Forecast accuracy
How far forecasts land from actual demand, and whether they lean high or low.
MAPE = average of |Actual − Forecast| ÷ Actual
WMAPE = Σ |Actual − Forecast| ÷ Σ Actual
Bias = (Σ Forecast − Σ Actual) ÷ Σ Actual
MAD = average of |Actual − Forecast|
No forecast is right, so the useful questions are how wrong and in which direction. Error measures tell you how much buffer the forecast forces you to carry; bias tells you whether the planning process itself leans optimistic or cautious.
MAPE averages the percentage error of each period, so a miss on a small period counts as much as a miss on a large one. WMAPE weights by volume and is the better choice when periods or products differ a lot in size.
Calculator
The fields start with a worked example. Replace them with your own figures; the results update as you type.
- MAPE
- 7.3 %
- WMAPE
- 7.0 %
- Bias
- +0.4 %
- Mean absolute deviation
- 9.5 units
How to read it
Check bias first. A forecast with a modest error but a steady positive bias quietly builds inventory month after month; one with a larger error and no bias is easier to buffer against.
Common mistakes
- MAPE is undefined when actual demand is zero and explodes when it is close to zero. Use WMAPE for intermittent items.
- Measure at the level where the decision is made. Accuracy by product family looks far better than by item and location.
- Compare against the forecast that was frozen at the lead time, not the one revised a week before the month ended.
Last reviewed 2 November 2026.