<![CDATA[Background Inaccurate production planning can result in excess inventory, material accumulation, and imbalances between incoming and outgoing stock in manufacturing companies. These issues were identified in the production of 19-liter bottled drinking water at PT ABC, highlighting the need for a forecasting method capable of providing more accurate demand estimates Purpose This study aims to determine the most appropriate demand forecasting method for 19-liter bottled drinking water by comparing several quantitative forecasting techniques based on historical demand data. Methodology A quantitative time-series approach was employed using monthly demand data for 19-liter gallons collected throughout 2022. Three forecasting methods—Single Exponential Smoothing, Linear Regression, and Double Exponential Smoothing with Trend—were evaluated. Model performance was assessed using Mean Absolute Percentage Error (MAPE), Mean Absolute Deviation (MAD), Mean Squared Error (MSE), and Tracking Signal analysis to determine forecasting accuracy and model validity Findings The historical demand data exhibited an increasing trend, indicating that Linear Regression was the most suitable forecasting method. Among the evaluated models, Linear Regression produced the lowest forecasting errors, with a MAPE of 0.41%, a MAD of 4,062.611, and an MSE of 26,179,663.49. Based on this model, the projected demand for the following six months was 1,001,833.2; 1,004,232.5; 1,006,631.9; 1,009,031.2; 1,011,430.5; and 1,013,829.9 gallons, respectively. Implications The findings provide a reliable basis for improving production planning, inventory management, and decision-making processes by reducing the risk of overstock and enhancing the efficiency of manufacturing operations. Future studies are recommended to incorporate longer historical datasets and compare conventional forecasting techniques with advanced machine learning or hybrid forecasting models to improve prediction accuracy under dynamic market conditions. Originality This study contributes by systematically comparing multiple quantitative forecasting methods to identify the most appropriate technique for the demand pattern of 19-liter bottled drinking water in a real industrial setting. The proposed approach offers a practical framework for selecting forecasting models based on demand characteristics, thereby supporting more effective production planning in the bottled water industry.]]>
