Most companies maintain warranty databases for purposes of financial reporting and warranty expense forecasting. In some cases, there are attempts to extract engineering information (e.g., on the reliability of components) from such databases. Another important application is to use warranty data to detect potentially serious field reliability problems as early as possible. When a serious problem arises, the existence of the problem will eventually be obvious. Early detection of serious problems through the use of sensitive statistical methods, allowing early action to mitigate potential reliability problems, could save large amounts of money and product good will.

This paper describes a detection procedure that has been designed for this purpose. In addition to the statistical decision rules, we suggest graphical tools for illustrating and describing the particular information in the data that caused the potential problem to be flagged. The methods are illustrated using data from an automobile warranty database.