Forecast accuracy calculator

Paste your actuals and your forecast — one value per line, or comma-separated — and this works out how accurate the forecast was, and just as importantly whether it leans high or low. It gives you the four numbers worth quoting: MAPE, WAPE, bias and the tracking signal.

Actuals

Forecast

—

MAPE
(mean abs % error)

—

WAPE
(volume-weighted)

—

Bias
(signed, vol-weighted)

—

MAD
(mean abs deviation)

—

Tracking signal

■ Actual ■ Forecast

Which number to trust

MAPE	Mean absolute percentage error — averages the % miss of each period equally. Easy to explain, but it over-weights small-volume periods (a 50-contact miss on a quiet Sunday looks as bad as a 500-miss on Monday) and blows up if any actual is zero. Good for a headline, poor for skewed data.
WAPE	Weighted absolute percentage error — total absolute error as a share of total volume. It weights by size, so the busy days count for more, and it copes with zeros. For contact-centre volumes this is usually the fairer headline number.
Bias	The signed error as a share of volume. Near zero is the goal. Persistently positive means you over-forecast (over-staff, waste); persistently negative means you under-forecast (under-staff, missed service). Bias is the error that compounds — chase it before you chase the last point of MAPE.
Tracking signal	Running sum of errors ÷ MAD. It flags drift: if it climbs past roughly ±4, your forecast is systematically off in one direction and the model needs attention, not just noise.

See Why chasing forecast accuracy is the wrong goal for what to do with these numbers, and Which forecasting method should I use? to pick the model.