FORECASTING FOR OPERATIONS

What is forecasting?
Forecasting is a planning tool by which historical data is used to predict the direction of future trends. Forecasting
is a technique that uses historical data to make informed decisions about future events or conditions.
What is forecasting in business?
Forecasting is a decision-making tool used by many businesses to help in budgeting, planning, and estimating
future growth. In the simplest terms, forecasting is the attempt to predict future outcomes based on past events
and management insight.
BASIC CATEGORIES OF FORECASTING METHODS
Forecasting methods can be divided into three main categories:

 Extrapolative or time series methods

 Causal or explanatory methods
 Qualitative or judgmental methods
Extrapolative methods use the past history of demand in making a forecast for the future. The objective of these
methods is to identify the pattern in historical data and extrapolate this pattern for the future. This process might
seem like driving while looking only through a rare view mirror. However, if the time horizon for which the
forecast is made is short, extrapolative methods perform quite well.
Casual methods of forecasting assume that the demand for an item depends on one or more independent factors
(e.g., price, advertising, competitor’s price, etc.). These methods seek to establish a relationship between the
variable to be forecasted and independent variables. Once this relationship is established, future values can be
forecasted by simply plugging in the appropriate values for the independent variables.
Judgmental methods rely on the expert’s (or manager’s) opinion in the making a prediction for the future. These
methods are useful for medium to long range forecasting tasks. The use of judgement in forecasting, at first
blush, sounds unscientific and ad hoc. However, when the past data are unavailable or not representative of the
future, there are few alternatives other than using the informed opinion of knowledgeable people. There are,
however, good ways and bad ways to solicit judgments for making a forecast. We will discuss some approaches
that structure and formalize the process of soliciting judgments so that individual biases are minimized. Often, in
operations situations, judgmental methods are employed in conjunction with extrapolative or casual method.
EXTRAPOLATIVE METHODS
Seek to identify patterns in past data. Most of these depend on four components of demand: horizontal, trend,
seasonal, and cyclical. The appropriateness of an extrapolative method will depend on which components of
demand are operating in given situations.
Components of Demand
The horizontal component of demand exists when demand fluctuates about and average demand. The average
demand remains constant and does not consistently increase or decrease. The sales of a product in the mature
stage of the product life cycle may show a horizontal demand pattern.
The trend component of demand refers to a sustained increase or decrease in demand from one period to the
next. For example, if the average monthly demand for a product has increased 10 to 15 percent in each of the
past few years, then an upward trend in demand exists. The sales of product in the growth stage of the product
life cycle tend to show an upward trend, whereas those in decline tend to show a downward trend.
The seasonal component of demand pertains to the influence of seasonal factors that impact demand positively
or negatively. For example, the sales of snow blowers will be higher in winter months and lower in summer
months every year, indicating a seasonal component in the demand for snow blowers.
The cyclical component of demand is similar to the seasonal component except that seasonality occurs at regular
intervals and is of constant length, whereas the cyclic component varies in both time and duration of occurrence.
For example, the impact of a recession on the demand for a product will be reflected by the cyclic component.
Recession occur at irregular intervals and the length of time a recession lasts varies. This component is present in
most economic data, such as GNP, personal income, and industry sales such as automobiles and major
appliances.

