Demand forecasting in the supply chain: case study of a Bangladeshi retailer

This study aims to forecast daily sales using time series analysis through two models: Holt-Winters and ARIMA. A real dataset containing 1,826 consecutive days of sales data was used as the basis for analysis. The first model, Holt-Winters, was optimized by adjusting parameters for level, trend, and seasonality to achieve the lowest possible MAPE. Subsequently, ARIMA (4,1,1) models were developed using Minitab software. Forecasted values from each model were compared against the actual sales values to compute the Mean Absolute Percentage Error (MAPE), which was used as the main metric for evaluating forecasting accuracy. The results showed that the Holt-Winters model achieved the lowest MAPE of 22.04, indicating the highest accuracy among the two models. These findings suggest that the Holt-Winters method is a suitable approach for short-term sales forecasting in daily time series data. This study provides useful insights for improving forecasting in business operations, especially for production planning and inventory management.