Stockertim Dictionary · Strategy and management · Technology
Mean absolute deviation
Also called MAD, mean absolute error, MAE
In short
Mean absolute deviation is a measure of forecast accuracy found by averaging the sizes of forecast errors over several periods while ignoring whether each was positive or negative.
- Intermediate
- Ratio or metric
- Checked
In detail
Mean absolute deviation, often shortened to MAD, answers the question of how far off a forecast typically is. Each period's error is taken without its sign, so an error of minus 20 counts the same as plus 20, and the average is calculated. Because positive and negative errors no longer cancel out, the figure shows the real size of misses even when the forecast is unbiased on average.
MAD is easy to compute and explain, and it is expressed in the same units as demand, such as units or orders, which makes it intuitive for planners. It is commonly used to compare forecasting methods on the same data, to choose smoothing constants and to size safety stock, since buffers need to cover typical errors over the lead time. Dividing the running sum of errors by MAD gives a tracking signal, which flags persistent bias.
Other accuracy measures suit other purposes. Mean squared error punishes large misses more heavily, which matters where big errors are costly, and mean absolute percentage error expresses errors as a share of demand, which allows comparison between products with very different volumes.
Formula
MAD = sum of |actual − forecast| / number of periods
- actual
- demand in each period
- forecast
- forecast for each period
- number of periods
- how many periods are averaged
Worked example
A forecast for a product at Tidewater missed by 10, −5, 20 and 5 units over four weeks.
- Sum of absolute errors: 10 + 5 + 20 + 5 = 40
- MAD: 40 / 4 = 10
The forecast is typically about 10 units away from actual demand each week.
In practice
Planning systems report MAD for each item and use it to set safety stock and to compare alternative forecasting methods before choosing one.
Watch out
- MAD in units cannot compare a high-volume item with a low-volume one; use a percentage measure.
- Low MAD does not rule out bias; check the sum of errors too.
Learn it properly
Official sources
- 6.4.3.2 Forecasting with single exponential smoothing (NIST/SEMATECH e-Handbook) · National Institute of Standards and Technology
- 6.4 Introduction to time series analysis (NIST/SEMATECH e-Handbook of Statistical Methods) · National Institute of Standards and Technology
Facts last checked on 3 November 2026.
Education, not advice: the Stockertim Dictionary explains what terms mean. It never recommends buying or selling anything.