Forecast accuracy (MAPE)

MAPE, or mean absolute percentage error, is the average size of your forecast misses as a percentage of what actually happened. Lower is better. A MAPE of 10% means your forecasts were off by 10% on average, up or down.

Formula

For each date:

Absolute % error = |actual − forecast| ÷ actual

Then:

MAPE = sum of absolute % errors ÷ number of dates

The bars mean "ignore the sign". An over-forecast of 5 and an under-forecast of 5 count the same.

Worked example

Hypothetical. Forecast and actual cars on hire for four days:

MAPE = (6.3 + 20.0 + 5.0 + 25.0) ÷ 4 = 14.1%.

Friday and Sunday were both over-forecast. That's a bias worth fixing, and MAPE alone hides it. Also track the plain average of (forecast − actual) to see which way you lean.

Why it matters for a rental operator

A forecast you don't score is a guess. MAPE gives you one number to watch month to month and a way to compare methods: last-year pick-up versus a three-year average, for example.

Watch out for small numbers. On a day where actual is 2 cars, a miss of 2 is 100% error. Measure at category or week level if your fleet is small.

See it in Rate Pulse

Rate Pulse doesn't forecast or score demand. Setting up forecasts and accuracy tracking on your own bookings is part of our revenue management work.