Wednesday, August 22, 2012

Can Defect Density have an UCL, if so Why it should?

While doing my usual morning readings, in a totally separate context, i had a question that started to haunt me. I tried to ignore but then the urge to think about it got stronger every fleeting minute. Question was/is - About Defect Density? Can DD have a UCL, if so Why it should? I started to retrospect on experience i have gained, learning's i learnt and start to contemplate on various perspectives. My perspective still going strong compells me to say Since LCL & UCL are benchmarks set for a certain metric to track that means they imply . ....At least and At Max. W.r.t DD it is basically the ratio of total number of Valid defects to the total uniquely executed test cases. Considering the two premise, i feel DD is a measure of Code quality more than tester's ability to find defects. From testing perspective, it perfectly makes sense to say that we will find atleast x% defects (read LCL). But when we start to call out that from very same testing perspective this is the Max defects we can find, we start to tread a wrong path. Isn't it? Because there is no heuristics, no model, no methodology to tell the max limit of defects testing team (or tester) can find. If on different platform we can always can calculate the Defect Injection rate (the total minimum # of defects that will be there)...but NEVER MAX. Secondly Even a developer cannot give a commitment to the max defects that can be found in his/ her code. Hence my strong opinion has it that no matter what we can never set a benchmark for Max density (UCL). If thats the case, then WHY is it that while formulating the SOWs, SLAs we always put a LCL & UCL for Defect Density as well. Is it not the time for a challenge to our canned approach? Let me know your that we can create a disruption in the way we have been working all this while... - Manav Ahuja