Statistical
Naive forecast accuracy: WMAPE and row percentage errors
Benchmark any demand forecast against a Naïve (Y_{t-1}) baseline using Mean Absolute Percentage Error (MAPE) and volume-weighted WMAPE.
Use a naive forecast as a baseline
A naive one-step forecast uses the previous period’s actual. A seasonal naive forecast uses the corresponding period of the prior season.
WMAPE divides summed absolute errors by summed actuals. MAPE averages row percentage errors; this lesson demonstrates WMAPE and individual APE, not an aggregate MAPE formula.
| Period (A) | Actual (B) | Naïve Forecast (C) | AbsError (D) | APE (D / B) |
|---|---|---|---|---|
| P1 | 100 | (seed) | — | — |
| P2 | 125 | 100 | 25 | 20.0% |
| P3 | 100 | 125 | 25 | 25.0% |
| P4 | 125 | 100 | 25 | 20.0% (WMAPE = 75/350 = 21.43%) |
WMAPE from total absolute error
Divide total absolute forecast error across evaluated periods (D3:D5) by total actual volume over the exact same periods (B3:B5).
=ROUND(SUM(D3:D5) / SUM(B3:B5), 4)Row Absolute Percentage Error (APE)
Guard Actual = 0 with IF(B3=0, "Undefined APE", ...) before dividing.
=IF(B3=0, "Undefined APE", ROUND(ABS(B3 - C3) / B3, 4))Inputs and interpretation
Compare forecasts over the same held-out dates, horizon and metric. APE is undefined for a zero actual, while WMAPE here requires nonnegative actuals and a positive sum. It is not equally weighted across observations. Seasonal naive forecasting uses the corresponding prior-season value; the worked table below demonstrates the one-step nonseasonal baseline.
Try the example
Use the original values or change them to test your understanding.