CHAPTER 2.

FORECASTING FEATURES COMMON TO ALL

FORECASTS
A Forecast is any attempt to try and
predict the future based on quantitative A wide variety of forecasting techniques
means (using quantitative techniques and are in use. In many respects, they are
formulas) or non-quantitative means such different from each other. Nonetheless,
as intuition, trends (experience), facts or certain features are common to all
opinion. The main purpose of forecasting forecasts:
is to be prepared based on the possible
future. 1. Forecasting techniques generally
assume that the same underlying causal
There are two uses for a forecast in the system that existed in the past will
production setting of the organization: the continue to exist in the future.
first is to help the managers plan the
system and the second is to help the 2. Forecasts are rarely perfect; actual
managers plan the use of the system. results usually differ from predicted
Planning the system involves long-range values. No one can predict precisely how
plans about the types of products and an often large number of related factors
services to offer, what facilities and will impinge upon the variable in question;
equipment to have, where to locate and so this, and the presence of randomness,
on. Planning the use of the system refers precludes a perfect forecast. In short,
to short-range and intermediate-range allowances should be made for
planning, which involves tasks such as inaccuracies.
planning inventory and work force levels,
planning purchasing and production, 3. Forecasts for groups of items tend to be
budgeting and scheduling. more accurate than forecasts for individual
items because forecasting errors among
Forecasting is not only used to predict items in a group usually have a canceling
demand. We can also use forecasting to effect.
predict profits, revenues, costs,
productivity changes, prices and 4. Forecast accuracy decreases as the
availability of energy and raw materials, time period covered by the forecast – the
interest rates, movements of key time horizon - increases.
economic indicators (e.g. GNP, inflation,
government borrowings) and prices of
stocks and bonds.

For simplicity purposes, we will only

discuss forecasting of demand in this
chapter.
ELEMENTS OF A GOOD FORECAST 3. Select a forecasting technique.

A properly prepared forecast should fulfill 4. Gather and analyze the appropriate
certain requirements: data.

1. The forecast should be timely. 5. Prepare the forecast.

2. The forecast should be accurate, and 6. Monitor the forecast. A forecast has to
the degree of accuracy should be stated. be monitored to determine whether it is
It will enable users to plan for possible performing in a satisfactory manner. If not,
errors and will provide a basis for reexamine the method, assumptions,
comparing alternative forecasts. validity of data, and so on; modify as
needed; and prepare a revised forecast.
3. The forecast should be reliable. It
should work consistently.

4. The forecast should be expressed in APPROACHES TO FORECASTING

meaningful units.
Forecasts Based on Judgment and
5. The forecast should be in writing. Opinion

6. The forecasting technique should be Judgmental forecasts rely on analysis of

simple to understand and easy to use. subjective inputs obtained from various
sources, such as consumer surveys, the
sales staff, managers and executives, and
panels of experts. Quite frequently, these
STEPS IN THE FORECASTING sources provide insights that are not
PROCESS otherwise available.

There are five basic steps in the

forecasting process: Forecasts Based on Judgment and
Opinion
1. Determine the purpose of the
forecast. What is its purpose and when 1. Executive Opinions. – this is a
will it be needed? This will provide an judgmental forecast conducted by a small
indication of the level of detail required in group of upper level managers or
the forecast, the amount of resources executives.
(manpower, computer time, pesos, etc.)
that can be justified, and the level of 2. Direct Consumer Contact
accuracy necessary. Composites. – this is a judgmental
forecast shared by the sales staff of the
2. Establish a time horizon. The forecast company. They are considered as a good
must indicate a time limit, keeping in mind source of information since they are the
that accuracy decreases as the time ones who are in constant contact with the
horizon increases. consumers.
3. Consumer Surveys. – interviews, 1. Trend. This refers to a gradual,
research, questionnaires that are long-term upward or downward movement
answered by the consumers themselves. in the data. Population shifts, changing
incomes and cultural changes often
4. Other approaches. – Another account for such movements.
approach is called the Delphi Method. It
involves managers and staff completing a 2. Seasonality. This refers to short-term,
series of questionnaires, each developed fairly regular variations generally related to
from the previous one, to achieve a factors such as weather, holidays, and
consensus forecast. vacations. Restaurants, supermarkets and
theaters experience weekly and even daily
“seasonal” variations.

Forecasts Based on Time Series Data 3. Cycles. These are wavelike variations
of more than one year’s duration. These
A time-series is a time-ordered sequence are often related to a variety of economic,
of observations taken at regular intervals political and even agricultural conditions.
over a period of time (e.g. hourly, daily,
weekly, monthly, quarterly, semi-annually, 4. Irregular variations. These are due to
annually, etc). The data may be unusual circumstances such as severe
measurements of demand, earnings, weather conditions, strikes, or a major
profits, shipments, accidents, output, change in a product or service. They do
precipitation, productivity, and the not reflect typical behavior and they
consumer price index. Forecasting should be clearly identified from the data.
techniques based on time series data are
made on the assumption that future 5. Random variations. These are any
values of the series can be estimated from residual variations that remain after all
past values. other behaviors have been accounted for.

Although no attempt is made to identify

variables that influence the series, these
methods are widely used, often with quite Techniques for Averaging
satisfactory results. Analysis of time series
data can requires the analyst to identify Historical data typically contain a certain
the underlying behavior of the series. This amount of random variation, or noise, that
can often be accomplished by merely tends to obscure systematic movements in
plotting the data and visually examining the data. This randomness arises from the
the plot. One or more patterns may combined influence of many - - perhaps a
appear: trends, seasonal variations, cycles great many - - relatively unimportant
and constant variations (variations around factors, and it cannot be reliably predicted.
an average). In addition, there can be Ideally, it would be desirable to completely
random or irregular variations. These remove any randomness from the data
variations can be described as follows: and leave only “real” variations, such as
changes in the
demand.
As a practical matter, however, it is usually Three techniques for averaging will be
impossible to distinguish between these discussed in this section.
two kinds of variations, so the best one
can hope for is that the small variations 1. Naïve forecasts. – The forecast for any
are random and the large variations are period equals the previous period’s actual
“real”. value.

Averaging techniques smooth fluctuations 2. Moving averages – Technique that

in a time series because the individual averages a number of recent actual
highs and lows in the data offset each values, updated as new values
other when they are combined into an become available.
average. A forecast based on an average
thus tends to exhibit less variety than the 3. Weighted Moving Averages –
original data. This can be advantageous Technique that assigns percentages to the
because many of these movements historical data.
merely reflect random variability rather
than a true change in level, or trend, in the M.A.D. Mean Average Deviation
series. The use of standard deviation to find the
potential range of forecasting error of a
Moreover, because responding to given technique. The most appropriate
changes in expected demand often entails technique is one that yields the lowest
considerable cost (e.g. changes in M.A.D.
production rate, changes in the size of the
workforce, inventory changes), it is
desirable to avoid reacting to minor
variations.

Thus, minor variations are treated as

random variations, whereas larger
variations are more likely to reflect “real” Where: Σ/e/ = total of the absolute values
changes, although these too are of the forecasting errors
smoothed to a certain degree. Averaging
techniques generate forecasts that reflect n = the number of forecasts
recent values of a time series (e.g. the
average value over the last several note: absolute value means that you will
periods). These techniques work best disregard the negative signs
when a series tend to vary around an
average (ideal), although they can also
handle step changes or gradual
changes (upward or downward) in the
level of the series.