Title: Improving Demand Forecasting Accuracy for Pharmaceutical Injection Products Using Time-Series Methods: Evidence from an Indonesian Pharmaceutical Manufacturer
Authors: Naniek Utami Handayani and Anggita Dwi Larasati
Volume: 10
Issue: 5
Pages: 97-109
Publication Date: 2026/05/28
Abstract:
The pharmaceutical industry requires an accurate demand forecasting system to support production planning, inventory control, and service level improvement, especially for injection products that have critical and fluctuating demand characteristics. This study aims to evaluate the most suitable time-series forecasting method for predicting demand for Injection Product X at an Indonesian pharmaceutical company and to provide a basis for improving the demand-planning process. The research uses a quantitative case study approach, drawing on historical monthly demand data from 2020 to 2024. Three time series methods are compared: Double Moving Average, Double Exponential Smoothing, and the Winter Method (Holt-Winters). Accuracy evaluation is conducted using Mean Absolute Percentage Error, Mean Absolute Deviation, and Mean Squared Error. The best method is then validated using moving range and used to forecast demand for the next 12 periods. The research results indicate that the conventional forecasting method based on simple averaging yields a high error rate, with a Mean Absolute Percentage Error of 56.211%, and is therefore categorized as inaccurate. After comparing methods, Double Exponential Smoothing yielded a Mean Absolute Percentage Error of 25.825%, Double Moving Average was 23.477%, and the Winter Method was 20%. The Winter Method became the best method because it captured the trend and seasonal components in the demand data for Injection Product X. Validation using the moving range showed that all error values were within control limits, indicating that the model is stable and suitable for use. The forecast results for the next 12 periods provide a quantitative basis for the company to prepare production plans, procure raw materials, control inventory, and anticipate the risks of overstocking and stockouts. This study confirms that selecting a forecasting method appropriate to the characteristics of the data pattern can improve forecast accuracy and strengthen the demand planning process in the pharmaceutical industry. The main contribution of this study lies in applying comparative time-series methods at the level of individual pharmaceutical products and in strengthening data-based forecasting SOP recommendations.