Predicting Under- and Overperforming SKUs within the Distribution–Market Share Relationship

•Machine learning performs comparably to velocity models for distribution and market share.•Weighted random forests accurately predict 83% of under- and overperforming SKUs (in the velocity model).•Reasons for under- or overperformance are suggested by the algorithm (the most important SKU features for prediction).•The two most important features for prediction are store size and parent brand performance.