Summary
Wanting to know which forecast error measure to use is a common question. There is no one-size-fits-all answer, but StockIQ gives you a small set of measures that each answer a different question, and they work best together.
For how each one is calculated, see the Forecast Error KPIs topic.
The measures StockIQ reports
- wMAPE- Weighted mean absolute percent error: the total units the forecast missed by, divided by the total units actually sold, over the Error Sample Length. It is a percent, so use it to compare items, categories and sites with very different volumes. Earlier versions of StockIQ labeled it MAE%.
- MAE- Mean absolute error, in units: how many units you can expect to miss by in a typical period. Use it to sanity-check a wMAPE on a low-volume item, where a miss of a few units is a large percent.
- Error $- The units missed valued at the item's average unit cost. Use it to decide which errors matter most to the business and where to spend planning time first.
- Bias %- Whether the forecast runs high or low over time. A good wMAPE with a large bias still means you are consistently over- or under-stocking.
- FVA- Forecast value added: whether the statistical forecast beats the simple benchmark, and whether the operational (planner-adjusted) forecast beats the statistical one. Use it to find manual overrides that are making the forecast worse.
Which one should I use?
- To track forecast quality across the business, standardize on wMAPE at a consistent lag (set by Forecast Error Lag in Global Demand Forecast Settings), and watch Bias % beside it.
- To prioritize work, sort by Error $.
- To judge whether your forecasting process is adding value, compare against the benchmark and read FVA. See What is a good forecast error to shoot for?
Measures StockIQ does not use
- MAPE- Mean absolute percent error averages each period's own percent error. A period with very low demand produces a huge percentage that swamps the result, a period with zero demand can't be scored at all, and tuning to MAPE pushes forecasts low. wMAPE avoids all three, which is why StockIQ uses it. MAPE is not the same number as wMAPE, even though the names are close.
- RMSE%- Root mean squared error percent penalizes large misses more heavily. StockIQ no longer offers it as an error calculation; every forecast error percent in StockIQ is wMAPE.