Worst Practices in Forecasting

Jan. 11, 2008

Sometimes the pursuit of "best in class" is a losing proposition.

The most fundamental worst practice, says Mike Gilliland, product marketing manager at SAS Institute, is to assume things work without bothering to measure them. "This can occur when purported 'best practices' are implemented without solid evidence that they are even a 'good' practice." Speaking at the recent APICS Conference in Denver, Gilliland listed some of the other most prevalent worst practices:

having unrealistic accuracy expectations
setting inappropriate performance targets
employing overly elaborate systems or processes
using inappropriate statistical models due to over-fitting to history, or "pick best" selection software
not utilizing a naive forecast
political tampering with statistical forecasts
blaming the forecast for all business woes.

Questioning the conventional wisdom, Gilliland asks, "How good do you really have to be at forecasting?" For many companies, he observes, achieving "best in class" forecasting could very well be a waste of resources if the company's major challenges lie in other areas. "By eliminating the worst practices in their forecasting efforts," he explains, "companies should be able to forecast as accurately as can reasonably be expected and achieve this as efficiently as possible."

See Also