Q] Which formula should we use to measure forecast
error: (Actual – Forecast) / (Forecast) or (Actual –
Forecast) / (Actual)? Should we use different formulas
for computing error at a product family, SKU, and DC?
[A] Both measures are used in business, though the second
one is used the most and for the right reason. We want to know
how our forecast deviated from the actual, and not how actual
deviated from the forecast. For that reason, in computing
percent forecast error, the error should be divided by the actual,
and not by the forecast. The same error metric should be used
for measuring the performance of SKUs, product family, and
products at the DC level. Whichever error metric you use, make
sure you are consistent. Otherwise, you cannot compare the
performance. Incidentally, some companies prefer to measure
forecast accuracy instead of forecast error.