WMAPE, MAPE, MAE & BIAS
Forecast Accuracy Lab
Measure forecast error size and direction, then find the SKUs where error is concentrated.
Processed locally in your browser. Your planning data is not uploaded.
01
Data
Edit the table or load the example.
| SKU | Forecast | Actual | Actions |
|---|---|---|---|
02
Result
Error size, direction, and concentration.
→
Your result will appear here.
Analyze manual rows or upload a CSV to see metrics and Pareto.
Methodology, assumptions and limitations
WMAPE = Σ|Forecast − Actual| ÷ Σ ActualMAE stays in operational units. MAPE averages row percentages and excludes Actual = 0. Bias uses Σ(Forecast − Actual) ÷ Σ Actual.
Bias convention
Positive means overforecasting; negative means underforecasting.
Limitations
- WMAPE and bias are unavailable when total Actual is zero.
- MAPE can be distorted by small Actual values.
- SKU aggregation sums absolute error and avoids cancellation.