A multi-echelon system was designed to generate statistical forecasts of menu-item demand in hospitals from one- through twenty-eight-day intervals prior to patient meal service. The three interdependent echelons were: (1) Forecasting patient census, (2) estimating diet category census, and (3) calculating menu-item demand. Eighteen weeks of supper data were utilized to analyze diet category distribution patterns and menu-item preferences, to test forecasting models, and to evaluate the performance of the forecasting system. A cost function was used to evaluate the efficiency of the mathematical forecasting system and manual technique over a nine-week period. The cost of menu-item forecast errors resulting from the use of adaptive exponential smoothing and Box-Jenkins formulations was approximately 40 per cent less than costs associated with the manual system.