Warranty prediction during product development: Developing an event generation engine

In 2010, high tech industries including computer makers, peripherals manufacturers, and medical equipment manufacturers spent a total of $8 billion on warranty. Reducing warranty costs improves the manufacturer\u27s profit and helps to reduce the overall cost of the product. An often cited principle is that approximately 80% of the eventual product cost is \u27locked in\u27 during the very early stages of product development, however, traditional methods of warranty analysis are not well suited to predict the warranty costs during these early stages. Thus, product development personnel need better tools to make good predictions about the warranty costs so that they can make better decisions to reduce warranty earlier in product development. In order to address this gap, previous research defined a warranty prediction framework, which at its core was a warranty event generation engine that integrated the disparate data sources available early in the product development process. The objective of this work was to create this event generation model, which would give the probability of occurrence for a warranty event, given the length of time of service for the system. The model developed in this work used different data sources namely, field data, product development data and engineering judgment data from our industrial partner. The datasets were then combined using a two-stage numerical Bayes method to predict the probability of occurrence of an event. Various test cases were created, by using the different datasets as priors and likelihoods. The results were then compared to actual field data set to understand how well the model performed. It was found that the model performed well and was able to produce a bounded solution. The future research agenda is to create a tool for product development professionals that will help them predict warranty costs