Evaluating the quality of a forecasting method can be dramatically complicated. In order to help the community finding out who’s delivering the best forecasts out there :-), we have just produced a video concerning accuracy measurements of sales forecasts. Not all metrics are born equal.
Special thanks to Ray Grover for the voice-over.
Reader Comments (2)
Hum, actually, I believe it’s the opposite, MAPE puts a lot of weight on low selling products :-) Then, the idea of aligning with the cost is sound: indeed, in order to very precisely benchmark to forecasting tools, one need to introduce a ‘cost function’ that truly reflect the cost/benefit of the forecast inaccuracies. In practice though the process is somewhat complicated, so most companies stick to simpler metrics.
7 years ago | Joannes Vermorel
MAPE certainly does have its limitations and weights the A Class items very heavily. Often it should be used in conjunction with (say) a Hit/Miss measure. eg For A Class Items a HIT is within 20%, B Class 50% and C Class 100%. MAPE is still however highly relevant. It can aligns to where the dollars are. Often the Sparse items don’t really rely on forecast in a Retail environment. When I get down to 2 units send in another case. Steve
7 years ago | Stephen Kitchen