Forecasting

Unit Introduction
Forecasting - one of the key elements of operations management. It tells us what
the customer will need at what time and in which quantity. It relates the
management functions of planning, organizing and controlling. Company serves
their customers and the society at large by producing various goods and services
in factories and plants. The market needs for such products are changing. And
their suppliers have to respond more quickly than ever before with product
delivery to survive. To do so, they have to place a higher emphasis on forecasting
to determine the demand level. Otherwise, if a firm produces less than the actual
demand, customers will go unsatisfied and if they produce more, unsold products
will pile up. Forecasting, done by using data from past events or through some
subjective judgment (like experience, guesses, hunches, etc.) and reduces
severity of any unexpected events. If firm knows how much to produce, it can
plan and organize operations accordingly. And if operations have been properly
planned and organized, control is easier and smoother. This is where forecasting
comes in. Thus forecasting also reduces the costs of adjusting operations in
response to unexpected deviations by specifying future demand. Furthermore it
helps improve the organization’s competitive edge. Keeping this entire in mind,
the following lessons are covered in this unit: Products and Forecasting;
Different Elements of Forecasting; and Different Approaches and Techniques of
Forecasting.
School of Business

Unit- 3 Page- 54
 Bangladesh Open University

Lesson One: Product and Forecasting

Lesson Objectives
After completing this lesson you will be able to:

Understand the meaning of forecasting and its importance

Identify the subjects of forecasting

Explain the factors that affect the product demand of forecasting

Discuss how to develop a workable forecasting system

Forecast and its Implications

The term Forecast can have different meanings in different disciplines (e.g.,
business, economics, and political communities). In operations management, we Forecast is an inference
rather adopt a specific definition of forecast, distinguishing it from the broader of what is likely to
concept of prediction. Therefore a forecast is an inference of what is likely to happen in the future.
happen in the future. It is estimated by systematically combining and casting
forward data about the past in a predetermined way. It is objective in nature but
not an absolutely certain prophesy. Even very carefully prepared forecasts can be
wrong. In fact, it is extremely rare for a forecast to be exactly right. A prediction,
on the other hand, is an estimate of a future event achieved through subjective
considerations other than just past data. In case of prediction subjective
consideration need not occur in any predetermined way. Therefore the
implications of forecasting and/or prediction are:
• Forecasting is both objective and subjective: As noted, forecasting is the art Forecasts are possible
and science of predicting future events, which may involve taking historical only when a history of
data and projecting them into the future with some sort of mathematical data exists and
reasonable inferences
model. It may be subjective or intuitive prediction of the future or it may be can be reached on the
of both, i.e., a combination of mathematical models adjusted by good basis of that data.
judgment. As definitions make clear, forecasts are possible only when a
history of data exists and reasonable inferences can be reached on the basis
of that data. For example, a canteen manager can use past data to forecast the
number of a regular product like tea, but suppose s/he wants to offer a new
product—black coffee. Since no past data exists to estimate the sales of the
new product, prediction, instead of forecasting, will be needed.
• Forecasting prepares an organization for a common future objective through
coordination: Forecasting helps develop and implement a continuous
improvement program specially tailored for any particular business and
organization culture and environment. It is an aid to other program which
may involve implementation of proven manufacturing and distribution
planning and control tools and techniques such as Manufacturing Resources
Planning (MRP), Distribution Resources Planning (DRP) and Just-In-Time
(JIT). In forecasting, demand and sales forecasting in particular, is essential
to the three major activities of any operations manager that are planning,
scheduling and controlling the system (Figure 3.1.1). Demand forecast drives
the production capacity and scheduling system in a firm and affects the
financial, marketing and HR or personnel planning functions. So, forecasting
is not only the critical part of operations manager’s function but also
important to financial, marketing and HR manager. Thus, it brings every
department together and strengthens corporate culture and environment.
Forecasts provide a basis for coordination of plans for activities in various
parts of a company. When all parts of the company base their work on the
same forecasts, they prepare for the same future and their efforts are mutually

Production Operations Management Page- 55

 School of Business

supportive. In other words, co-ordination of activities is accomplished when

adequate forecasts prepare an organization for a common future objective.
• Forecasting aids in both short-term and long-term planning. Forecasts are
Forecasting is important important inputs for long-range planning and strategic decisions. At the same
for designing the long
term as well as for the time they are an important basis for shorter-range decisions in day-to-day
short term decision operations. Operations managers try to forecast a wide range of future events
making. that potentially affect success. Most often the concern is forecasting customer
demand for products or services. Managers may want long-run estimates of
overall demand or short-run estimates of demand for each individual product.
Businesses must develop forecasts of the demand level, which the company
should prepare to meet. Since service operations generally cannot store their
output as inventory, they must try harder to estimate the level of future
demand. When service organizations fail to accurately forecast their
demands, they often face a double-edged sword. Overstaffing leading to
waste and non-optimal operation on one hand and understaffing resulting in
loss of business, time and customers on the other.

Activity: Assume you want to start a garment business. Now, how the
forecasting techniques can help you to be successful in that business.

Subjects of Forecasts
A successful operation In making decisions, managers need to make inferences about the future in
plan must inference several subject areas. Such three relevant areas are technological developments,
about economic, business conditions, and the expected level of demand. With respect to these,
technical and demand
forecast.
organizations use three major types of forecasting in planning the future of their
operations. These are, economic, technological and demand forecasting. Figure
3.1.1 depicts these three different kinds of forecast in organizational setting:

Information from
Internal and External
Environment

Economic Demand Technological

Forecast Forecast Forecast

Planning Scheduling Controlling

Product/Process Design Operations Scheduling Inventory Management
Investment Planning Production Timetable Manpower Management
Capacity Decision Investment Planning Cost Control

Business
Operations
Figure 3.1.1: Forecasting and business operations Feedback

Unit- 3 Page- 56
 Bangladesh Open University

Economic forecasts address the business cycle by predicting inflation rates,

money supplies, recession, infrastructure, housing and other planning indicators.
Businesses can gain useful ideas about the expected long and intermediate range
business conditions and business growth from such forecasts. Fiscal policy and
monitory policies are also considered in economic forecasting.

Technological forecasts are concerned with rates of technological progress,

which can result in the birth of innovative products requiring new plants and Technological forecasts
equipment. Technological changes provide many businesses with new products are concerned with
technological changes in
and materials to offer, new methods to be implemented, while some businesses business environment.
with obsolete products and/or processes fail in the face of the competition.
Forecasts about technological developments are very important for businesses in
dynamic fields like computer technology, telecommunications services etc.

Demand forecasts are projections of anticipated level of demand for a

Demand forecast is
company’s products or services. These forecasts are also called the sales forecast. often called the sales
Demand forecasts are very important instruments for planning and monitoring forecast.
decisions. In effect, the whole of operations management is guided by demand
forecasts of goods and services. The discussions of this text are generally guided
towards forecasting demand.

Factors Affecting Product Demand Forecast

Demand forecast is given much importance in operations management. Operations manager
Operations managers most often have to forecast customer demand for their often have to forecast
customer demand to
products or services. Managers may need long-run estimations of overall demand
know– what factors are
or short-run estimates of demand for individual products. So, they have to know responsible for their
what factors are responsible for a particular product’s demand. product demand.

