Assessing forecast model performance in an ERP environment

Abstract

PurposeThe paper aims to describe and apply a commercially oriented method of forecast performance measurement (cost of forecast error – CFE) and to compare the results with commonly adopted statistical measures of forecast accuracy in an enterprise resource planning (ERP) environment.Design/methodology/approachThe study adopts a quantitative methodology to evaluate the nine forecasting models (two moving average and seven exponential smoothing) of SAP®'s ERP system. Event management adjustment and fitted smoothing parameters are also assessed. SAP® is the largest European software enterprise and the third largest in the world, with headquarters in Walldorf, Germany.FindingsThe findings of the study support the adoption of CFE as a more relevant commercial decision‐making measure than commonly applied statistical forecast measures.Practical implicationsThe findings of the study provide forecast model selection guidance to SAP®'s 12+ million worldwide users. However, the CFE metric can be adopted in any commercial forecasting situation.Originality/valueThis study is the first published cost assessment of SAP®'s forecasting models.