MAPE: formula, example and what counts as good forecast accuracy

What is MAPE?

MAPE, the mean absolute percentage error, measures how far forecasts were from actual results on average, as a percentage of the actual values. A MAPE of 12% means forecasts missed by 12% on average.

It is probably the most widely used forecast accuracy metric, because a percentage is easy to explain to anyone.

MAPE formula

Over n periods, here is how to calculate MAPE:

MAPE = (100% ÷ n) × Σ |actual − forecast| ÷ |actual|

Take each period’s error as a share of the actual value, ignore the sign, and average the results.

A MAPE example

Actual demand over three weeks was 100, 80 and 120, and the forecasts were 90, 100 and 110. The percentage errors are 10%, 25% and 8.3%, so MAPE is (10 + 25 + 8.3) ÷ 3, about 14.4%.

MAPE vs MAE

MAE, the mean absolute error, averages errors in units instead of percentages. MAPE is easier to compare across items of different sizes; MAE behaves better when demand is low. MAPE breaks down when actual demand is zero, because the percentage error is undefined, and it treats over- and under-forecasts unevenly, which is why slow movers are often measured with MAE or a weighted MAPE instead.

What is a good MAPE for demand forecasting?

A frequently quoted rule of thumb calls under 10% highly accurate, 10 to 20% good, 20 to 50% reasonable and over 50% inaccurate. What is achievable depends on the item: steady sellers can be forecast tightly, new or erratic items cannot. Compare methods on the same items over the same held-back period rather than against a universal target.

Last reviewed 2026-11-01.

Related terms

  • Holdout validation — Holdout validation tests a forecasting method on data it has not seen: you hide the most recent period, fit the model on everything before it, and score the forecast against what actually happened in the hidden period.
  • Holt double exponential smoothing — Double exponential smoothing, known as Holt’s method, is a forecasting technique that tracks two things over time, the level of demand and its trend, and updates each with its own smoothing weight as new data arrives.
  • Service level — In inventory, service level is the probability of not running out of an item during one replenishment cycle.

All glossary terms →