Causal features versus correlated features

Advanced forecasting should be judged by horizon, segment and decision value—not one aggregate score. This note examines causal features versus correlated features with that evaluation discipline.

Practitioner lens

For causal features versus correlated features, validation must respect the information set available at each forecast origin. Features published after the decision date create leakage and overstate likely performance.

First experiment

Use rolling-origin backtests, horizon-specific losses and interval coverage; keep feature availability time-correct.

This is an original educational brief. It is a starting point for discussion, not a claim that one technique fits every supply chain.