Mean absolute percentage error (MAPE)
Mean over periods of |actual − forecast| ÷ actual; periods with zero actual are excluded, so for rare sizes and accessories with zeros the measure misleads: use WAPE
- Formula
Mean over periods of |actual − forecast| ÷ actual; periods with zero actual are excluded, so for rare sizes and accessories with zeros the measure misleads: use WAPE- Unit
- %
- Aggregation rule
- Average over dates
- Source system
- Plan and budget model; POS: receipts
- Owning role
- Analyst (BI)
- Frequency
- Monthly
Passport
- Grain
- SKU × period
- At zero or no data
- Periods with zero actuals are dropped; if all are zero, “—”. For rare sizes use WAPE.
- Limits
- A simple average over periods: small actuals give huge errors. The measure misleads on accessories and rare sizes.
- Check example
- Actual, period 1, units: 120; Forecast, period 1, units: 100; Actual, period 2, units: 80; Forecast, period 2, units: 90; Actual, period 3, units: 60; Forecast, period 3, units: 70.
(Math.abs(a-b)/a+Math.abs(c-d)/c+Math.abs(e-f)/e)/3*100. Result: 15.3% - Related metrics
Version 1.0 · valid from 2026-11-16 · checked 2026-11-16
Related pages
- Term: Mean absolute percentage error (MAPE)
- Calculation: Sales forecast: method, horizon, error
Definition
The mean over periods of the relative error: the gap between forecast and actual as a share of that period’s actual.
A period with zero actual is excluded and small periods inflate the measure: for rare sizes, accessories and low-sales weeks use WAPE. There is no single standard.