I just ran across some great stats on the cost of software defects. These are quotable, so I thought I would share.
The following is a quote from Capers Jones in his book, Estimating Software Costs. For those unfamiliar with Capers work, he is one of the driving forces in the field of software estimation–especially in the world of traditional development processes. He’s been working full time doing software estimation since 1971, and now runs a consulting company on software estimation.
The most expensive and time-consuming work of software development is the work of finding bugs and fixing them. In the United States the number of bugs or defects in requirements, design, code, documents, and bad-fixes averages five per function point. Average defect removal efficiency before delivery is 85 percent. The cost of finding and repairing these defects averages about 35 percent of the total development cost of building the application. The schedule time required is about 30 percent of project development schedules. Defect repair costs and schedules are often larger than coding costs and schedules. Accuracy in software cost estimates is not possible if defects and defect removal are not included in the estimates. The author’s software cost-estimating tools include full defect-estimation capabilities, and support all known kinds of defect-removal activity. This is necessary because the total effort, time, and cost devoted to a full series of reviews, inspections, and multistage tests will cost far more than source code itself.
Emphasis is mine. Capers also asserts that the cost of producing paper documents often costs more than the actual coding.
In regards to requirements on a traditional project, requirements grow at an average of 2% per month of development. When measured with function points, requirements growth can often exceed 50% of the volume of the original requirements. </ p itself.