Is It Good Enough To Ship? Predicting Software Quality

Software quality often gets lots of lip service and little else-until its absence triggers a disaster and stuff hits the wall. Don Beckett shares work he did to determine when the software for a satellite destined to orbit the earth was sufficiently stable to risk being launched. Failure would have cost hundreds of millions of dollars. Don shows how he modeled this problem to answer the “launch/don't launch” question. Beginning with an analysis of the factors that determine acceptable quality and the issues that confront defect collection, Don overviews how defect discovery follows a Rayleigh curve distribution that anyone can use for predicting defects remaining in a system. He shares a model of how staffing and scheduling trade-offs will almost certainly impact defect creation rates. Don discusses the advantages and challenges of using “industry wide” defect data to predict software quality and offers practical tips on what to look for when comparing your systems to industry benchmarks.