No single forecast metric answers every planning question. Accuracy measures error size, bias measures direction, and Pareto shows where attention matters.
Analyze forecast accuracyMAE: operational units
MAE is average absolute error in the original unit. It is easy to translate into labor, inventory, or capacity, but difficult to compare across scales.
WMAPE: portfolio error
WMAPE divides total absolute error by total actual. It weights high-volume items naturally and remains defined for individual zero-actual rows when total actual is positive.
| Metric | Question | Limitation |
|---|---|---|
| MAE | How many units? | Not scale-free |
| WMAPE | Weighted portfolio error? | Volume dominates |
| MAPE | Average row %? | Excludes zeros |
| Bias | Systematic direction? | Errors cancel |
MAPE: row percentages
MAPE averages absolute percentage error by row and excludes Actual = 0. Small actuals can dominate, so always show exclusions.
Bias: direction
Supply Planning Lab defines Bias = Σ(Forecast − Actual) / ΣActual. Positive means overforecasting; negative means underforecasting.
Pareto: concentration
Aggregate absolute error by SKU, sort descending, and find the few items accounting for most error. This directs planner time better than a portfolio average alone.
Use a metric set
Pair WMAPE or MAE with bias, zero-demand counts, item segmentation, and error concentration. Compare metrics at a consistent grain and horizon.
Test your own assumptions.
The calculator shows methodology, scenarios, and interpretation without signup.
Analyze forecast accuracy