Business Forecasting

CHAPTER 4Forecasting Performance

The evaluation of forecasting performance is a perennial topic for discussion. Dozens of metrics have been developed, with some (such as Mean Absolute Percent Error) gaining wide adoption despite being unsuited for some common situations. While no single metric has gained universal acceptance, we covered many of the options in our 2015 collection, Business Forecasting: Practical Problems and Solutions. This brief chapter provides four new perspectives on issues relating to forecasting performance.

In the first article, Steve Morlidge attempts to move beyond the arcane debate between experts, where little attention is paid to how an error metric can be used to improve the process of forecasting. He describes a control system that provides timely, actionable feedback, intended to promote the right behavior to effect forecast improvement.

Next, Patrick Bower argues from a similar starting point, rejecting the “exercises in intellectual preening that offer little in terms of practical guidance.” Bower then provides 10 practical guidelines for selecting and implementing a forecast performance measure.

In the third article, Stefan de Kok introduces us to an entirely new metric, Total Percentage Error (TPE). TPE measures the full range of uncertainty in our forecasts, providing a tool for better decision making. He further extends the concept to provide an enhanced version of Forecase Value Added (FVA), which he dubs Stochastic Value Add (SVA).

Finally, ...