In service logistics, OEMs or life-cycle product support service providers are often required to make decisions based on sparse data. This is due to the low-frequency failures of highly reliable equipment. As a result, most OEMs or service providers face tremendous challenges in attempting to achieve the contractual service levels, both in the initial phase when historical data is not available and during service contract execution. This paper contributes to the existing research in two respects:
– We distinguish aleatory uncertainties (due to stochastic failure processes) from epistemic uncertainties (due to imprecise knowledge of the failure processes).
– We show that the contractual service level could be jeopardized by ignoring epistemic uncertainties in the decision making process. However, this adverse effect can be eased by risk pooling and an efficient use of the available information.
We first study the service performance (service level and cost) under aleatory uncertainties only. In this situation the parameter of the failure process is assumed given. To model the reality of imprecise knowledge of the stochastic process we then assume that only the distribution of the parameters is given. We compare these two scenarios by calculating the expected shortfall in contractual service levels when decisions are made without taking epistemic uncertainties into account.
We than study the risk pooling effects on the expected shortfall in both initial phase and contract execution phase. We employ Bayesian inference to update our state of knowledge using operations data during the contact execution phase.
Our numerical experiments show that the expected shortfall can be reduced by pooling risks.
We applied our method in a case study at Fokker Services B.V., a performance-based aircraft maintenance, repair and overhaul (MRO) service provider. Our results are very promising and are supported by engineers’ and experts’ opinions.