June July Aug Sept
Cars sold 45 52 41 36 49 47 43 48
Mr. X wants to predict the car sales for the month of November for the same
year. You are requested to satisfy his wish.
17. Below are given figures of production (in 000 units)
Year 1985 1987 1988 1989 1990 1991 1992
Production 77 88 94 85 91 98 90
Fit a straight line by the least square method. Forecast the production in 1986.
What is monthly increase in production?
18. Using 1974 as the origin, obtain a straight line trend by the method of least
square.
Year 1970 1972 1973 1974 1975 1976 1979
Value 140 144 160 152 168 176 180
Find the trend value of missing year of 1971, 1977 and 1978.
19. Forecast the probable value of x when y is 17 by using simple regression
model.
X 5 9 13 17 21
Y 3 8 13 18 23
20. The data relating age of husband and wife is given in following table.
Age of wife 18 20 22 23 27 28 30
Age of husband 23 25 27 30 32 31 35
If the age of wife is 45, then what will be the age of husband.
21. From the following table compute line of regression for estimating blood
pressure (BP)
Age 56 42 72 36 63 47
BP 147 125 160 118 149 128
Forecast the B.P. of man whose age is 50 years.
22. A panel of judge A and B graded seven debaters and independently awarded
the following marks.
Debaters 1 2 3 4 5 6 7
Marks by 'A' 40 34 28 30 44 38 31
Marks by 'B' 32 39 26 30 38 34 28
 FORECASTING CHAPTER - 20 55

An eighth debater was introduced and awarded by 36 marks by judge 'A'. The
judge 'B' was not present at that time. If the judge B were also present, then
how many marks would you expect by him for that eighth debater.
23. Construct the indices of seasonal variations from the following time series data
on consumption of cold drinks of 1000 bottles.
Years 1995 1996 1997 1998
Quarters
1st 90 75 80 85
2nd 75 80 78 75
3rd 85 78 75 80
4th 70 75 72 81
24. Determine the seasonal index for each quarter in order to measure the
seasonal variations from the data given below.
Quarters
Year
Fall Winter Spring Summer
1990 - - 90 30
1991 150 120 98 35
1992 160 115 95 30
1993 152 108 100 28
1994 145 115 107 32
1995 152 120 - -
25. Following data relates to the number of customers in post office in a particular
month.
Days
Weeks
S M T W T F S
I 145 200 180 180 110 90 10
II 170 280 275 200 165 150 8
III 55 252 250 160 90 52 7
IV 75 220 142 112 70 60 10
Calculate the daily indices to measure seasonal variations.
56 / OPERATIONS MANAGEMENT

26. Calculate the seasonal averages and seasonal indices for the following time
series.
Month 1974 1975 1976
Jan 15 23 25
Feb 16 22 25
March 18 28 35
April 18 27 36
May 23 31 36
June 23 28 30
July 20 22 30
August 28 28 34
September 29 32 38
October 33 37 47
November 33 34 41
December 38 44 53
27. Calculate the seasonal indices by using ratio to moving average method with
the reference of.
(a) Multiplicative model
(b) Additive model
Quarters
Year
Q1 Q2 Q3 Q4
1991 68 62 61 63
1992 65 58 66 61
1993 68 63 63 67

  