When you start a project, you need to plan how you’re going to assure its quality. Therefore, you need to know how much effort to put into quality assurance. Can we predict whether an artifact will be good or bad, before we start to construct it? At an early stage, can we estimate the number of defects that will be present and our own effectiveness in finding them? Commonly, government projects have their quality assurance outsourced to specialist verification and validation companies. This paper comes from such a company.

The authors show that adding expert opinion to historical data makes for better predictions. They focus solely on requirements analysis artifacts, and they present their hybrid defect content and effectiveness early prediction (HyDEEP) method. The research consists of a painstaking, statistical, focused, and convincing study of five real projects. It shows that the defects present and the proportion found are best predicted by combining expert opinion with historical data. The authors present interesting tables of the factors used by the experts. These factors, when present, lead to higher defect rates and more effective defect discovery.

Although managers can get some ideas from it, the paper is really for researchers. It includes a good survey of the literature. The paper ignores agile processes and focuses on just the first phase of a classic waterfall process. The results may apply to design and code artifacts, too.