Volume Prediction as a Part of Demand Planning Solution

This master thesis develops and refines predictive methodologies to improve demand forecasting accuracy in demand planning. By enriching company sales data with exogenous variables such as price indicators and macroeconomic factors, a comprehensive predictive framework was created. The study identified SARIMA, XGBoost, and LSTM as the most accurate models for given data characteristics, achieving a significant improvement in demand forecasts compared to previously used methods. Utilizing Azure’s cloud-based architecture and IBM Planning Analytics, the solution facilitated precise product-level predictions. This improved demand planning solution provides significant benefits for demand planners, including time savings, improved inventory management, and better decision-making processes.