Forecast error (WAPE)

Of all the units we sold, how many did the forecast get wrong?

Current value (inventory case data): 77.0% - ✕ Act. Open the case.

Field	Value
Definition	Sum of absolute errors divided by sum of actuals.
Formula	sum(/A - F/) / sum(A)
Unit	%
Granularity	SKU x week; rolled up by SKU class.
Direction	Lower is better ↓
Target and thresholds	≤ 30.0% (amber ≤ 45.0%)
Parameters	–
Owner	Demand Planner
Source	Demand (UCI Online Retail II) + baseline forecast computed in this repo
Input	table forecast, columns actual_units, forecast_units
Recommended chart	line - Weekly trend; seasonal peaks annotated.
Related KPIs	Forecast error (MAPE), Forecast bias

How to read it

Volume-weighted: big SKUs dominate, which is what inventory and service feel. The headline planning-accuracy KPI.

Common misreadings

• Computed at a high aggregation level, errors cancel out (SKU A over, SKU B under); compute at SKU level, then sum.

Worked example (from the test suite)

actual_units	forecast_units
100	90
10	20
0	5

Expected result: 22.73%. Sum |A-F| = 10 + 10 + 5 = 25. Sum A = 110. 25 / 110 = 0.22727. Same data as MAPE: 22.7% vs 55%.

This example is asserted in tests/test_formulas.py against the Python function and in tests/test_sql.py against the SQL below.

Implementations

Python function (the single implementation used by the site):

@kpi("wape")
def wape(df: pd.DataFrame) -> float:
    return safe_ratio(
        (df["actual_units"] - df["forecast_units"]).abs().sum(), df["actual_units"].sum()
    )

Reference SQL (tested on SQLite):

SELECT SUM(ABS(actual_units - forecast_units)) * 1.0 / SUM(actual_units) FROM forecast