Selecting the Appropriate Forecasting Method: The Toolbox Approach

Today, most companies seem to use simple methods that are easy to comprehend and
mostly those that involve judgment by company employees. On the other hand, most
forecast practitioners generally use forecasting methods with which their decision
makers feel comfortable, even though these methods may not be the most effective
ones. One method widely used results in goal setting rather than forecasting. Here,
companies begin their planning process with a corporate goal to increase sales by
some percentage. This target often comes down directly from senior management as
an edict. Everyone then proceeds to back into their targets based on what each
business unit manager thinks they can deliver. If they don’t meet their prospective
targeted goal when totaled, senior management either assigns individual targets to
each business unit or puts a financial plug in place hoping someone will over-deliver.

Underlying Methodology Assumption

The basic assumption underlying the application of any forecasting method (statistical
or judgmental) is that the actual outcome observed will follow some pattern associated
with seasonality, trend, and/or causal relationships plus some random influences. This
is algebraically written as: Actual Outcome = Pattern + Randomness. This simple
equation is really saying that even when the average patterns of the underlying data
have been identified, some deviation will exist between the forecast and the actual. Our
purpose as practitioners is to minimize these deviations (errors) in the forecast by
selecting the appropriate method.

As forecast practitioners, we believe our primary responsibility is to provide senior

management with only accurate point estimates. As such, we tend to rely on the one-
methodology -fits-all-situations forecasting mentality when, in fact, we should
consider ourselves business analysts who use an array of methods to predict the future
depending on each individual situation. Like a medical doctor, we should carry a
toolbox of generic methods and, depending on the ailment, choose the technique (tool)
that best fits the situation. Unfortunately, before we can apply the toolbox approach to
forecasting, we need to better understand the methods available along with their
advantages and disadvantages.

Available Methodologies

Most forecasting methods fall into two broad categories: qualitative (also known as
“judgmental”) methods—those that rely on the subjective assessments of a person or
group of persons—and quantitative (also known as “mathematical” or “objective”)
methods—those that rely on past sales history alone or those that arc built ona
relationship between past sales and some other variable(s). Although one may have a
firm grasp on these two categories of methods, it is important to realize that some
subjective assessment is usually involved in all types of sales forecasts. Subjectively
derived forecasts use intuitive or gut feelings based on the experience and savvy of
people who understand not only what is presently occurring in the marketplace but
also what is likely to occur. The most widely used judgmental techniques are
independent judgment, committee judgment, sales force estimates (also known as
“sales force composites”), and juries of executive opinion.

Judgmental methods are often perceived as “last resort” techniques (i.e., “We don’t
have the hard data needed to use some mathematical technique, so we are forced to
make a rough estimate.”). Often, judgmental methods provide very accurate forecasts.
The major advantages of judgmental methods are their low cost to develop (there’s no
need for expensive computer hardware/software); executives usually have a solid
understanding of the broad-based factors and how they affect sales demand; and sales
forecasts can be developed fairly quickly. But, they are always biased toward the user
group that develops them; they are not consistently accurate over time due to the
subjective nature of development; some executives may not really understand the
firm’s sales situation since they are too far removed from the actual marketplace; and
they are generally not well suited for firms with a large number of products (i.e.,
stockkeeping units [SKUs]).

There are two broad segments within the quantitative category: time series (which I
refer to as reactive or one- dimensional methods) and causal (which I refer to as
multidimensional or proactive methods). Time series are techniques built on the
premise that future sales will mimic the pattern of past sales. In other words, time
series methods rely on the identification of patterns (i.e., seasonality, trend, and/or
cyclical) within the past sales history of items being forecasted and assume those
patterns will continue into the future. The most basic time series method is called the
“naive” model, because it assumes future sales will replicate past sales. It is a little
naive to assume that sales will be exactly the same as the prior year given the dynamics
of the marketplace. Another time series method is called “moving averaging.” Moving
average techniques are also called “smoothing” models since they level out small,
random fluctuations. The most widely used time series methods are called
“exponential smoothing.” The basic premise of exponential smoothing is that the sales
volumes for the most recent periods have more impact on the forecast and therefore
should be given more weight. Among the array of exponential smoothing techniques
are Brown's double exponential smoothing, Holt's two-parameter exponential
smoothing, and Winter's three parameter exponential smoothing.
A more advanced time series method is called “decomposition.” This technique is
based on the assumption that sales are affected by four basic elements: trend,
seasonal influences, cyclical influences, and random (irregular) influences. The most
advanced time series technique is called Box-Jenkins (also known as auto regressive
integrated moving average [ARIMA] models), which combines the key elements from
both time series and regression models. Here, autocorrelation coefficients identify the
association between a variable at one time period and with the same variable at some
other time period. Thus, autocorrelation is really the correlation of a variable with itself.
In spite of its forecasting success, the Box-Jenkins approach is still the least used time
series method. This is because of its complexity that discourages many forecast
practitioners and managers from using it.

The major advantages of time series methods are: They are well suited to situations
where sales forecasts are needed for a large number of products; they work very well
for products with fairly stable sales; they can smooth out small, random fluctuations;
they are simple to understand and use; they can be easily systematized and require
little data storage; software packages for such methods are readily available; and they
are generally good at short- term forecasting. The major disadvantages of time series
methods are: They require a large amount of historical data; they adjust slowly to
changes in sales; a great deal of searching may be needed to find the weighted (alpha)
value; they usually fall apart when the forecast horizon is long; and forecasts can result
in great error because of large fluctuations in current data.

