Analyzing and Interpreting Test Results - Software Testing: Testing Across the Entire Software Development Life Cycle - page 188

Figure 12.8

Prior development project defect tracking log

This curve demonstrates characteristics of most testing project results. There
is an initial burst of defect discovery that peaks sometime in the fi rst third of the
project. Then the defect discovery rate trails back down as the software becomes
more stable and exhibits fewer and fewer defects.
Point A is the peak (maximum) number of defects discovered during any week of
the project. Figure 12.8 shows us that the discovery peak for this project was 749 de-
fects. Point B is the week during the project in which the peak number of defects was
discovered. Figure 12.8 shows us that the discovery peak for this project occurred
during week 10. Point C is the week when the project was completed; in this example
it is week 24. Finally, Area C under the curve from the project start date to the project
completion date is the total number of defects discovered during the project. Figure
12.8 shows us that testing during this project discovered a total of 11,497 defects.
There is a body of research that suggests that the higher the defect discovery
peak (Point A) and the earlier in the testing the discovery peak occurs (Point B), the
fewer defects will be found by customers after project completion. [49] The impli-
cation is that if you compare defect curves from project to project and you see the
defect discovery peak getting higher and occurring sooner in the project, then you
are letting fewer and fewer defects get past the project completion to the customer.
Recall from Chapter 1 that the most expensive defects to fi x are customer-discovered
defects. Any trend you see in testing results from project to project that suggests
there will be fewer customer-detected defects also suggests that signifi cant software
support savings are being realized by your organization.
The same research suggests that if you extend the defect discovery curve beyond
project software completion Point C to Point D on the x -axis, then Area D under the
curve between Point C and Point D will approximate the number of defects found by
customers. Figure 12.9 shows how the defect curve in Figure 12.8 might be extended
to estimate Area D.