Importance of Counting             

In a previous post we discussed the perception that QA may be counting too much.  Before we go deeper into that consideration, one of the other concerns is the reliability of the measurement.  We define Reliability as getting the same result if the measurement is taken under the same conditions twice.  Whether that result is correct is a different question but if the measurement is different when taken under the same conditions, it is not only invalid; it is also unreliable.  This can cause real problems, as it has in some projects.  Measurements taken by the vendor differed from the ones taken by the customer.  Not surprisingly the vendor measurement showed less outstanding defects than the customer’s ‘equivalent’ measurement.  It was eventually traced to one label used on only a few defects that was being excluded by an earlier filter on the vendor side.   Once that condition was removed, the counts agreed.