Forecast accuracy is usually reported as MAPE, the mean absolute percentage error between forecast and actual demand. Accuracy alone is not enough: bias — a forecast consistently high or low — does more damage than random error, because it compounds into stockouts or excess stock.
Your actuals vs forecast
One period per line, as actual, forecast (comma, tab or space separated). Add as many periods as you like.
Frequently asked questions
What is a good MAPE for demand forecasting?
MAPE (mean absolute percentage error) is how far off, on average, your forecast was from actual demand, as a percentage. There is no universal target — it depends on the product and how volatile demand is — but the real test is whether your forecast beats a naive "next period equals this period" guess. A sophisticated forecast that is no better than naive is adding no value.
What is the difference between accuracy and bias?
Accuracy metrics like MAPE and MAD measure the size of the error, ignoring direction. Bias measures direction — whether the forecast is consistently too high or too low period after period. A forecast that is sometimes 10% high and sometimes 10% low is inaccurate but unbiased; one that is consistently high is biased, and bias often has a human or incentive cause.
What is the tracking signal?
The tracking signal is the running sum of forecast errors divided by the mean absolute deviation (MAD). It behaves like a control chart: when it drifts beyond about ±4 the forecast is out of control and systematically biased, and should be recalibrated rather than left to run.