Exponential Smoothing Methods

In these methods, the weight assigned to the previous period’s demand decreases exponentially as that data
gets older. Thus, recent demand data receives a higher weight than does older demand data.
Classification of Exponential Smoothing Models
The basic exponential smoothing model does not incorporate trend and seasonality components in the
demand data for forecasting. Winters (1960) and Pegels (1969) have develop models that are capable of
incorporating these effects.
Demand data can have no trend, a linear trend, or a ratio trend. A linear trend is defined by the units per
period by which the expected demand changes (increases or decreases). An example of a linear trend would be
an average increase in demand of 10 units per period. A ratio trend is defined by the percentage of demand by
which demand is changing, as, for example, when demand increases at an average rate of 15 percent per period.
Thus, in a ratio trend, there is a compounding effect on the absolute value of demand.
Similarly, demand data can have no seasonality, additive seasonality, or ratio seasonality. The distinction
between additive and ratio seasonality is that the former is defined as the number of units above the average and
the latter is defines as a ratio of period sales to average sales. With additive seasonality, therefore, the
amplitudes of the demand pattern remain more or less constant; with ratio seasonality, the amplitudes become
more pronounced as demand levels increase over time.
CASUAL OR EXPLANATORY METHODS
When we have enough historical data and experience, it may be possible to relate forecasts to factors in the
economy that cause the trends, seasonals, and fluctuations. Thus, if we can measure these causal factors and can
be determined their relationships to the product or service of interest, we can compute forecasts of considerable
accuracy.
The factors used in causal models are of several types: disposable income, new marriages, housing starts,
inventories, and cost-of- living indices as well as predictions of dynamic factors and / or disturbances, such as
strikes, actions of competitors, and sales promotion campaigns. The causal forecasting model express
mathematical relationships between the causal factors and the demand for the item being forecast. There are
two general types of causal models: regression analysis and econometric methods.
Regression Analysis
Forecasting based on regression methods establishes a forecasting function called regression equation. The
regression expression equation expresses the series to be forecast in terms of other series that presumably
control or cause sales to increase or decrease.
QUALITATIVE OR JUDGMENTAL METHODS
In this age of management science and computers, why must we resort to qualitative methods to make some
of the most important predictions of future demands for product or services, predictions on which great risks
involving large investments in facilities and market development hinge?
The Delphi Method
Technological forecasting is a term used in conjunction with the lonest term predictions, and the Delphi
technique is the methodology often used as the vehicle for such forecasting. The objective of the Delphi
technique is to probe into the future in hopes of anticipating new products and processes in the rapidly changing
environment of today’s culture and economy. In the short range, such predictions can also be used to estimate
market sizes and timing.
The Delphi technique draws on a panel expert in a way that eliminates the potential dominance of the most
prestigious, the most verbal, and the best salespeople. The Delphi technique was first developed by the RAND
Corporation as a means of achieving these kinds of results.
The panel experts can organize in a various way, and it often includes individuals from both inside and outside
of the organization. Each panel member is an expert on some aspects of the problem, but no one is an expert on
an entire problem. In general, the procedures involve the ff.
1. Each expert in the groups makes independent predictions in the form of brief statements.
2. The coordinators edit and clarifies these statements.
3. The coordinator provides a series of written questions to the experts that include feedback supplied by
the other experts.
4. Steps 1 to 3 are repeated several times. In practice, convergence is usually obtained after a small number
of rounds.
Market Surveys
Market surveys and analyses of consumer behavior have become quite sophisticated, and the data that result
are extremely valuable inputs for predicting market demand. In general, the methods involve the use of
questionnaires, consumer panels, and tests of new products and services. The field is a specialty in itself and
beyond our scope. There is a considerable amount of literature dealing with the estimation of new product
performance based on consumer panels (Ahl, 1970 ) and analytical approaches (Bass, 1969; Claycamp and Liddy,
1969) as well as simulation and other techniques (Bass, King, and Pessemeier, 1968). Proposed products and
services can be compared with the products and known plans of competitors, and new market segments can be
exploited with variations of product designs and quality levels.
Historical Analogy and Life Cycle Analysis
Market research studies can sometimes by supplemented by referring to the performance of an ancestor of
the product or service under consideration and applying an analysis of the product life cycle curve. For example,
the assumption can be made that color television would follow the general sales pattern experienced with black
and white television but that it would take twice as long to reach a steady state ( Chambers, Mullick, and Smith,
1971). Such comparison provides guidelines during initial planning phases and maybe supplemented by other
kind of analyses and studies as actual demand becomes known. Chase and Aquilano (1977) focus their attention
on the life cycle of products in studying the problems of production management.
Scenario Based Forecasting
Long- range forecasting has been criticized because of its lack of predictive accuracy. Ascher (1978) has
documented that the results of long- range forecasts in several areas have been inaccurate because the core
assumption upon which the forecasts were predicted proved to be wrong. It is reasonable to assume that the
further the forecast horizon, the less accurate the forecast is likely to be. Long- range forecast, however, are an
integral part of long- range planning, so they must be made.
ACCURACY OF FORECASTING METHOD
The accuracy of forecasting methods is one of the most important criteria for comparing alternative
methods of forecasting. Cost, ease of application, and specific requirements of a planning situation are other
factors that influence the choice of a forecasting method. Mak-ridakis and Winkler (1983) empirically estimated
the impact of the number and choices of forecasting methods of accuracy of forecast when the results of the
methods used are simply averaged to provide the final forecast. Their main findings were as follows:

 Forecasting accuracy improves as the forecasts from more methods are combined to provide the
final forecast; the marginal impact of including an additional method decreases as the number of
methods increase.
 The risk of larger error in forecasting that might result from the choice of a wrong method is
diminished when the results of two or more methods are combined.
 Variability in forecast accuracy among different combinations of forecasting methods decreases as
the number of methods increases

OTHER FORMS OF FORECASTING

Decomposition methods are extrapolative in nature, but they differ from exponential smoothing methods.
The key difference is that, instead of extrapolating a single pattern as is the case with exponential smoothing,
each component of demand is extrapolated separately.
Box and Jenkins (1970) have proposed a framework for analyzing and modeling time series data. This
framework is based on well- developed statistical theory. It provides an approach for identifying patterns in data
and methodology for extrapolating this pattern into the future. The basic framework consists of three stages:
identification, estimation, and diagnostic testing.
The Fourier series forecasting method represents time series using mathematical function consisting of a
constant term plus the sum of several sine and cosine term. The method is useful when data have seasonal
patterns.

IMPLICATIONS FOR THE MANAGER

If there is a single, most important set of data for managers, it is forecast data. Virtually every important
decision in operations depends in some measure on a forecast demand. Managers are keenly aware of their
dependence on forecasts. Indeed, a great deal of executive time is spent worrying about trends in economic and
political affairs and how events affect the demand for products or services. Quantitative techniques are probably
least effective in “calling the return” that may result in sharply higher or lower demand because quantitative
methods are based on historical data.