Does software testing methodology rely on flawed data?
- by Konrad Rudolph
It’s a well-known fact in software engineering that the cost of fixing a bug increases exponentially the later in development that bug is discovered. This is supported by data published in Code Complete and adapted in numerous other publications.
However, it turns out that this data never existed. The data cited by Code Complete apparently does not show such a cost / development time correlation, and similar published tables only showed the correlation in some special cases and a flat curve in others (i.e. no increase in cost).
Is there any independent data to corroborate or refute this?
And if true (i.e. if there simply is no data to support this exponentially higher cost for late discovered bugs), how does this impact software development methodology?