This study develops demand and time forecasting models to improve inventory management and order fulfillment processes for a textile company’s warehouses in Türkiye. The dataset consists of historical weekly sales and warehouse records from 2017 to 2022. Time series methods, including Holt-Winters and Seasonal ARIMA (SARIMA), as well as machine learning algorithms such as LightGBM and XGBoost, were applied and compared. Forecasting performance was evaluated using Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). Among the tested models, SARIMA achieved the best forecasting performance with the lowest error rates, outperforming both the traditional Holt-Winters method and machine learning models. The results show that accurate forecasting with SARIMA improves order planning efficiency, reduces excess inventory, and minimizes stock waste. Additionally, the forecasting models provided useful insights into supplier delivery behavior, supporting more proactive supply chain management. Due to its scalability, the proposed forecasting approach can be adapted to other industries that face similar challenges related to demand volatility and warehouse operations.