Demand of a particular product of a particular company, as you know, is a result

of many forces in the market, like average income of the consumers, price and
availability of related goods, to name a few (Figure 3.1.2). Basically, however, it
depends on the size of the market of that product and the share of the market that
the company captures. A number of forces that are beyond the company’s
control, as well as others that the company can at least influence, act to determine
the level of demand that the company receives. Examples of such forces are
environmental factors, political and legal happenings etc.
Advertising
Marketing
After-sale Service
Product/Service Design DEMAND
Credit policy
Quality

Competing products

Figure 3.1.2: Variables affecting demand for product or service

Status of Economy
Sales are influenced by demand and demand is influenced by a large number of
factors. One factor that influences demand is the status of the economy. As the

Production Operations Management Page- 57

 School of Business

business cycle goes through the phases of depression, recovery, and boom,
demand shifts and changes accordingly (Figure 3.1.3). The following actions will
discuss these influential factors of the business cycle.
• Recession: The contractionary phase of the trade cycle which follows a peak
and ends with the trough. The term recession is generally reserved for the
When the economy mild version of this phase, unlike the slump, which is a severe version. If the
moves into recession, underlying growth rate of output (or income) is sufficiently positive, a
output and income fall
recession may be marked by a fall in the growth rate with no absolute fall in
resulting into a fall in
consumption and output. When the economy moves into recession, output and income fall
investment. resulting into a fall in consumption and investment. Tax revenues also begin
to fall and government expenditure on benefits begins to rise. Wage demands
are moderate as unemployment rises. Imports decline and inflationary
pressures ease.
• Depression: It is the severe downturn phase in the business cycle. Any
contractionary phase of the trade cycle could be called a depression but the
Economic activity is at a term is usually reserved for the most severe cases, such as the downturn in
low in comparison with output in the USA and Europe in particular which occurred in 1929 and the
surrounding years. years following (the great depression). Economic activity is at a low in
comparison with surrounding years. Mass unemployment exists, so,
consumption, investment and imports will be low. There will be few
inflationary pressures in the economy and prices may be falling.
• Recovery: The expansionary phase of the trade cycle during which output
begins to increase. Recovery follows a trough and ends with the peak. In this
period national income and output begin to increase. Unemployment falls.
Consumption, investment and imports begin to rise. Workers feel more
confident about demanding wage increases and inflationary pressures begin
to mount.
• Boom: The expansionary phase of the trade cycle. The term is usually only
applied to particularly fast rates of upward divergence from the secular trend.
National income is high during boom period. It is likely that the economy
will be working at beyond full employment. Consumption and investment
expenditure will be high. Tax revenue will be high. Wages will be rising and
profits increasing. The country will be sucking in imports demanded by
consumers with high incomes and businesses with full order books. There
will also be inflationary pressures in the economy.

Income

Peak/boom
Recovery
Recession
Trough/depression
Time

Figure 3.1.3: The business cycle

• Inflation: Inflation is a sustained rise in the general price level. The inflation
may be Demand pull or Cost push. Demand-pull inflation is caused by excess

Unit- 3 Page- 58
 Bangladesh Open University

demand in the economy. Output increases and unemployment decreases in

such case. Cost-push inflation is caused by increases in the costs of
production in the economy. Such impulse leads to a drop in output. However, Inflation adversely
unemployment may not decrease. During an inflationary period, prices of affects the standard of
some products (for example, food) rise faster than others, inducing diversions living by reducing the
purchasing power of
of resources to those sectors. Inflation adversely affects the standard of living income.
by reducing the purchasing power of income, especially of those who are on
fixed incomes, such as pensioners. Similarly, it adversely affects lenders. If,
households expect prices to be higher in the future, they will be tempted to
bring forward their purchases. However, rising inflation tends to reduce
consumption. The negative effect on consumption more than offsets the
positive effect on consumption caused by the bringing forward of purchases.

Government actions and reactions like careful use of monetary (i.e. determining
supply of money and interest rates) and fiscal policies (i.e. taxation and spending)
are intended to mitigate the severity of the variations of business cycle. However
constant change in demand is still a fact of life in much of the business world.

Activity: Do you think product demand of forecast of your business will

be affected by the socio-economic factors? Justify.

The Product Life Cycle (PLC) and Forecasting

Products are born, they live, and they die. They are cast aside by a changing
society. Normally, a product passes through five stages in its product life cycle:
(i) Development, (ii) Introduction, (iii) Growth, (iv) Maturity or Steady state, and
(v) Decline or Phase out (Figure 3.1.4)1. (also discussed in the unit one lesson 2).
i. Product development: Product development begins when the company
finds and develops a new product idea. During product development,
sales are zero and company’s investment cost add up.
ii. Introduction: A period of fluctuating and slow sales growth as the
product is introduced in the market. Profit is nonexistent in this stage
because of the heavy expenses of product introduction.
iii. Growth: A period of rapid market acceptance and substantial profit
improvement.
iv. Maturity: A period of stability and slowdown in sales growth because the
product has achieved acceptance by most potential buyers. Profit
stabilizes or decline because of increased marketing outlays to defend the
product against competition.
v. Decline: The period when sales show a downward drift and profits erode.

A product, whether good or service, cannot reach the same percentage of

population or target market every day of its existence. Lets think about the
products various stages of the life cycle in the music industry- Digital audio tapes
are in the introductory stage, compact disk are in the growth stage, cassettes are
in the maturity/saturation stage, and eight track tapes are in the decline stage.

1
In many cases we combine development, and testing and/or introduction as one stage.

Production Operations Management Page- 59

 School of Business

Development Testing Growth Maturity Decline

PLC

Figure 3.1.4 The product life cycle (PLC)

Hence the operations manager need to carefully analyze the PLC of any product
Products in the first two during sales forecast, especially for the longer ones. However, products in the
stages of their life cycle first two stages of their life cycle need longer forecasts than that of those in the
need longer forecasts maturity and decline stages. Again introduction and growth stages are quite
than in the maturity and
decline stages. turbulent with gradual increase in sales. While at maturity the demand reaches a
steady state and finally starts to decline. The forecasting are also useful in
projecting different staffing levels, inventory levels, and factory capacity as the
product passes from the first to the last stage.

A large number of other factors influence demands particularly when one

considers secondary influences, that is, factors that affect the customer’s demand
for goods and services, which in turn affects the demand for the company’s
product or service. These can be ethical issues, consumer association,
competitors’ actions, consumers’ preferences, and other social phenomena.

Uses of Forecasting
Forecasting is often used in organizations for three purposes. These are,
• New output introductions: Forecasting is used to decide whether
environmental demand is sufficient to generate the returns desired by the
organizations. If demand exists but at too low a “price” to cover the “costs”
the organization will incur in producing the output then the organization
should reject the opportunity.
• Capacity needs: Forecasting is used to determine long time capacity needs
for facility design. An accurate projection of demand for a number of years
in the future can save the organization great expense in expanding, or
contracting, capacity to accommodate future environmental demands. Due to
competitive forces in the environment, even in the non-profit sector, an
organization that produces inefficiently, because of excess idle capacity, or
insufficiently to meet demand, is courting disaster.
• Production planning: Forecasting is used to ascertain short-term fluctuation
in demand for production planning, workforce scheduling, materials
planning, and so forth. These forecasts are of special importance to
operations management and crucially affect operational productivity,
bottlenecks, master scheduling, meeting promised delivery dates, and other
such issues of concern to top management and the organization as a whole.

How to Develop a Workable Forecasting System?

In the organization production personnel, inventory control personnel, and
marketing personnel should cooperate in determining the forecast. This
cooperation provides checks and balances for the forecasting procedure. In order

Unit- 3 Page- 60
 Bangladesh Open University

to create a workable forecasting system, it is best to assign specific

responsibilities as follows:
• The marketing people should forecast only product groups or promotion
items and items for which a definite trend is expected or has been
detected.
• Production and inventory management people should be responsible for
tracking forecasts, and for reporting deviations of a significant magnitude
to the marketing people for interpretation.
• Transformation of sales forecasts into production plans should be done
by production and inventory control management, since, in order to level
the capacity requirements, the production plan will not necessarily follow
the sales forecast.
• The expeditious interpretation of deviations from forecast, when
reported, is the responsibility of the marketing people. Unless an
interpretation has been put forward and revision is required by the
marketing people, production and purchasing will not stop.
In other words, a workable forecasting system is a system in which production,
inventory and marketing personnel work together, in a coordinated effort, from
the initial forecast, through actual production, down to forecasting revision, as
required.

