Analyzing Artificial Intelligence in Demand Planning Forecasting Algorithms

SAP IBP offers approximately 20 forecasting algorithms, such as regression models and machine learning-based algorithms (e.g., gradient boosting, extreme gradient boosting). These can account for additional effects like price, promotion calendars, and events by integrating external variables.

Users can include multiple forecasting algorithms in a model, allowing the system to select the best fit. Curve-based forecasting enables grouping of products with similar sales histories, generating a normalized reference curve for improved forecast accuracy—useful for items like spare parts with aligned behavior.

Demand sensing leverages machine learning and pattern detection for daily/weekly short-term forecasts, adjusting based on correlations with orders, deliveries, snapshots, calendar events, and detected change points. This supports rapid adaptation to changing demand signals.

Organizations can also integrate their own machine learning models for tailored pre-processing or forecasting approaches.

• Advanced Forecasting Algorithms integrate 20+ methods including Gradient Boosting, XGBoost, and ARIMAX/MLR with internal and external drivers like price, promotions, and events to improve forecast quality and accuracy across demand planning.
• Demand Sensing and Pattern Recognition uses XGBoost to forecast weekly demand disaggregated by working days and holidays, with pattern detection adapting rapidly to changing demand signals, reducing stock-outs and freeing planners' capacity.
• Curve-Based and Custom ML Integration groups similar products using normalized reference curves for spare parts and seamlessly integrates custom machine learning models into SAP IBP's forecasting process to optimize accuracy and responsiveness.