Integrating Demand Forecasting and Inventory Policy for Home Textile Products under Limited Historical Data: Evidence from a Bed Linen Manufacturer

Abstract

Demand uncertainty remains a critical challenge for textile manufacturers, particularly for bed linen products that vary in size, design, consumer preference, promotional cycles, and seasonal demand patterns. In practice, inaccurate demand planning may lead to excessive finished goods inventory or, conversely, stockouts that reduce sales opportunities and customer service performance. This study aims to develop an integrated model that links demand forecasting with inventory policy for bed linen products at PT XYZ. The study uses weekly sales and inventory data for seven bed linen size variants from August 2024 to June 2025. The research process includes exploratory demand analysis, demand pattern classification, time series forecasting model development, forecast accuracy evaluation, and the integration of forecasting outputs into inventory control decisions through safety stock and reorder point calculations. Several classical time series forecasting methods, including Moving Average, Exponential Smoothing, Holt Winters, and ARIMA/SARIMA, are compared using MAE, MAPE, and RMSE. Due to the limited historical dataset, rolling origin cross validation is applied to obtain a more reliable model evaluation. The proposed framework is expected to identify the most suitable forecasting method for each demand pattern and provide a more structured basis for inventory decision making. By integrating forecasting results into inventory policy, this study is expected to help reduce overstock and stockout risks, improve service level, and support more adaptive production planning in the textile manufacturing context.