A non Parametric Estimation of Service Level in a Discrete Context

Abstract

An exact method for the estimation of the cycle service level has been proposed for periodic review stock policies in a discrete demand context for any known i.i.d. demand distribution. However, the implementation of this method in real environments has previously to manage some important and eventually cumbersome issues such as: (i) the identification of the appropriate demand distribution and its validation; (ii) the estimation of the parameters of the demand distribution; and (iii) the calculation of temporal aggregates of the demand distribution in order to estimate the expected service level. This paper shows some difficulties linked to these issues and proposes an alternative approach based on the observed demand frequencies, so that these issues are avoided and the accuracy of the service level estimation seems to be improved.