Many forecasters use Mean Absolute Percentage Error (MAPE) when evaluating theeffectiveness of their forecasting models. This paper discusses why MAPE may not, in fact, improve forecast model selectionbecause error metrics likes MAPE and RMSE often fail to pick up spikes which represent critical, demand altering events.Here we detail the various limitations of this approach using theoretical and real-world examples, and present alternatives toMAPE, which, depending on the situation, can provide much greater forecast ...

From Issue: Building a Holistic Supply Chain with Consumption-Based Forecasting & Planning
(Fall 2020)

The Limitations of MAPE & The Error Metrics You Should Be Using