Free tool
Forecast accuracy calculator
Enter the actual and forecast for each period to get MAPE, WAPE, accuracy and bias, with zero-actual periods called out rather than hidden.
Your periods
Enter the actual and forecast for each period.
| Period | Actual | Forecast | Error | Remove |
|---|---|---|---|---|
| 10 | ||||
| 20 | ||||
| 10 | ||||
| 10 | ||||
| 10 | ||||
| 10 |
Estimated forecast accuracy (1 − WAPE)
89.23%
- MAPE
- 11.01%
- WAPE
- 10.77%
- Bias
- +1.54%
- MAE
- 11.67 units
Bias uses forecast minus actual, so a positive bias means the forecast ran ahead of actual demand.
The formula
MAPE = mean(|actual − forecast| ÷ actual) × 100, over periods where actual is not 0
WAPE = SUM(|actual − forecast|) ÷ SUM(actual)
Accuracy = 1 − WAPE
Bias = SUM(forecast − actual) ÷ SUM(actual)
MAE = SUM(|actual − forecast|) ÷ number of periods
- Actual
- What actually happened in the period.
- Forecast
- What was forecast for the same period, on the same unit and time bucket.
- MAPE
- Mean absolute percentage error: the average of each period's own percentage miss. A period with actual = 0 has no defined percentage and is left out of the average, not treated as 0% or 100%.
- WAPE
- Weighted absolute percentage error, sometimes written WMAPE: total absolute error divided by total actual volume, rather than an average of per-period percentages. A period with actual = 0 still adds its error to the numerator; it only drops out of MAPE.
- Bias
- Signed error over the same total-actual denominator as WAPE. This calculator defines bias as (forecast − actual) ÷ actual, so a positive bias means the forecast ran ahead of actual demand and a negative bias means it ran behind. Some sources define it the other way around; check which convention a number you are comparing against uses before treating the sign as agreement or disagreement.
- MAE
- Mean absolute error in the original units, unscaled by actual volume.
Worked example
- Six periods of actual and forecast demand give six errors: 10, 20, 10, 10, 10, 10 units, for a total of 70 against total actual demand of 650.
- WAPE weights every row by its own actual volume: 70 ÷ 650 = 10.77%, so accuracy (1 − WAPE) is 89.23%.
- MAPE instead averages each row's own percentage error, 10%, 16.67%, 12.5%, 6.67%, 11.11%, 9.09%, for a mean of 11.01%, a different number because it treats a small period's miss as equal to a large period's.
- Bias uses the same total-actual denominator but keeps the sign: (+10 -20 +10 -10 +10 +10) ÷ 650 = +1.54%, positive because the forecasts ran ahead of actual demand more often than behind it.
When it applies
- Comparing forecast performance across periods or SKUs on a total-volume basis, where WAPE is the more common ASCM-aligned choice because one large period cannot be swamped by several small, noisy ones.
- Diagnosing a single item's typical percentage miss with MAPE, once periods with zero actual demand are set aside from that average.
- Checking whether a forecasting process runs systematically high or low over time using bias, separately from how large the errors are.
When it breaks down
- New or intermittent-demand items with many zero-actual periods: MAPE excludes every one of them, which can leave too few periods to average meaningfully, and WAPE's total-actual denominator can be thin for the same reason.
- As a single number without also checking bias: two forecasts can share the same MAPE or WAPE while one runs consistently high and the other consistently low, which changes the fix.
- Comparing MAPE and WAPE figures from different sources without confirming which one a report means by "forecast error," since the two answer different questions from the same inputs and are not interchangeable.
Common mistakes
Treating a zero actual as a 100% or a 0% error
Dividing by zero is undefined, not large or small. This calculator drops periods with actual = 0 from the MAPE average and reports how many were excluded, rather than assigning them an arbitrary percentage.
Averaging MAPE across SKUs of very different volume
A straight average of per-SKU MAPE values gives a slow-moving item the same weight as a top seller. WAPE, computed once across the combined actual and error totals, does not have this problem.
Reading a signed bias without checking the sign convention
Some tools define bias as actual minus forecast instead of forecast minus actual, which flips which direction is "over" and which is "under." This calculator uses forecast minus actual: positive means the forecast ran ahead of demand.
Questions
What is a good MAPE for forecasting?
There is no single target: a good MAPE depends on how many items are aggregated, how volatile the underlying demand is, and the length of the period being forecast. A stable, aggregated monthly category and a single volatile SKU forecast weekly are not comparable on the same scale, so benchmark a MAPE against that same item's own history, or against a naive forecast (next period equals this period), rather than a published number.
What is the difference between MAPE and WAPE (and bias)?
MAPE averages each period's own percentage error, so every period counts equally regardless of size, and it is undefined wherever actual is zero. This calculator drops those periods from the average and reports how many were excluded rather than treating them as 0% or 100% error. WAPE (also written WMAPE) instead divides total absolute error by total actual volume, so large periods weigh more and a zero-actual period still adds to the error total without breaking the calculation. Bias is a third, separate figure: a signed version of the same total-actual ratio that shows whether error runs systematically high or low, which a magnitude-only measure like MAPE or WAPE cannot show.
How is forecast accuracy calculated?
This calculator reports accuracy as 1 − WAPE, where WAPE is total absolute error across all periods divided by total actual volume. That makes accuracy a volume-weighted measure: it is not simply the average of each period's own percentage accuracy, and one large period's miss affects it more than several small ones.
Can forecast accuracy be negative?
Yes. When 1 − WAPE is used, a WAPE over 100% (total absolute error larger than total actual volume) makes accuracy negative, and this calculator never floors or clamps the result, so a badly miss-forecast set of periods will show a negative number rather than 0%.
Is forecast accuracy 1 − MAPE?
Not on this page. 1 − MAPE is a common shorthand, but this calculator defines accuracy as 1 − WAPE because WAPE is volume-weighted and does not break on a zero-actual period the way a per-period MAPE average does. The two will usually be close but are not the same number, so check which definition a reported accuracy figure is using before comparing it to this one.
Related
Published by Skuvelo. Results are estimates computed from the figures you enter, not a reading of your own sales or stock.
A calculator answers once. Skuvelo keeps answering.
This tool computes one number from what you type. Skuvelo computes it continuously, for every SKU, from your own sales and stock.