<![CDATA[This study aims to optimize inventory control for snack products in the retail industry, characterized by highly fluctuating demand and elevated Days Sales Inventory (DSI) levels. Focusing on Class A products with high DSI offers the greatest potential impact on inventory management efficiency. The approach integrates Holt-Winters forecasting, ARIMA (1,1,1), and Random Forest Regression with quantitative models such as Economic Order Quantity (EOQ) and Periodic Order Quantity (POQ), as well as stock control techniques including Safety Stock and Reorder Point, to determine the optimal order quantity and ordering time. Sales data for 21 weeks were processed to generate sales forecasts for the subsequent 31 weeks, covering weeks 22 through 52, using all three forecasting methods. The evaluation metrics indicate that Random Forest Regression achieved the best performance, with a Mean Absolute Error (MAE) of 42.4, Mean Absolute Percentage Error (MAPE) of 13.9%, and a Root Mean Squared Error (RMSE) of 46.7, The results show a significant reduction in DSI and total costs, contributing positively to strengthening the company’s cash flow. Further analysis over the 31-week period using the POQ method resulted in a decrease in DSI from the actual level of 111 days to 71 days, and also reduction in total cost from IDR 14.933.114 to IDR 10.104.863, representing a difference of IDR 4.828.250. In addition to the integrated forecasting and EOQ–POQ methods, it is recommended to enhance the adaptation of dynamic forecasting models that are more responsive to changes in demand patterns and to develop real-time monitoring systems using ERP or IoT technology to minimize the risks of stockouts and product spoilage. This research provides both practical and academic contributions toward achieving more efficient and sustainable inventory management for snack products. ]]>
Copyrights © 2026
