Forecasting, in its most basic form, is the practice of making predictions about future values of a variable, outcome and/or decision. The question is whether forecasting is necessary, and why. The answer to the first is a resounding yes, mainly due to the need for businesses to plan ahead and anticipate various events. For instance, firms use forecasts of independent variables (such as sales) to help guide decisions about dependent variables (such as the required level of inventory) or to estimate the impact of events that may be out of their control (e.g. a new competitor, a change in tax rates or the impact of a hurricane). Forecasting thus facilitates the formulation of rational and consistent strategies that help firms achieve their objectives. Indeed, if the future can be predicted, it can be controlled and leveraged by the business to provide an advantage over competitors. Given the importance of forecasting to businesses, good forecasts are therefore a necessity.
What is a ‘good’ forecast?
A logical question at this point would be how do we determine whether a forecast was ‘good’? Generally, the public judges a forecast solely on its accuracy: a good forecast is assumed to be one that is equal (or very close) to what actually happened (or the realised value). But forecasting should never be confused with prophecy. The chances of forecasting the exact value of a random variable is almost zero. Rather, good forecasts:
- incorporate all available information and are as accurate as possible given this information;
- are unbiased, neither consistently higher nor consistently lower than realised values; and,
- are timely, i.e. they are developed early enough to inform adequate planning and allow sufficient time to adjust.
How do we attain “good” forecasts, science or art?
‘Good’ forecasts usually emerge from striking a balance between art and science. The scientific aspect of forecasting usually refers to the quantitative methods used; this approach often gets the most attention. In fact, in the early stages of my career, I held a common, but misguided view that good forecasts came from statistical methods and powerful computers. I theorised what I considered to be a brilliant forecasting approach. I would collect a host of data; inspect the data; build a few models; compare the forecasting performance of these models; and, choose the best performing one. However, as time progressed, one thing became abundantly clear: business forecasting should not use such a simplistic approach! Naïve quantitative methods help us to identify existing relationships and/or established patterns; however, they almost exclusively assume that the future is like the past.
On the flip side is forecasting using judgment, what many people refer to as their gut instinct. Humans possess unique knowledge and insight that provide a great deal of value in business decisions. However, like quantitative methods, relying on the judgment alone can be detrimental. Judgments forecasts are subjective, vary due to psychological factors, and can be clouded by personal or political agendas.
The ideal approach
The ideal solution would be to utilize an approach that merges the science (i.e. quantitative methods) with the art (i.e. the forecaster’s insight and knowledge of the market). This approach is known as scenario analysis and has been used by forecasters since the Second World War. It was arguably best applied in the business world by Shell Oil Company, which used this approach to consider the rise and fall of oil prices and effectively plan for its business consequences. Scenario analysis provides a consistent approach to incorporating this human insight into forecasting and also allows the researcher to produce an image of the future under various potential outcomes.
Scenario analysis starts from the premise that the future is unlikely to be like the past: human choice and other unexpected events need to be factored into the equation. The aim of scenario analysis is to map the so-called possibility space (i.e. identify potential alternative outcomes) and evaluate the impact that these outcomes could likely have on the business. Rather than simply forecasting sales using a naïve method, one considers sales under different market conditions, e.g. marketing campaigns of varying degrees of success, new entrants into the market, industrial strikes, mergers, etc. Scenario analysis therefore allows the business analyst to mix the science with the art to derive competitive advantages for the enterprise.
More often than not, forecasts are put together using either quantitative methods or judgement alone. At AE, our forecasting methods integrate the two, merging art with science. Our approach is conceptually very simple. We use historical data to generate statistical forecasts as well as a map of the possibility space taking into account human insight and judgment. It is important, however, that forecasts are internally consistent, are based on logical storylines, are plausible and are associated with certain sign posts that can be used by managers to identify when this scenario is likely to occur. We have only touched the surface of the concept of forecasting in this post, but we hope that you have gained a better understanding of the state of the art when it comes to business forecasting.