Need for Judgment in Forecasting

Forecasting, as we have seen, is an art or a special skill rather than an exact
science. The key inputs in a science are constant laws of nature, whereas the key Judgment must be
inputs of forecasting are information, analysis, experience, and informed exercised to see that
judgment. There are no natural laws that make the relationships between demand appropriate forecasting
and other variables continue to behave as they have done in the past. Economic methods are developed
and properly applied.
conditions, competitors’ actions, consumers’ preferences, and other social
phenomena often are whimsical. Judgment must be exercised to see that
appropriate forecasting methods are developed and properly applied. In an
organization judgment is used to determine what data to collect for possible
evaluation and use in forecasting. A forecaster must exercise judgment in
evaluating tabulated data and graphs to discern whether there is some pattern to
demand or whether changes in demand appear to be related to some factors that
can be identified and used to forecast future changes in demand.

Often in the business world a purely subjective approach is used by the

operations managers, where the forecaster forms an opinion of the relative ‘Jury of executive
opinion’ where a group
influences of many factors on demand and estimates the resulting demand. For of executives make
example, in a subjective approach called jury of executive opinion where a group subjective forecasts
relying on experience,
of executives make subjective forecasts relying on experience, intuition and intuition and personal
personal understanding. Even if a mathematical model of forecasting is used, one understanding.
must still decide which variables to evaluate in developing it and which of many
possible equations to use.

On the contrary, if a quantitative measure of forecast is used to evaluate possible

models, judgment must be used to select the appropriate measure, because often
they point to different models as being the best. For instance, a manager may face
a dilemma when they try to choose between exponential smoothing and moving
average (explained in detail in lesson Three) to forecast a certain product’s

Production Operations Management Page- 61

School of Business

demand. Here only statistical techniques like forecasting errors may not be
enough to make a decision.

For as long as the model is used, someone should judge whether the model that
was developed is appropriate for the purpose for which it was developed.
Because it is important to see if the demand data of past periods is still
appropriate to use for current and expected future conditions. That is whether the
same forces are still acting on demand, whether they are exerting the same
relative influence, and whether they can be expected to continue to do so. In
some cases, forecasters may also develop versatile models that will adapt to
changing conditions.

Summary
Forecasting is one of the key elements of operations management. A forecast is
an inference of what is likely to happen in the future. It is estimated by
systematically combining and casting forward in a predetermined way data about
the past. Forecasting involves taking historical data and projecting them into the
future with some sort of mathematical model. It may be subjective or intuitive
prediction of the future or it may be both, i.e., a combination of mathematical
models adjusted by good judgment. But a history of data must exist from which
reasonable inferences can be reached for forecasting. Forecasting prepares an
organization for a common future objective through department activity co-
ordination. It is important for both short-term and long-term planning.

Organizations use three major types of forecasting (economic, technological and

demand forecasting) in planning the future of their operations. All forecasts lead
to demand forecasting. Demand of a particular product of a particular company is
a result of many forces in the market, like average income of the consumers,
price and availability of related goods. Basically, it depends on market size of
that product and the captured market share of that company. A number of forces
that are beyond the company’s control (environmental factors, political and legal
happenings), as well as others that the company can at least influence (Product
life cycle), act to determine the level of demand that the company receives. To
develop a workable forecasting system, production personnel, inventory control
personnel, and marketing personnel should cooperate in determining the forecast.
This cooperation provides checks and balances for the forecasting procedure. The
key inputs of forecasting are information, analysis, experience, and informed
judgment. Economic conditions, competitors’ actions, consumers’ preferences,
and other social phenomena often are whimsical. Judgment must be exercised to
see that appropriate forecasting methods are developed and properly applied.

Unit- 3 Page- 62
 Bangladesh Open University

Discussion questions

1. “Forecasting of sales is the key to many other types of forecasts.”

Explain.
2. “Forecasting is more focused towards science than just being a simple
technique.” Judge the above comment.
3. Why is economic forecast important? In developing an economic
forecast, what problems can be encountered?
4. In recent times, technological forecasting has been given great emphasis.
Why?

Production Operations Management Page- 63

 School of Business

Lesson Two: Different Elements of Forecasting

Lesson Objective
After completing this lesson you will be able to:

Identify different steps in forecasting system

Explain the importance of time horizons in the business forecasting
process

Discuss the forecasting errors

Steps in a Forecasting System

To forecast the demand of an item, you can use the following steps. These steps
present a systematic way of initiating, designing, and implementing a forecasting
task, through regular updating of data.

• Determine the uses of the forecast: At the beginning of a forecasting

process, you have to decide what objectives you are trying to obtain or
why do you need to forecast?
• Select the items that are to be forecasted: After deciding upon the
objectives, the next step becomes selecting the products—the goods or
services—that have to be forecasted about.
• Determine the time horizon of the forecast: The time horizon of the
forecast being made is of great importance in ensuring its effectiveness
and accuracy of the process. You have to make cautious determination
whether it will be short, medium or long term. More discussion on this
topic follows shortly.
• Gather the data needed to make the forecast: Demand forecast may rely
on information of various sorts—historical, causal, environmental, etc.
Only relevant data should be considered to ensure the precision of the
forecast.
• Validate the forecasting model: Among the forecasts prepared using
several techniques, you have to calculate their respective errors and
compare them. The best approach has to be chosen.
• Make the forecasts: Now, you have to set your target production level as
per the forecasts of the best model.
• Implement the result: Once the demand forecast is final, you produce as
per the forecast. Intermittent or periodic checks and feedback are then
conducted on the forecast.

Forecasting Time Horizons

You can distinguish among a variety of forecasting activities by considering-
The time horizon of how far into the future they focus. While detailed forecasts of specific items are
forecasting carries much
weight in operations used for short-term operations, overall product demands are forecasted for longer
management. terms to prepare for investment, capacity and layout (Figure 3.2.1). The time
horizon of forecasting carries much weight in operations management, as
planning and scheduling operations often rely on forecasts of a range of time
spans. Forecasting the time horizon can be divided into three specific ranges:

1. Short range forecast: A forecast that has a time span of one year but is
generally less then three months. The use of such forecast is for purchase

Unit- 3 Page- 64
 Bangladesh Open University

planning, job scheduling, workforce levels, job assignments and

production levels.

2. Medium-range forecast: Medium range or intermediate range forecast

generally spans from three months up to three years. It is useful in sales
planning, production planning and budgeting, cash budgeting, and
analyzing various operating plans.

Forecasted Time
Topic Horizon

Item Short
Demands Run

Aggregate Intermedi-
Demands ate Run

Strategies & Long

Facilities Run

Present Five Years +

Figure 3.2.1: Decisions involved in various time horizons

3. Long range forecast: It is usually used for new products, capital

expenditures, facility location or expansion, and research and
development.

Features of Short, Medium, and Long Range Forecasts

Medium range and long-range forecasts have three features that distinguishes
them from short range forecasts, these are:
• Short-term forecasting usually employs different methodologies than do Short-range forecast
long term ones. Mathematical techniques such as moving average, tends to move more
accurate than long-
exponential smoothing, trend extrapolations are common to short-range range forecast.
projection.
• Short-range forecast tends to move more accurate than long-range
forecast. Factors that influence demand change every day. Thus, as time
horizon lengthens, accuracy of forecast will likely to diminish. Therefore
sales forecast needs to be updated regularly in order to maintain their
value and integrity. After each sales period, the forecast should be
reviewed and revised.

Production Operations Management Page- 65

 School of Business

• Intermediate and long term forecasts deals with more comprehensive

issues and support management decision regarding planning of products,
plants and processes.

