Information Technology Reference
In-Depth Information
12.6 PRIOR DEFECT DISCOVERY PATTERN
USEFULNESS
There is no current way to precisely predict beforehand how many or what kinds of
defects your test team will discover in new software. The quest for such a prediction
is often called looking for the “silver bullet.” The software engineering researcher
who fi nds the silver bullet will indeed strike gold!
Depending on the circumstances and maturity of the software development or-
ganization, predevelopment defect predictions can range from an educated guess to
reasonably accurate predictions. Project managers are understandably hesitant to
take any proactive steps based on educated guesses. Some historically based defect
prediction techniques can yield answers
10% of the actual defect totals at the
end of the project. Managers are more likely to take proactive steps based on his-
torical defect prediction techniques that have proven to be credible over a series of
projects.
The educated guess is used when the next software project is nothing like prior
projects for the development and testing teams. There is no project history basis for
gleaning predictions about the next project. The educated guess comes from the ex-
perts on the software development project that have “done something similar.” If this
explanation does not give you a warm, fuzzy feeling about the value of an educated
guess, then you understand the implications correctly. The most generous statement
one can make is that an educated guess in the absence of historical defect data is a
better starting point than nothing … but not by much.
/
12.6.1 Prior Project Defect History as a Starting Point
If there is project history either from similar software applications or from previous
releases of the same software application, then that project history can be leveraged
for next project defect predictions. The value of the project history for defect predic-
tion is directly proportional to the level of defect tracking detail that was captured in
the project history. As we have seen in the previous sections of this chapter, the prior
project can pick and choose what to include in the defect tracking log. As you would
intuitively expect, the more the details in the prior project defect tracking log, the
more useful that defect tracking log history becomes as a predictor for the next proj-
ect. We will examine the usefulness of minimal historical detail in this section and
conclude with a reference to a commercially available method that uses considerably
more historical detail for defect discovery prediction.
The minimum useful historical defect information is a prior project's defect
log that uniquely identifi es each defect discovered and its date of discovery. An
informative curve can be drawn on a two-dimensional graph showing numbers
of defects discovered by week versus project week. Assume for the purposes of
demonstration that the prior project was completed in 24 months and produced a
software application containing 200,000 lines of code commonly referred to as
200KLOC. The defect log discovery curve by week might look something like
Figure 12.8.
Search WWH ::




Custom Search