Whatever the nature of your forecast, there are a few
principles that, if applied in the beginning of your forecasting exercise, will
give you a step ahead of the complexity. These principles, which I call the
forecasting framework, are composed of three dimensions: factorial analysis, past versus future, and cost versus accuracy.
The first step in any good forecasting practice is to split
your target variable. Reduce its complexity. Define the lower variables, which
actually influence the one you want to forecast. Another simple example: profit
(revenues — cost) and revenues (price x sales volume). Assuming price in the
short term might be fixed, to predict your revenues you just need to predict
your volume. So what are the factors that influence the variability of your
volume? New launches of competitors, customer demands?
Once you have identified all the potential influencing
factors, assess them in terms of impact and magnitude. Through the assessment
you will be able to score each of the factors in terms of their contribution to
movement in your target variable.
Depending on the nature of your key factors, you can use
different techniques to assess their impact. Each testing technique depends on
data availability and the nature of each single factor. Once again, precision
is welcome but not necessarily mandatory. As you will have the chance to test
and improve your forecasting, the success factor here is to identify the
eligible key influencing factors to focus on.
Past versus future
Once you have identified key factors, you need to analyze
them. A first step is to split them into categories. Are they existing factors
or are they new?
Do you have past history to rely on? If you’re a startup,
and want to assess the credit risk of your client, how do you do that? You
don’t have a time series; they are all new to your business.
Conversely, if you have a long-standing relationship with
your key clients, and your demand is composed of cyclical order, analyzing a
time series might provide interesting insight.
Some key factors have a long time series (i.e., pharma sales
volume of a branded product launched 20 years ago, IMS data in some markets).
Some do not (i.e., the reaction of your target customers to the launch of a new
digitalized strategy on Facebook). If you are lucky enough to have a time
series, use that. For instance, as said before, you can use regression analysis
and see the effect of your competitors’ ADV on your top line. But, if you are
missing that, here is what could be done.
If your key factors do not show any usable time series, you
may find using a proxy to be worthwhile. In this trick, you find a proxy to
your key factor. A proxy is a variable that can explain the movement of your
key factor. But unlike your key factor, it is measurable. For instance, if you
want to target the affluent and growing middle class in the emerging markets,
you may find it difficult to measure that. What is the percentage of salary
they will be spent on acquiring brand-name drugs? Use a proxy. For instance:
you have data from the French pharma market that shows the average percentage
of salary spent by the middle class on medicines. You can adapt this proxy, applying
a multiple that simulates the lag in achieving a similar percentage. Be
creative, you are forecasting; this is a learning exercise.
Cost versus accuracy
Now, you have identified your key factors. You understand
the different ways to measure them. The last question then is about forecasting
Measurability is partially influenced by the very first two
points, but it is also a separate variable to be addressed. Indeed,
measurability is all about a trade-off—if you spend more, will your accuracy be
Clearly, if you have to address a multimillion-dollar
manufacturing business, pursuing higher levels of accuracy in your forecast
will allow you to save a huge amount of money. Conversely, when exploring the
adoption of new and challenging strategies, the most important thing might be
the possibility to learn and the consequences of your decision.
This is the most important lesson in forecasting: It’s a
process, and most of the time it’s not perfect. Mistakes and inaccuracy are
part of the game. You have to refine your forecasting practice, and you have to
work to identify areas of improvement. But the value of the forecasting process
is clear in that it will teach you a lot.
Luca De Angeli,
FP&A, is finance and strategy director at Finox Biotech in Kirchberg,
Read the expanded
version of this article at www.AFPonline.org/Exchange.