What are MAPE and WAPE?
MAPE is the mean absolute percentage error. It calculates the percentage error of each product and takes the average. A product that sells a few units can then have a very large error, and it distorts the result.
WAPE is the weighted absolute percentage error. It adds up all the errors and divides by the total demand, so each product counts in proportion to its volume. For a range of products, WAPE is usually the better measure to report. Report bias too, because an error can be too high or too low.
The formula
MAPE = average of ( |actual − forecast| ÷ actual ) × 100
WAPE = Σ |actual − forecast| ÷ Σ actual × 100
Bias = Σ (forecast − actual) ÷ Σ actual × 100
- A positive bias means that the forecast was too high.
A worked example
Three products in one week
- Product 1: actual 100, forecast 90. The error is 10, or 10%.
- Product 2: actual 10, forecast 20. The error is 10, or 100%.
- Product 3: actual 200, forecast 210. The error is 10, or 5%.
- MAPE = (10% + 100% + 5%) ÷ 3 = 38.3%. WAPE = 30 ÷ 310 = 9.7%. Bias = 10 ÷ 310 = +3.2%.
MAPE says 38% because one small product missed. WAPE says 9.7%, which describes the forecast of the whole range better.
MAPE and WAPE in Miridia
Miridia Planner measures the accuracy of each SKU in a backtest against its own history, and it weights the accuracy by value when it rolls up to a product family or a region.