Beat the method you already use
A forecast earns a place in a decision only after its error is compared with the baseline method, on the horizon that decision actually uses.
If it does not beat that method, it stays off the board. A chart that looks sophisticated and loses to the spreadsheet is not an improvement.
What we instrument
Error against the baseline, and whether the score is visible where the decision is made. A model that nobody opens is not delivered.
Lead time — how much earlier a decision can move — is recorded after the fact, not advertised before.