The accuracy of business forecasts typically depends on choosing the statistical model best suited to the nature of the data cranked into the system. For example, historical data depicting seasonal trends might call for a different forecasting model than data reflecting business interruptions such as strikes. The Forecast Pro line of software developed by Business Forecast Systems Inc. (BFS) incorporates an improved "expert selection" approach that tests the data properties and then applies a rule-based system to select the most appropriate forecasting model. As a result, even people without backgrounds in statistics or economics can use the system to develop more accurate forecasts -- even large-scale forecasts involving thousands of items. The accuracy of the BFS approach was confirmed in a competition sponsored this year by the International Journal of Forecasting in which its expert-selection technique outperformed all other commercial entries and most of the academic models. "If you improve the accuracy of your forecast, you will improve the accuracy of your planning -- which can result in substantial savings," says Eric Stellwagen, vice president at BFS. The software can be used by marketing departments to forecast sales by product line -- or by production planners to predict inventory requirements at the SKU (stock-keeping unit) level. In addition to incorporating the expert-selection feature into its Forecast Pro standalone software products, BFS recently developed a "dynamic link library" that allows other vendors to seamlessly integrate its forecasting routines into their packages that require such functionality. Demand-planning systems from Bridgeware Inc. and Made2Manage Systems Inc. were among the first to embed the BFS technology. John Teresko, John Sheridan, Tim Stevens, Doug Bartholomew, Patricia Panchak, Tonya Vinas, Samuel Greengard, Kristin Ohlson, and Barbara Schmitz contributed to this article.