Activity: Do you think you should use same types of steps for the short to
the long range business forecasting? Why or why not? Discuss.

Characteristics of Different Ranges of Forecasts

The comparison of different ranges of forecast with respect to their application,
characteristics and forecast techniques is tabulated below (Table 3.2.1).
Table 3.2.1: Comparison of Different Ranges of Forecasts
Ranges of Horizon or Application Characteristics Forecast Method
forecasts Time Span
Long- Generally 5 Business planning: • Broad, general • Technological
range years or • Product planning • Often only • Economic
more • Research qualitative • Demographic
programming • Marketing
• Capital planning studies
• Plant location & • Judgment
expansion
Medium- Generally 1 Aggregate planning: • Numerical
range season or 2 • Capital & cash • Estimate of • Collective
years budgets reliability opinion
• Sales planning needed • Time series
• Production • Not necessarily • Regression
planning at the item • Economic index
• Production & level correlation or
Inventory combination
budgeting • Judgment
Short- Generally Short-run control: • May be at item • Trend
range less than 1 • Adjustment of level for extrapolation
season or 1 production & planning of • Graphical
year employment levels activity • Judgment
• Purchasing • Should be at • Exponential
• Job scheduling adjustment of smoothing
• Project assignment purchases and
• Overtime decisions inventory

Forecasting Errors
Forecasting is mainly done to find the near to possible data of the nature of
The forecast error is the demand or to determine quantity of production based on previous data or
numeric difference of
forecasted demand and hypothesis. For example, if the forecast for the month of November for a product
actual demand. is 326 units, it may be that, in reality the quantity demanded would be less or
more then 326 units; the difference of the two is known to be the error, i.e.,
forecasting error. Therefore the least is the error; the better is the forecast.
Forecast not only helps us to determine the demand but also helps us predict how
much inventory is to be kept; therefore the excess cost can be minimized.

When an organization evaluates different forecasting methods, the operations

managers need a measure of its effectiveness. Forecast error is the scorekeeping
mechanism most commonly used. A calculation of average error made by a
forecast model over time provides a measure of how well a forecast matches the
pattern of past data. This measure is often used as an estimate of how well the

Unit- 3 Page- 66
 Bangladesh Open University

model will fit the demand pattern one is trying to predict. Such a measure of
alternative forecast models provides a basis for comparison to see which model
seems to do the best job. Forecast error is the numeric difference of forecasted
demand (Fi) and actual demand (Di). Obviously organizations never prefer a
forecast method yielding large errors. The most commonly used techniques for
measuring overall forecast error is Mean Absolute Deviation (MAD). This value
is computed by taking the sum of the absolute values of the individual forecast
errors (Fi - Di) and dividing by the number of period of data (n).

∑F
i =1
i − Di
MAD =
n
Here in MAD, we find the difference between the forecasted demand and actual
demand. If the forecast is absolutely correct, i.e., the actual demand equals the
forecasted demand, the error will be zero. As forecasting continues, the forecast
error is recorded and accumulated, period by period. Note that MAD is an
average of the forecast errors. Errors are measured without regard to sign, i.e.,
MAD expresses magnitude, not the direction of errors.

Bias, another measure of forecast error, calculates the forecast error with regard Bias is calculated as the
to direction and shows any tendency consistently to over or under forecast. Bias sum of the actual
is calculated as the sum of the actual forecast error for all periods divided by the forecast error for all
periods divided by the
total number of periods calculated. If the forecast repeatedly overestimates actual total number of periods
demand, Bias will have a positive value; consistent underestimation will be calculated.
indicated by a negative value.
n

∑ (F
i =1
i − Di )
Bias =
n

Assume that, a retail shop owner forecasted the demand for a butter to be 50 units
for each of the next three weeks. The actual demand turned out to be 40, 56, and
70 units. His forecast errors, MAD and Bias, are calculated as follows:

MAD = (|50 – 40| + |50 – 56| + |50 – 70|)/3 = 12 units

Bias = [(50 – 40) + (50 – 56) + (50 – 70)]/3 = - 5.3 unit

Summary
To make reliable forecasts, an organization should have a systematic way of
initiating, designing, and implementing regular updating of data. Forecast itself is
a system comprising of mainly seven steps. At first, organization has to
determine the uses or objectives of forecasting. After deciding upon the
objectives, the next step becomes selecting the products—the goods or
services—that have to be forecasted about. The next steps are determining the
time horizon and collecting data. Determining an appropriate forecasting
technique or model follows data collection. Then the operations managers have
to make the actual forecast based upon the activities taken up so far and
implement the result. Intermittent or periodic checks and feedback that are part
of result implementation are then conducted on the forecast.

Production Operations Management Page- 67

School of Business

Determining appropriate time horizon is critical for forecasting. The time horizon
of forecasting carries much weight in operations management, as planning and
scheduling operations often rely on forecasts of a range of time spans.
Forecasting the time horizon can be divided into three specific ranges: short-
range, medium-range and long-range. Short-range forecasts have a time span of
one year but is generally less then three months. The use of such forecasts are for
purchase planning, job scheduling, workforce levels, job assignments and
production levels. Medium range or intermediate range forecasts generally span
from three months up to three years. It is useful in sales planning, production
planning and budgeting, cash budgeting, and analyzing various operating plans.
Long-range forecasts are usually used for new products, capital expenditures,
facility location or expansion, and research and development. When an
organization evaluates different forecasting methods, the operations manager
needs a measure of its effectiveness. Forecasting error is the scorekeeping
mechanism most commonly used. Forecasting error is the numeric difference of
forecasted demand and actual demand.

Unit- 3 Page- 68
 Bangladesh Open University

Power Shortage
In 1964 the National Power Survey by the FPC indicated that electric utility
companies of the northeastern section of the U.S. had several excess capacity. Case
The forecasts of the northeastern section of the industry were a declining rate of Analysis
growth for peak capacity demand. The past 6% growth trend was moving below
this level, causing additional over-capacity.

Based in part on this study, the Kilowatt Company adopted a plan that assumed a
declining growth pattern. They increased their promotional efforts and reduced
some of their rates to encourage electric usage. In addition, they joined a program
of pooling generating capacity with other utilities in the region. This would
permit switching capacity when demands temporarily created a shortage for one
of the member companies. Thus reducing the capital investment requirements.

The actual growth in demand in 1966 exceeded 7%; in 1967 it was about 1.5%; it
soared to almost 14% in 1968, and was almost 6% in 1969. As a result, instead of
over-capacity, Kilowatt and other utilities in the region found themselves with a
severe shortage of generating capacity. This shortage resulted in requests to
customers to reduce their use of electric power at certain times. It even led to
occasional power blackouts like the famous incident in New York in 1965.

By 1966, Kilowatt found themselves in a position where their peak load capacity
was more than 60,000 kilowatts less than the forecast peak load. Because of the
slower increase in demand and an increase in capacity, in 1967 the forecast
exceeded actual demand by more that 100,000 kilowatts. However, in 1968 and
1969, the forecast was more than 200,000 kilowatts below actual.

Case questions
1. How could such a wide disparity between forecasted usage and actual usage
occur in view of the forecasting techniques available?
2. What factors could affect peak load growth?
3. What factors might have contributed to a greater growth rate than forecasted?
4. What can Kilowatt do to overcome their present difficulties and what should
they do to avoid such large forecasting errors in the future?

Production Operations Management Page- 69

School of Business

Discussion questions
1. How do some investments in forecasting results in a negative return?
2. How accurate should the government forecast need to be and why?
Explain with the help of two such situations.
3. Is there any difference between forecasting demand and forecasting
sales?
4. Explain how an organization can become confused when the distinction
between forecasting and planning is not clear.

Unit- 3 Page- 70
 Bangladesh Open University

Lesson Three: Different Forecasting Approaches/Techniques

Lesson Objective
After completing this lesson you will be able to:

Explain different qualitative forecasting techniques

Identify different quantitative forecasting techniques.

