Improved testing using failure cost and intensity profiles

The cost of field failures is arguably a very important factor in judging the quality of software. It would be nice if field failure costs could be used before the software is released so as to improve its quality. We introduced the expected cost of field failures (Ntafos, 1997) as a measure of test effectiveness and demonstrated that significant benefits are possible even with rather crude cost and failure rate estimates. We discuss the main issues that arise in trying to reduce the cost of field failures through better test effort allocation. We introduce improved models and strategies and present analytical and experimental (simulation) evaluations