FORECASTING DEMAND AT THE OUTLET–PRODUCT LEVEL IN DISTRIBUTION USING MACHINE LEARNING ALGORITHMS: A CONCEPTUAL MODEL
Keywords:
distribution, demand forecasting, machine learning, XGBoost, Random Forest, FMCG, SKU, conceptual modelAbstract
In the distribution of fast-moving consumer goods (FMCG), the basis for building a
correct plan for sales agents and van-sellers is an accurate forecast of future demand
for each outlet and product. This article — the first part of a three-part series —
proposes a conceptual and methodological model for forecasting weekly demand at
the outlet–product (SKU) level using machine learning (ML) algorithms. Based on
the data schema of a real distribution information system (DMS/SFA), it describes
the feature set, the choice of models (Random Forest, XGBoost), and the evaluation
methodology (MAE, RMSE, MAPE, R²). The article provides an integrated
framework for practical implementation; empirical verification is intended to be
carried out on real data.
