Logistics unit converter
Switch between the units your supply chain speaks.
Data & AnalysisDATA & ANALYSIS
Forecast Accuracy, WAPE, MAPE & Bias Calculator
Evaluate paired actual and forecast data with WAPE, MAPE, signed bias, MAE, and RMSE, plus a period-by-period CSV export.
Forecast accuracy & bias calculator
Evaluate paired actual and forecast data with WAPE, MAPE, signed bias, MAE, and RMSE, plus a period-by-period CSV export.
YOUR RESULT LIVE CALCULATION
Weighted absolute percentage error
10.34% WAPE
POSITIVE BIAS = OVER-FORECASTActualForecast
- MAPE (nonzero actuals)
- 9.94%
- Forecast bias / actual total
- 3.45%
- Mean absolute error
- 10
- Root mean squared error
- 11.18
- Zero-actual periods
- 1
- Total actual / forecast
- 580 / 600
MAPE excludes zero-actual rows; WAPE and percentage bias are N/A when total actual is zero. Errors can exceed 100%. Rows remain in CSV order; this evaluates forecasts, it does not generate them.
Period errors
6 rows
| Period | Actual | Forecast | Forecast - actual | Absolute error | APE % |
|---|---|---|---|---|---|
| Jan | 100 | 110 | 10 | 10 | 10 |
| Feb | 120 | 115 | -5 | 5 | 4.17 |
| Mar | 80 | 100 | 20 | 20 | 25 |
| Apr | 0 | 10 | 10 | 10 | N/A |
| May | 150 | 140 | -10 | 10 | 6.67 |
| Jun | 130 | 125 | -5 | 5 | 3.85 |
BEHIND THE NUMBER
How the calculation works
Rows must be aligned observations for one consistent item or cohort, horizon, and unit of measure. Positive bias means over-forecasting. MAPE averages absolute percentage errors only where actual is greater than zero; WAPE includes the absolute error of zero-actual rows but requires a positive total actual. MAE and RMSE use the original demand units. Percentage errors can exceed 100%; the tool does not disguise them as a bounded accuracy score.
WAPE = sum(|forecast - actual|) / sum(actual) × 100; bias = sum(forecast - actual) / sum(actual) × 100
When this calculation helps
Use paired actual and forecast values for the same item scope and time periods. Examine bias alongside an error measure so persistent underforecasting is not hidden by positive and negative differences canceling.
Check the interpretation
A zero-actual period makes an individual percentage error undefined. Do not replace that undefined percentage with zero or silently remove difficult periods. The aggregate and period measures answer different questions and should retain their stated denominators.
This measures entered forecasts, not future forecast quality. Zero total actual produces N/A percentage metrics. Rows remain in CSV order; use unique period labels and comparable forecast horizons. Dataset sharing is disabled.
Forecasting: Principles and Practice on accuracy measuresGood to know
How are zero actuals handled?
Zero-actual rows are excluded from MAPE, but their absolute error contributes to WAPE. WAPE and percentage bias are undefined when all actual demand is zero; MAE and RMSE are still reported.
Is forecast bias signed?
Yes. The convention here is forecast minus actual, divided by total actual. A positive percentage indicates aggregate over-forecasting; a negative percentage indicates under-forecasting.