Different Forecasting Approaches

Forecasts can range from instantaneous hunches to elaborate statistical equations.
There is no specific &
There is no way to be sure that a specific technique is best for a specific situation. best technique of
Some are simple and more common, while some complex ones may be more forecasting for a
accurate. Broadly, there are two general approaches to forecasting dominant in specific situation.
operations management. These are,
i) The qualitative approach and
ii) The quantitative approach
Subjective or qualitative intuitive forecast incorporates factors such as the
decision-maker’s guesses, hunches, intuition, emotion and personal experiences,
A combination or
and value system in reaching a forecast. Quantitative forecast employs a variety blending of the two
of mathematical models that use historical data or casual variables to forecast styles is usually most
demand. Some firms use one approach, while some use the other. In practice a effective.
combination or blending of the two styles is usually most effective. Within the
quantitative approach are two basic types of models, which are distinguished, by
the type of the data they use. The demand-based models mainly rely on historical
data about the item that is being forecasted. The causal models, on the other
hand, might formally relate the demand to other variables believed to influence
the demand. The Figure 3.3.1 shows the different forecasting methods, which is
detailed in the following sections.

Forecasts

Qualitative Quantitative

Jury of Sales force Consumer Delphi Nominal Causal or

executives composite Technique Time explanator
market group
series y
survey

Smoothing Decomposition

Simple Moving Weighted Exponential Additive Multi- Regression

average average moving smoothing plicative Economic
average

Figure 3.3.1: Different forecast methods

Production Operations Management Page- 71

 School of Business

The Qualitative Methods of Forecasting

1) Jury of executive opinion: Under this method, the opinions of a group of
high-level managers often in combination with statistical models, are pooled
to arrive at a group estimate of demand.
2) Sales force composite: In this approach each sales person estimates what
sales will be in his/her region. This forecast are then reviewed to ensure they
are realistic and then combined at the district and national levels to reach an
over all forecast.
3) Consumer market survey: This method solicits input from customers or
potential customers, regarding their future purchasing plans. It can help not
only in preparing a forecast, but also in improving product design and
planning for new products.
4) Delphi method: This is a group process intended to achieve a consensus
forecast. There are three different types of participants in the Delphi method–
Delphi method is a
group process intended (1) decision-makers/experts, (2) staff personal/coordinator and (3)
to achieve a consensus respondents. The key to the Delphi technique lies in the coordinator and
forecast avoiding direct experts. The decision makers usually consist of a group of five to ten experts
inter-personal relations.
who will be making actual forecasts, the staff personal or coordinator assists
the decision makers by preparing, distributing, collecting and summarizing a
series of questionnaires and survey results. The respondents are a group of
people often located in different places, whose judgments are valued and are
being sought. This group provides inputs to the decision-makers before the
forecast is made. The procedure works as follows:
a. A coordinator poses a question, in writing, to each expert of a panel.
b. Each expert writes a brief prediction.
c. The coordinator brings the written predictions together, edits them,
and summarizes them.
d. On the basis of the summary, the coordinator writes a new set of
questions and gives them to the experts. These are answered in
writing.
e. Again, the coordinator edits and summarizes the answers, repeating
the process until the coordinator is satisfied with the overall
prediction synthesized from the experts.
An advantage of this method is that direct interpersonal relations are avoided.
Hence personalities do not conflict, nor one strong-willed member dominates
the group. On the other hand, Delphi technique is time consuming.
The nominal group 5) Nominal group discussion technique: Like Delphi technique, the nominal
technique affords group discussion involves a panel of experts. Unlike the Delphi technique,
opportunity for the nominal group technique affords opportunity for discussion among the
discussion among the
experts. experts. Here seven to ten experts are asked to sit around a table in full view
of one another, but they are asked not to communicate with one another. A
group facilitator hands out copies of the question needing a forecast. Each
expert is asked to write down a list of ideas about the question. After a few
minutes, the group facilitator asks each expert in turn to share one idea from
his or her list. A recorder writes each idea on a flip chart so that everyone can
see it. The experts continue to give their ideas (usually between 15-25) in a
round-robin manner until all the ideas have been written on the flipchart. No
discussion takes place in this phase of the meeting.

Unit- 3 Page- 72
 Bangladesh Open University

During the next phase of the meeting, the experts discuss all the ideas that
have been presented. Often similar ideas are combined. When all discussion The advantage of the
has ended, the experts are asked to rank the ideas, in writing, according to nominal group
priority. The consensus is the mathematically derived outcome of the discussion is time
individually rankings. The advantage of the nominal group discussion is time saving.
saving. But this technique can fail if the coordinator fails to allow creativity
and encourage discussion. There is also the risk of getting all the experts
together. This also can become costly if the experts are to be brought from
foreign countries.

Activity: Think about the process of ‘new product design’ of your

firm. In this case what qualitative methods of forecasting you will
chose? Why? Discuss.

Quantitative Method of Forecasting

Naïve (Time Series) Quantitative Models: The time series model predicts the
simple assumption that the future is a function of the past. In other words, they
look at what has happened over a period of time and use a series of past data for
forecasting. Some of the tools are:
a) Simple average
b) Moving average
c) Exponential smoothing

(a) Simple Average

Simple average (X) is the simplest way to forecast assuming that the demand in
the next period will equal to the average of the past demands. Here the demands
of the previous periods are equally weighted.
n

sum of demands for all periods

∑ Di
i =1
X = =
number of periods n
Where,
Di = the demand in the ‘i’th period
n = number of periods
By averaging, we try to detect the pattern or central tendency of demand. The
demand for any one period will probably be above or below the underlying
pattern, and demands for several periods will be dispersed or scattered around the
pattern. Therefore, if we average all past data the extreme values will be offset
and the result will be an average that is representative of the period by reducing
the chances of being misled by gross fluctuations that may occur in any single
period. But we have to be careful about the fact that, if the underlying pattern
changes over time simple averaging will not detect this change.

Example (i)
If a mobile communication company sales 100, 110, and 96 mobile connection in
the month of January, February and March respectively, we can assume that the
demand for April will be 102 [(100+110+96)/3]. This may not be exact but for a
starter it could be on the best-cost effective solution for some organization. In
many cases we give different weights to the past data, especially giving more
weight to the most recent data, making it simple weighted average.

Production Operations Management Page- 73

 School of Business

sum of demands for all periods ∑ WiDi

i =1
X = =
number of periods ∑ Wi
Where,
Di = the demand in the ‘i’th period
Wi = the weights of the ‘i’th period

Example (ii)
In the above forecast example if weights of 20%, 30% and 50% are given for the
demand figures of January, February and March respectively, then the demand
forecast for April will be 101 [(100x0.20 + 110x0.30 + 96x0.50)/1.00].

(b) Moving average

Moving average is
useful if we can A moving average forecast uses a number of recent actual data values from
assume that market several of the most recent periods to generate a forecast. Once the number of past
demands will stay periods to be used in the calculations has been selected, it is held constant.
fairly steady over a Moving average is useful if we can assume that market demands will stay fairly
period of time.
steady over a period of time. A 4-month moving average is found by simply
summing the past 4-month data and divided by 4 and so on. The average
“moves” over time, in that, after each period elapses, the demand for the oldest
period is discarded and the demand for the newest period is added for the next
calculation. This overcomes the major shortcoming of the simple averaging
model. Moving average can be of two types:
i) Simple moving average, where demands of all periods are equally
weighted.
ii) Weighted moving average, which allows varying, not equal, weighting of
the old demands.

