Forecasting Accuracy Analysis of Catering Raw Material Stock Using Simple Exponential Smoothing Based on Mean Absolute Percentage Error (MAPE)

Abstract

In the catering industry, inaccurate inventory management often leads to significant food waste or stockouts due to highly volatile raw material demand, and conventional intuition-based procurement methods are no longer sufficient to maintain operational efficiency. This research applies to the Simple Exponential Smoothing (SES) algorithm to forecast raw material requirements and evaluates its accuracy using the Mean Absolute Percentage Error (MAPE) metric. Twelve months of historical transaction data from a local catering business were analyzed, categorized into basic commodities, proteins, and vegetables, with the SES model calibrated by testing smoothing constants ( ) across the range of 0.1 to 0.9. The findings indicate that stable items such as rice achieve the highest accuracy at a low of 0.2, yielding a MAPE of 4.25% — classified as Very Good. Highly volatile items such as proteins and fresh vegetables require a high of 0.8–0.9 to remain responsive, producing MAPE values between 12.40% and 18.15%, classified as Good. These results confirm that SES offers a defensible, data-grounded decision-making structure that measurably reduces forecasting errors and improves procurement cost management in the catering sector.