Abstract

Every business decision depends on making a forecast of the consequences of the decision. Although most organizations do forecasting, most do so badly. They ask either for a point prediction -- a single "best guess" forecast, when everyone knows that this is an oversimplification of the truth, or for a simple range forecast, which is likely to result in biased predictions more often than not. In this article, the authors propose a better approach, one that takes seriously the uncertainty in forecasting and the most common errors in the way people think about this uncertainty.