Software testing and sequential sampling

The paper briefly describes sequential sampling and explains how such sampling procedures can be integrated into a software testing strategy to provide benefit. Using first-pass failure rate, (a simple software testing metric), as a quality indicator, the authors show theoretically and empirically how sequential sampling plans can be used to establish a mechanism with known statistical confidence/risk which allows for decisions to stop testing when product quality either meets the target or is worse than the expectation. Quality assessment, as well as quality certification roles, can be addressed. When sequential sampling is coupled with other simple statistical techniques, information can also be provided to help identify fault prone areas. Data from releases of the 5ESS Switch project are used in several examples to validate the usefulness of the techniques. >