Forecasting Fleet Warranty Returns using Modified Reliability Growth Analysis

Forecasting the performance of a product on the market allows for quick correction of design and engineering dependent failures, repairs of eventual breakdowns and forecasting repair and warranty expenses. Warranty data can be used as a base for product reliability prediction according to various literature theories. In the first part of this paper, these theories are analyzed underlining the pros and cons. Neither classical reliability theory, nor Peugeot-Citroen model or RGA (reliability growth analysis) seem to be able to model the fleet behaviour in terms of failure prediction. Therefore the pros of each model have been grouped up to build up a new hybrid model. Then the paper describes the new model that is based on RGA but modifications have been necessary in order to cope with the problem of missing data relevant to the so called untraceable vehicles. Censored data occurred because of the fact that the data used comes from field tests (from customers) instead of in-house tests (from professional testers). Therefore the RGA model has been implemented with estimation of fleet width, taking into account cancellations and thefts. Finally, the paper explains how the model seems to offer a wide applicability to any firm/product provided with data coming from the field. The model is currently applied by one of the main European motorcycles producers. Even if the development is still in progress, the company productively uses it in order to estimate the number of repairs requested under warranty and some cases are presented here.