Mean Absolute Percentage Error (MAPE)

MAPE shows, on average, how far forecasts are from actual results in percentage terms, making it easy to understand and to compare across series of different sizes.

For example, suppose actual sales in three months were 100, 200 and 400, and the forecasts were 110, 180 and 420. The absolute percentage errors are 10%, 10% and 5%. The MAPE is the average, about 8.3%.

MAPE is popular in business forecasting because percentages are intuitive. A MAPE of 5% suggests forecasts are typically within about 5% of actual results.

It has limitations. It cannot be calculated when actual values are zero, and it becomes very large and unstable when actual values are close to zero. It also treats over-forecasts and under-forecasts asymmetrically in some situations. Alternatives, such as weighted versions of MAPE or scaled error measures, are sometimes used instead.