How do we measure ourselves that we performed better than the previous release in terms of defects? If the number of defects found in the current release are less than the previous one, do we say we have improved? I feel it is not a good measure as the complexity of the features implemented in the previous release might have been more than the current release.
Can someone suggest me how do I normalize the complexity here? How do I use the normalized day to see where I'm with respect to the previous release?
Do you have different levels/priorities for defects (severe, medium, low)? Does that normalize it? Would adding a "complexity" category to the defect do it? This could be 2 fields like "developer complexity" (how hard to code) and "QA complexity" (how hard to test). Or how about some cyclomatic complexity type metric to the code? I would say the number of defects found in the field is a good measurement if some normalization is done or not. Maybe relate the defects to usage analysis...if the bugs are found where the code is 'used' more, then this would be worse for QA compared to bugs found in rarely used code or usage scenarios. On the flip side, bugs found pre-release in rare scenarios would be more 'points' for QA. Just some thought off the top of my head. Good luck and please let us know if you figure out anything.