A Bayesian approach for the determination of warranty length

The purpose of many reliability studies is to decide the warranty length. However, a review of the classic literature on reliability reflects that it is a common practice to only recommend the use of low quantiles of the estimated distribution of time to failure. We propose a methodology that takes into account the following aspects: the reliability of the product, the consumer appreciation of the competitiveness of the warranty scheme, the effect on the image of the company when the product fails under the warranty period, and the costs that the manufacturer incurs to fulfill the warranty. The approach is based on the formulation of a utility function. Regarding the data structure, we contemplate complete sampling or censorship: type 1, II, and random. We illustrate the methodology with the determination of warranty length of brake linings.