In spectral analysis of time series data in statistics, mean absolute percentage error (MAPE) is measure of accuracy in a fitted time series value in statistics, specifically trending. It usually expresses accuracy as a percentage. The MAPE is a useful measure for measuring the accuracy of a single unit in a time series.
However, in practical application of time series analysis, such as in demand or economic forecasting, it is often necessary to aggregate a sum of separate but related time series. One such example is in forecasting the sales outlook for a variety of products. Though MAPE will tell the accuracy of each product, it has a major drawback in that a simple average might distort the analysis, effectively "washing out" high-accuracy series with low-accuracy series. To compensate for this shortcoming, forecast practitioners will calculate some weighted value of the individuals MAPEs. This is known as the real aggregated percentage error (RAPE)[1] or weighted absolute percentage error (WAPE); both are common and accepted forecast-industry terms. They vary slightly in that WAPE will give a higher weight to series with higher observations (volumes, units of sale, etc.), where RAPE will use some other criterion for weighting the individual accuracy, such as margin, impact (higher regressor values), or importance to the forecaster or analyst.
Note: There is in fact no mention of "RAPE" on the cited page (or anywhere in the book referenced)!