A simple moving average (SMA) is calculated as follows:

n

sum of demands for periods

∑ Di
i =1
SMA = =
chosen number of periods n
Where,
Di = the demand in the ‘i’th period
n = chosen number of periods
i = 1 is the oldest period in the n-period average
= n is the most recent period
Example: (i)
Month Actual 3-month moving average
Sales forecast
January 120
February 130
March 110
April 150 (120+130+110)/3 = 120
May - (130+110+150)/3 = 130
When there is a detectable trend or pattern, weights can be used to place more
emphasis on recent values. This makes the technique more responsive to changes
because more recent periods may be more heavily weighted. Selection of weights

Unit- 3 Page- 74
 Bangladesh Open University

requires experience and luck, because there is no set formula to do so. The
Weighted Moving Average (WMA) is calculated by using the following model:
n
WMA = ∑ C t Dt
t =1
Where,
n = chosen number of periods
t = 1 is the oldest period in the n-period average
= n is the most recent period
Di = the demand in the ‘i’th period
Ct = Weight against a particular periodic demand 0 ≤ Ct ≤ 1
∑ Ct = 1

Example (ii)
In the previous example if we weight most recent period twice as heavy as other
two periods by setting, C1 = 0.25, C2 = 0.25, C3 = 0.50, the demand forecast for
the month of May will be; WMA = 0.25 x 130 + 0.25 x 110 + 0.50 x 150 = 135

(c)Exponential smoothing
Exponential smoothing is a sophisticated weighted moving average forecasting
method that is still fairly easy to use. It involves very little record keeping of past
data. Exponential smoothing is distinguishable by the special way it weights each
past demand. The pattern of weights is exponential in form. Demand for the most
recent period is weighted most heavily; the weights placed on successively older
periods decrease exponentially. The basic exponential smoothing formula for
creating a new or updated forecast (Ft) uses two pieces of information:
(i) actual demand for the most recent period (Dt-1)
(ii) most recent demand forecast (Ft-1)

As each time period expires, a new forecast is made using:

Ft = α Dt-1 + (1 - α) Ft-1
Where,
F = Forecast
α = smoothing coefficient (0 ≤ α ≤ 1)
D = demand
t = is the period
t – 1 = immediate previous period.

Why is the Model called Exponential Smoothing?

Since,
F t = αD t-1 + (1- α) F t-1 (1)
Then,
F t-1 = αD t-2 + (1- α) F t-2 (2)
Similarly,
F t-2 = αD t-3 + (1- α) F t-3 (3) and so on ….

Production Operations Management Page- 75

School of Business

If we replace F t-1 in equation (1) by its equivalent value from equation (2),
we get,
F t = αD t-1 + (1- α) [αD t-2 + (1- α) F t-2]
= αD t-1 + α (1- α) D t-2 + (1- α)2 F t-2 (4)

If we continue expanding by replacing F t-2 in equation (4) by its equivalent from

equation (3), we get,

Ft = αD t-1 + α (1- α) D t-2 + α (1- α)2 D t-3 + (1- α)3 F t-3

= α(1- α)0 D t-1 + α (1- α)1 D t-2 + α (1- α)2 D t-3 + (1- α)3 F t-3

The above model self-explains why it is called exponential smoothing (please

note in each of the terms the exponents are increasing).

Smoothing Coefficient Selection

The selection of smoothing coefficient (α) is very important for effective use of
the exponential smoothing model. We have to be careful while taking the value
of α. A high value of α places heavy weight on the most recent demand and a
low α value weights recent demand less heavily. For example, when α = 0.8, the
model becomes,

Ft = α(1- α)0 D t-1 + α (1- α)1 D t-2 + α (1- α)2 D t-3 + (1- α)3 F t-3

= 0.80 D t-1 + 0.16 D t-2 + 0.032 D t-3 +…

On the other hand, when α = 0.2, the model becomes,

Ft = α(1- α)0 D t-1 + α (1- α)1 D t-2 + α (1- α)2 D t-3 + (1- α)3 F t-3

= 0.20 D t-1 + 0.16 D t-2 + 0.128 D t-3 + ……

Hence, for a new product or for items with a dynamic or unstable demand, a
higher value of α (0.7 to 0.9) is more justifiable. In case of a product that is in the
market for a longer period of time and the demand of the product is of less
variance, the value of α is to be taken from the range of 0.1 to 0.3, to smooth out
any sudden noise2 that might have occurred. When demand is slightly unstable,
smoothing coefficient of 0.4, 0.5, 0.6 might provide most accurate forecasts.

Example (i)
Say there is a new product, and the forecast for the month of November has to be
calculated considering that the demand of the product in September is 300 and in
October is 350. The forecast for September was 200 units. As the product is new,
the value of α is to be taken from the range of 0.7 to 0.9. In this case let the value
be 0.7:

2
Noise is any dispersion or deviation of demand from the actual demand pattern.

Unit- 3 Page- 76
 Bangladesh Open University

Therefore,

F October = α D September + (1- α) F September

= 0.7 x 300 + (1-0.7) 200
= 270
F November = α D October + (1- α) F October
= 0.7 x 350 + (1-0.7) 270
= 326
Therefore the forecast for the month of November = 326.

Summary
There are different forecasting approaches that range from instantaneous hunches
to elaborate statistical equations. Some are simple and more common, while
some complex ones may be more accurate. Broadly, two types of general
approaches to forecasting dominates operations management. They are the
qualitative approach and the quantitative approach. The Qualitative Methods
include jury of executive opinion, sales force, composite consumer market
survey, Delphi method and nominal group discussion technique, etc. Among
these techniques, Delphi method and nominal group discussion are most used.
An advantage of this method is that direct interpersonal relations are avoided. On
the other hand, Delphi technique is time consuming. Like Delphi technique, the
nominal group discussion involves a panel of experts. Unlike the Delphi
technique, the nominal group technique affords opportunity for discussion among
the experts. The advantage of the nominal group discussion is time saving. But
this technique can fail if the coordinator fails to allow creativity and encourage
discussion. Quantitative Methods include Naïve (time series) quantitative models
that in turn include simple average, moving average, exponential smoothing etc.
The time series model predicts the simple assumption that the future is a function
of the past. In other words, they look at what has happened over a period of time
and use a series of past data for forecasting.

Production Operations Management Page- 77

 School of Business

Brother’s Packing Box Company

Brother’s Packing Box Company is a small, closely held private corporation

operating for 8 years and located in Chittagong EPZ, Bangladesh. The stock of
Case
the company is divided among three brothers. The principal shareholder is the
Analysis founding brother, Abdul Mannan Khan, who had worked for a similar company
for 20 years in England.

Brother’s Packing Box Company supplies corrugated (folded) cartons to many

RMG factories situated in the EPZ area. Abdul Mannan Khan attributes his
success to the fact that he has been able to capture the EPZ market by supplying
boxes to many small firms as well as the big companies. Abdul Mannan Khan
has enforced the policy that no single customer can account for over 20 percent
of sales as he recognizes the danger of becoming too dependent on one client.
Two of his biggest clients account for 20 percent of sales each, and hence are
limited in their purchases. A.M Khan has convinced the purchasing agents of
these two companies to add other suppliers. This alternative supply protects
Brothers might have in paper shortage etc.

Brother’s Packing Box Company has currently over 250 customers with orders
ranging from a low of 100 boxes to set orders for 5,000 folded packing boxes per
year. Boxes are produced in 16 standard sizes with special printing to customer’s
specifications. Brother’s printing equipment limits their print to two colors. The
standardization and limiting printing allows Brother’s to be price competitive
with other box manufacturers. But they can also provide the service for small and
“emergency” orders.

Such personal service, however, requires tight inventory control and close
production scheduling. So far, A.M. Khan has always forecast demand and
prepared production schedules through experience. But because of the ever-
growing number of accounts and changes in personnel in customer purchasing
departments, the accuracy of his forecasting has been rapidly declining. The
inventory levels of finished boxes along with number of back orders are on the
increase.