The basic premise of causal models is that future sales of a particular product are
closely associated with changes in some other variable(s). For example, changes in
sales can be associated with changes in price, advertising, sales promotions, and
merchandising. Therefore, once the nature of that association is quantified, it can be
used to forecast sales. The most widely used causal methods are simple regression,
multiple regression, and econometrics. In recent years—due to the development of
more advanced software—a technique called robust regression has been gaining wide
acceptance by forecast practitioners. The major advantages of causal methods are:
They are available in most software packages; they are inexpensive to run on
computers; they are covered in most statistics courses, so they have become
increasingly familiar to managers; they provide accurate short- and medium-term
forecasts; and they are capable of supporting “what-if” analyses. The major
disadvantages of causal models are: Their forecasting accuracy depends on a
consistent relationship between independent and dependent variables; an accurate
estimate of the independent variable is crucial; many managers—due to a lack of
understanding—view it as a “black box” technique; they are more time-intensive to
develop and require a strong understanding of statistics; they require larger data
storage and are less easily systematized; and they tend to be more expensive to build
and maintain.

Selecting Method(s) Based on Product Portfolio Analysis

A forecast practitioner developing a sales forecast can choose from among all the
methods discussed, but all these methods are not equally effective for any given
situation. The key factors to consider evolve around the completeness and stability of
the data set(s) being forecast. Like a doctor, we need to assess each situation and
prescribe the appropriate treatment. In this case, the practitioner should use an
appropriate method(s). Unfortunately, most systems and most practitioners use one
methodology to forecast all their products, which results in poor performance, rather
than apply the appropriate method(s) depending on the situation.

If you consider your product portfolio as falling on two intersecting planes—one that is
either incomplete or complete, and the other either unstable or stable—you can
determine which method to apply in almost any situation. Incomplete data refer to
having limited sales history for a particular item, not having all the required causal
variables, and/or not having any data, for that matter. Complete refers to having all the
required data for a particular item, as well as all the causal variables. Unstable refers
to data having aberrations or being random with no distinct pattern associated with it.
Stable refers to data that have a distinct pattern associated with it, such as seasonality
and trend. Understanding the strengths and limitations of the methods discussed, you
can plot them in their corresponding quadrants based on data availability within your
product portfolio. See Figure 1 for the details.

With the help of this diagram, you can categorize methods based on data availability
within your product portfolio. Those products that have incomplete data and are
unstable can only be forecasted using sales force composites, independent judgment,
committee judgment, and simple moving average, most of which are qualitative
methods or judgmental (see lower left-hand quadrant). On the other hand, if you have
very stable data but its historical data are incomplete, then you are forced to use time
series methods. Such methods would include Winter's, Box- Jenkins, and Census X 11
(see upper left-hand quadrant). 1f you have complete data, including causal
information, and the data are stable, indicating a distinct pattern, you can use simple
regression, multiple regression, and/or econometric techniques (see upper right-hand
quadrant). Finally, if the data are complete, including all causal variables, but the data
are unstable, displaying aberrations or outliers, you are forced to use robust regression
(see lower right-hand quadrant), which adjusts for outliers using a weighting approach.
The approach is too complicated to explain in this short space.

Now that we have established the dimensions by which we can categorize what
methods can be applied based on data availability and stability, we can now apply a
third dimension based on “business strategy.” See Figure 2 for the details.

For example, if the corporate business strategy changes from a factory “push” strategy,
where sales volumes are pushed through the channels of distribution by discounting
products, to a demand “pull” strategy, where volumes are pulled through the channels
of distribution by growing consumer demand, you are forced to use more sophisticated
techniques. In this situation, the company needs to model consumer behavior, which
requires additional causal data, as well as longer historical sales volume.

By reviewing the product portfolios, we found, on average, 10 percent of the products

fell in the lower left-hand quadrant, 50 percent fell in the upper left-hand quadrant, 35
percent fell in the upper right-hand quadrant, and five percent fell in the lower right-
hand quadrant. Furthermore, of those 35 percent that fell into the upper right-hand
quadrant, 85 percent represented corporate unit and dollar volumes. See Figure 3 for
details. This means 50 percent of your product portfolio can be forecasted using
advanced time series techniques that can be easily automated with minimal
resources. As such, resources can be utilized to provide more analysis and develop
causal models for 35 percent of the products falling in the upper right-hand quadrant.
All of these require more resources. In fact, 85 percent of the products in most product
portfolios can be forecasted using advanced time series and causal methods.
Unfortunately, the forecasting needs of most companies fall in the lower left-hand
quadrant. See Figure 4 for details.

Conclusion

Whatever the method chosen, be it judgmental, time series, or causal, all presume that
the past can be drawn upon to predict the future. Consequently, each class of methods
uses the past differently and possesses a different set of strengths and weaknesses.
Your products, goals, and constraints should be considered when selecting the
forecasting method(s). There is always a tendency to use the one- methodology-fits-all
philosophy, because it's very appealing from an implementation standpoint. However,
we need to realize that forecasting methods are really generic tools that can be applied
simultaneously across groups of products based on the corporate product portfolio.
This “toolbox” approach for selecting forecast methods is much more effective, not to
mention more accurate. It provides us with the framework to focus our resources on
more sophisticated techniques that capitalize on market opportunities, resulting in 4
increased customer value.