What is forecast value added?
Forecast value added compares the error of the forecast before and after each step in the process. A step adds value when it reduces the error, and it removes value when it increases the error.
The usual steps are a naive forecast, a statistical forecast, the changes of a planner, and the consensus of the team. FVA shows which of these steps deserve the time that people spend on them.
The formula
FVA of a step = error before the step − error after the step
- Use the same error measure for each step, for example WAPE. A positive FVA means that the step helped.
A worked example
One quarter of Cold Craft forecasts
- The naive forecast, which repeats last year, has a WAPE of 30%.
- The statistical forecast has a WAPE of 22%. Its FVA is 30 − 22 = +8 points.
- After the manual changes of the planner, the WAPE is 25%. The FVA of those changes is 22 − 25 = −3 points.
The statistical model helped, and the manual changes made the forecast worse. Change fewer numbers by hand, and give each change a reason.
Forecast value added in Miridia
Miridia Planner measures forecast value added at each stage: naive, statistical, planner-adjusted, and consensus. Each forecast run is archived, so the accuracy is tracked honestly over time.