Once, a colleague of mine became enamored with a new forecasting technology whose “targets” he consistently beat. In time, however, his performance became markedly closer to the forecast “targets,” often missing them by only a small amount. He wondered if his business forecasting tool was still effective. Indeed it was since, over time, the forecast was accurately predicting his overall performance.
But he didn’t see it that way. He was treating the predictions as goals to beat rather than as insight into the general direction of his sales. As we’ll see, human biases plays a significant role in how we build and interpret models—more than we’d like to admit. Many would prefer only forecasts they can beat, but this is problematic.