Later orders are more common. A second warehouse has recently been leased
due to the overcrowded conditions in the main warehouse. Plans are to move
some of the slower moving boxes to the leased space. There has always been an
increase in demand for boxes prior to the winter season when more and more
orders of knit and woven garments are shipped out to many countries. Such
seasonality in demand has always substantially increased the difficulty of making
a reliable forecast.

Mr. A. M. Khan feels that it is now important to develop an improved forecasting

method. It should take both customer growth and seasonality into consideration.
He believes that if such a method can be applied to forecasting total demand, it
can also be used to forecast demand for the larger customers. The requirements
of the smaller customers could then be integrated to smooth production and
warehousing volume.

Mr. A. M. Khan has complied the following demand data.

Unit- 3 Page- 78
 Bangladesh Open University

Sales (in hundreds of boxes)

Month 1996 1997 1998 1999 2000
January 12 8 12 15 15
February 8 14 8 12 22
March 10 18 18 14 18
April 18 15 13 18 18
May 14 16 14 15 16
June 10 18 1 18 20
July 16 14 817 20 28
August 18 28 20 22 28
September 20 22 25 26 20
October 27 27 28 28 30
November 24 26 18 20 22
December 18 10 18 22 28
195 216 209 230 265

Case questions
1. Develop a forecasting method for Brother’s and forecast total demand for
2001.
2. How might Mr. A. M. Khan improve the accuracy of the forecast?
3. Should Mr. Khan’s experience with the market be factored into the
forecast? If so, how?

Production Operations Management Page- 79

School of Business

Discussion questions
1. Forecasting methods need to be monitored and controlled properly—
why?
2. How does correlation play a key role in forecasting?
3. How can a manager evaluate the reliability and the level of confidence of
a forecast?
4. In what situation you might use quantitative forecasting method though
having historical data?

Unit- 3 Page- 80
 Bangladesh Open University

Lesson Four: Time Series Decomposition

Lesson Objectives
After completing this lesson you will be able to:

Understand the different components of Time Series

Explain the general forms of Time Series Decomposition model

Analysis the casual quantitative model

To systematically analyze historical data for forecasting, we commonly use a

time series analysis. Here we plot demand data on a time scale, study the Analyzing time series
means breaking down
movements, and look for consistent shapes or patterns. A time series demand past data into
might have a variety of components. Analyzing time series means breaking down components and then
past data into components and then projecting them forward. A time series projecting them forward.
typically has four components: trend, seasonality, cycle and random variation.
The Trend component is the general upward or downward movement of the
average level of demand over time. Consider the Figure 3.4.1 which shows the
example of the trend component.

Time
Figure 3.4.1: The Trend Components
The Seasonal component is the recurring fluctuations of demand above and
below the trend value that repeats with a fairly consistent interval. Seasonal
pattern can repeat itself after a period of time, which can be in the form of days,
weeks, months or quarters or with some other interval. A classification of the
seasonal component appears in the Figure 3.4.2.

Period of pattern Season of length No. of season in pattern

Week Day 7
Month Week 4 – 41/3
Month Day 28-31
Year Quarter 4
Year Month 12
Year Week 52

Time
Figure 3.4.2: The seasonal components

Production Operations Management Page- 81

 School of Business

The Cyclical component as shown in Figure 3.4.3, is the recurring upward or

downward movement that repeats with a frequency that is longer than 1 year. As
many as 15 or 20 years of data may be required to determine and describe the
cyclical component. Cycles are pattern in the data that occur every several years,
they are tied into the business cycle and are of major importance in short term
business analyses planning.

Time
Figure 3.4.3: The Cyclical Components

The Random component depicted in Figure 3.4.4, is the series of short, erratic
movements that follow no discernible pattern.

Time
Figure 3.4.4: The Random Components

The pattern is the general shape of the time series. Although some individual data
points do not fall on the pattern, they all tend to cluster around it. To describe the
points clustered about a pattern, we use the term noise. Low noise means all or
most of the points lie very close to the pattern. High noise means many of the
points lie relatively far away from the pattern. When random component of a
time series has fluctuations that deviate substantially from the average level of
demand, it is often useful to smooth out the data by averaging several
observations to make the basic pattern more apparent which is known as Time
Series Smoothing.

Time Series Decomposition model

When a seasonal variation is evident in the demand pattern and the effect of
The Time Series
Decomposition model is
seasonality is to be included in the forecast it is commonly used as Time Series
separation of the overall Decomposition. Therefore, the Time Series Decomposition is separation of the
series into some of its overall series into some of its basic components that are more likely to have
basic components that recognizable and more predictable patterns. These basic components can then be
are more likely to have projected into the future and recombined to form a forecast. The four basic
recognizable and more
predictable patterns. components of time series as we have noted are Trend, Cyclical, Seasonal, and
Random. The decomposition approach is based on the assumption that these
components act independently of one another. If they are projected into the
future, it is assumed that the forces that have caused them to occur in the past
will continue.

Unit- 3 Page- 82
 Bangladesh Open University

General forms of time series decomposition models

Time series has two common models: these are, (1) Multiplicative model and (2)
Additive model.

(1) Multiplicative model: The most common model is the Multiplicative model,
in which forecast is done by multiplying all the time series components.

TF= T x S x C x R

(2) Additive model: In the other model in the time series i.e. the Additive model,
forecast is done through adding the time series components.

TF= T + S + C + R
Here,
TF = time series forecast
T= trend component
S= measure of seasonality, either a ratio or an amount to add
C= measure of cyclical adjustment, either a ratio or an amount to add
R= random component

Visual inspection of a plotted time series is often helpful to determine the type of
model that most appropriately represents the data (Figure 3.4.5).

No trend & no seasonality No trend & seasonality

Linear trend & additive seasonality Linear trend &, multiplicative seasonality

Nonlinear trend & additive seasonality Nonlinear trend & multiplicative seasonality

Figure 3.4.5: Different linear and non-linear additive and multiplicative seasonality

Production Operations Management Page- 83

School of Business

Defining and separating trend and cycle components may necessitate considering
data for a period of 15 to 20 years, which (data) might not be available.
Moreover, very few products remain in stable condition this long—either due to
competitive situation or due to change in product life cycle stage. Hence, it is not
unusual to estimate trend from considering data of 2 years or more as available.
The random component is assumed to be averaged out over the multiple
observations. Consequently, in practice only the seasonal component and the
combination of trend-cycle component is considered. The following example of
time series decomposition will help you to calculate the trend component and the
seasonal coefficient.

Example: The Table 3.4.1 pertains to the quarterly sales data of 1- liter-pack
milk of Shelaidah Dairy Ltd. For 1999 and 1998. The data has been further
calculated to facilitate the time series decomposition. In this example the cyclical
component is not considered, as we did not have enough data for the purpose.
The random component is smoothed out as we took the quarterly sales. Only
trend and seasonal component is taken into account. The 4-quarter moving
average shows an upward trend in the sales.
Table 3.4.1 Quarterly sales data shelaidah dairy ltd.
Year Quarter Quarter Sales 4-quareter moving
number (units) average
1998 First quarter 1 4744 -
(Jan—Mar)
Second quarter 2 1517 -
(Apr—Jun)
Third quarter 3 1526 -
(Jul—Sep)
Fourth quarter 4 2243 2507.50
(Oct—Dec)
1999 First quarter 5 5074 2590.00
(Jan—Mar)
Second quarter 6 1397 2560.00
(Apr—Jun)
Third quarter 7 1948 2665.50
(Jul—Sep)
Fourth quarter 8 5375 3448.50
(Oct—Dec)

From the above figure we will calculate the trend value and the seasonal
coefficient.

• Calculation of trend component

The lowest point of the trend line (at the beginning of quarter # 1) shows sales of
2600 units (appx.) and the highest point (at the end of quarter # 8) shows sales of
3400 units (appx.).
Trend component = (3400—2600) / 8 = 100
Here, the divisor 8 is the number of quarters considered for calculating the trend
line.

