Requirements attributes to predict requirements related defects

Literature suggests that requirements defects are a very costly problem to fix. Understanding how requirements changes influence the overall quality of software is important. Having some defect predictors at the requirements stage may help the stakeholders avoid making choices that could bring about catastrophic defect numbers at the end or at least be prepared for it. In this paper, six requirements-related attributes are analyzed to discover if they can be used for determining the occurrences of requirements-related defects. We measured two types of attributes: point and aggregate. The point attributes include time estimates, priority and ownership. The aggregate attributes include the number of indirect stakeholders, the number of related stories and the story creation time. Our analysis is based on data from the development of the IBM Jazz system. Our result shows that the number of indirect stakeholders and the number of related stories are good predictors for the number of defects, but other attributes show no or little correlation with the defects.