How to predict software defect density during proposal phase

The author has developed a method to predict defect density based on empirical data. The author has evaluated the software development practices of 45 software organizations. Of those, 17 had complete actual observed defect density to correspond to the observed development practices. The author presents the correlation between these practices and defect density in this paper. This correlation can and is used to: (a) predict defect density as early as the proposal phase, (b) evaluate proposals from subcontractors, (c) perform tradeoffs so as to minimize software defect density. It is found that as practices improve, defect density decreases. Contrary to what many software engineers claim, the average probability of a late delivery is less on average for organizations with better practices. Furthermore, the margin of error in the event that a schedule is missed was smaller on average for organizations with better practices. It is also interesting that the average number of corrective action releases required is also smaller for the organizations with the best practices. This means less downtime for customers. It is not surprising that the average SEI CMM level is higher for the organizations with the better practices