Abstract:
Accurate demand forecasting is crucial for efficient production and inventory management in the food sector, where product perishability and variable demand present significant challenges. This study evaluates the forecasting accuracy and operational efficiency of various time series forecasting models, including the Moving Average, Exponential Smoothing, Holt-Winters, ARIMA, and SARIMA models. The primary aim is to investigate how forecasting accuracy influences the connection between these models and completed product inventory management. Data from PT. XYZ, covering a period of 38 months (2021-2024), was utilized to evaluate the models, with performance measured using MAPE, RMSE, and MAE. The empirical findings demonstrate that the Holt-Winters model yields minimal error levels, especially with MAPE and RMSE. Nevertheless, it often underestimates during peak demand periods, resulting in inventory deficiencies. In contrast, the SARIMA model, albeit demonstrating marginally elevated error values, was more adept at identifying seasonal patterns and delivering more precise forecasts during peak demand periods, rendering it the favored option for inventory management. Mediation research indicates that forecasting accuracy strongly mediates the relationship between forecasting models and the management of finished products inventory. The empirical evidence indicates that enhanced forecasting accuracy results in superior inventory management, streamlined production timelines, and diminished operational inefficiencies. This study recommended the implementation of SARIMA instead of Holt-Winters in food manufacturing environments to improve demand forecasting and inventory control, especially during periods of heightened demand.
