This method was developed via an extensive data mining activity of the practices employed by software development organizations. It mathematically correlates defect density to a variety of factors such as people, techniques, execution, inherent risks, etc. Data was collected on a variety of application and industry types. The algorithms are updated on a regular basis as SoftRel, LLC collects more industry data.
Softrel's defect density predictions are recommended in the IEEE 1633 Recommended Practices for Software Reliability, 2016.
People, practices, techniques, processes and inherent risks are evaluated using a detailed survey and then one of 7 groups ranging from World Class to Distressed is determined from the assessment survey responses. That percentile group is then used to measure defect density. The smaller the percentile group, the lower the defect density.
The prediction can be done as early as the proposal phase. Even if you have absolutely no historical information and no component related information, a measurement can still be made using industry data. If you have COTS components, the above process still applies. If data is not available from COTS suppliers, there are default techniques available.
Software defect density prediction