Unit- 3 Page- 84
 Bangladesh Open University

• Calculation of seasonal coefficient

Dividing the actual sales by the sales value on the trend line and averaging the
corresponding quarters will give the corresponding seasonal coefficient. If the
actual sales and sales value of the trend line is close to each other a seasonal
coefficient of 1.00 can be taken.

Quarters Year 1 Year 2 Seasonal coefficient (S)

First quarter (4744/2600)=1.82 (5074/3050)=1.66 (1.82+1.66)/2=1.74
Second quarter (1517/2700)=0.56 (1397/3150)=0.44 (0.56+0.44)/2=0.50
Third quarter (1526/2850)=0.54 (1948/3300)=0.59 (0.54+0.59)/2=0.565
Fourth quarter (2243/2950)=0.76 (5375/3400)=1.58 (0.76+1.58)/2=1.17

Calculation of trend value:

Trend value = 2600 + Trend component * (Number of quarter)
= 2600 + 100 * (Number of quarter)

Forecasting for different quarters:

Quarter Quarter Trend Value Seasonal Index Forecast

(Year 2000) Number (T) (S) (T*S)
First quarter 9 2600 + 100 * ( 9 )=3500 1.74 6090
Second quarter 10 2600 + 100 * (10 )=3600 0.50 1800
Third quarter 11 2600 + 100 * (11 )=3700 0.565 2090
Fourth quarter 12 2600 + 100 * (12 )=3800 1.17 4446

Casual Quantitative Model: Casual model, as noted, incorporates dependency of

the variables or factors that might influence the quantity being forecast.

Regression model
Regression model is a causal forecasting technique output that establishes a
relationship between variables. There is one dependent variable and one or more
explanatory variables. Historical data establishes a functional relationship
between the two variables. If there is one explanatory variable it is called simple
regression, otherwise, it becomes multiple regression. If the model takes the
shape of a linear equation, we call it simple linear regression or multiple linear
regression models. Normally least square curve fitting technique is used to
develop the models for fitting a line to a set of points.

Simple Linear Regression

The simplest and most widely used form of regression involves a linear
relationship between two variables. A plot of the values might appear like that in
Figure 3.4.6.

Production Operations Management Page- 85

School of Business

Y Regression line

x x x
sample points x x x
x x x
x x x
x x x
x

X
Figure 3.4.6: A straight line fitted to a set of sample points

The objective in linear regression is to obtain an equation of a straight line that

minimizes the sum of squared vertical deviations of data points from the line.
This least square curve has the equation:

Y = a + bX
Where,
Y → predicted (dependent) variable [normally drawn on the Y-axis]
X → explanatory (independent) variable [normally drawn on the X-axis]
a → value of predicted variable (Y) when x = 0 (Y-intercept value of the
line)
b → slope of the line (change in Y corresponding to a unit change in X)

The coefficients a and b of the regression line are computed using the following
two equations:

n( ∑ xy ) − (∑ x )(∑ y )
b=
n(∑ x 2 ) − ( ∑ x ) 2
where, n → number of paired observations

a=
∑ y − b∑ x = y − b x − −

n
− −
where, y → mean of x vlues & x → mean of y values

Example
XYZ Fashions has a chain of twelve stores in Dhaka. Sales figure and profits for
the stores are shown in the following table. (Figures are in ‘000taka).

Sales (X) Profits (Y) Sales (X) Profits (Y)

70 1.5 120 2.0
20 1.0 140 2.7
60 1.3 200 4.4
40 1.5 150 3.4
140 2.5 70 1.7
150 2.7
160 2.4

Unit- 3 Page- 86
 Bangladesh Open University

a) Does a linear model seem reasonable?

b) Develop a regression model for the data and predict profit for a store
assuming sales of Tk. 100,000.

Solution
a) If we plot the points (Figure 3.4.7) we can see that the points seem to scatter
around a straight line. Hence, we can conclude that a linear model seems
reasonable.

5.0 - x

4.0 - x
X
X
3.0 - x x
X x
2.0 - x X x
x
1.0 -

0-
0 20 40 60 80 100 120 140 160

Figure 3.4.7: Scatter Diagram for the Sales & Profits Figures of XYZ Fashions

b) Now for developing the regression model let’s compute the quantities
Σx, Σy, Σxy, and Σx2.

x y xy x2
70 1.5 105 4,900
20 1.0 20 400
60 1.3 78 3,600
40 1.5 60 1,600
140 2.5 350 19,600
150 2.7 405 22,500
160 2.4 384 25,600
120 2.0 240 14,400
140 2.7 378 19,600
200 4.4 880 40,000
150 3.4 510 22,500
70 1.7 119 4,900
2
Σx =1320 Σy = 27.1 Σxy = 3529 Σx = 179,600

Substituting into equations, we find:

n( ∑ xy ) − (∑ x )(∑ y ) 12(3529) − 1320(27.1)

b= = = 0.01593
n(∑ x ) − ( ∑ x )
2 2
12(179,600) − 1320 2
Thus the regression equation is: Y = 0.506 + 0.01593X

Production Operations Management Page- 87

School of Business

It is important to note that the regression line should only be used for the range of
values from which it was developed; the relationship may be non-linear outside
the range.

a=
∑ y − b∑ x = y − b x = 2.26 − 0.01593 x 110 = 0.506
− −

n
− −
where, y → mean of x vlues & x → mean of y values

Trend Equation
A linear trend equation has the form:
Yt = a + bt
where,
t = specified number of time period from t=0
Yt = forecast for period t
a = value of Yt at t=0
b = slope of the line.
This equation is identical to the linear regression line, except that t replaces x in
the equation. Hence,

n( ∑ ty ) − ( ∑ t )(∑ y )
b=
n( ∑ t 2 ) − ( ∑ t ) 2
where, n → Number of periods, y → Value of the time series

Example
Consider the trend equation, Yt = 45 + 5t. Here the value of Yt when t = 0 is 45,
and the slope of the line is 5. The slope indicates that, on the average, the value
of Yt will increase by 5 units for each one-unit increase in t.

a=
∑ y − b∑ t = y − b t − −

n
− −
where, y → mean of y values & t → mean of t values

Unit- 3 Page- 88
 Bangladesh Open University

Summary
A summary of advantages and disadvantages of the prominent forecasting
techniques for comparison shows the following table.

Model Name Advantages Disadvantages

Qualitative Approach
Jury of Executive • Quick and simple to use • Subject to biases of opinion
Opinion and judgement
• Brings in a variety of
viewpoints • Unsupported by data
Sales force • Salespeople’s intimate • Salesmen may understate or
composite market knowledge is overstate the forecast to make
technique utilized it favorable
• Participation in forecasting
motivates sales-force
Delphi Technique • Brings in various • No scope for interaction
viewpoints • Time consuming
• Interpersonal relations are • Ultimately the coordinator
avoided takes the decision
• Personality conflict and
domination by any member
is avoided
Nominal Group • No biases or influences • More expensive, like cost of
Discussion accommodation
• Brings in different opinions
• Takes less time • Technique can fail if the
coordinator fails to allow
• There is interaction among creativity and encourage
participants discussion
• Chances for biases
Naïve (time series) Quantitative Models
Simple average • Extreme values are offset • Cannot be used for seasonal
• Gross fluctuations are and unstable products
avoided • Cannot detect changes of
patterns over time
Moving average • Most recent data are • Data of too many periods
considered may make it less responsive
to changes in demand pattern
Exponential • Needs very little record • Inexperienced smoothing
Smoothing keeping coefficient (α) selection may
lead to inaccurate forecast
Casual Quantitative Models
Regression • Usually useful for long term • The process is Complex and
Analysis trend analysis cumbersome
• Requires much data

Production Operations Management Page- 89