Data & drivers
Hierarchical forecasting
Hierarchical forecasting predicts at more than one grain — SKU, category, chain — and reconciles so the pieces add up. It is how you keep a detailed plan that still matches the number finance already believed.
Bottom-up respects the noisy leaves and can look wild at the top. Top-down is stable and unfair to the SKUs. Reconciliation (MinT and friends) is the compromise: forecast all levels, then project onto a coherent set.
You need a hierarchy that matches how people buy and report. A pretty product tree that ops does not use will reconcile into a fiction.
Cold starts often live in a hierarchy: the child shrinks toward the parent until it has a spine of its own.
Same folds, five candidates
Seasonal naive
16.4%
Smoothing
14.1%
Gradient boost
11.2%
Foundation
9.7%
Fine-tuned
10.1%