Information Technology Reference
In-Depth Information
Carrying the Project A analysis forward from the start of the project, we are
expecting the Project A trend to continue to reinforce our interpretation that testing
is doing a better job of fi nding defects during this project. Both coordinates of the
Project A peak tend to reinforce our positive interpretation: The Project A peak oc-
curs at a larger number of defects discovered—1213 defects for Project A compared
with 749 defects for the prior project. The Project A peak also occurs sooner in the
current than it did in the prior project: defect discoveries peaked at the 6-week mark
for Project A compared with the 9-week mark for the prior project.
Carrying the Project B analysis forward from the start of the project, we are expect-
ing the Project B trend to continue to reinforce our interpretation that testing is not neces-
sarily doing a better job of fi nding defects during this project. Both coordinates of the
Project B peak tend to reinforce our suspicions and begin to persuade us that the news is
worse than originally expected: Testing is not as effective as last project at discovering
defects. The Project B peak occurs at a smaller number of defects discovered: 607 defects
for Project B compared with 749 defects for the prior project. The Project B peak also
occurs later in the next project than it did in the prior project: Defect discoveries peaked
at the 12-week mark for Project B compared with the 9-week mark for the prior project.
The critical success factor for the development team is to decide to minimize these
undiscovered defects in a proactive, cost-conscious way. A starting point for minimizing
undiscovered defects before development completion is to fi nd ways to predict the ap-
proximate number of expected undiscovered defects. Recall from Figure 1.2 that these
undiscovered defects cost $14,000 on average to correct after the customer fi nds them.
The fi nancial argument for redoubling the testing and correction effort before project
completion would be different for a predicted 10-20 undiscovered defects (maximum
estimated correction cost around $280K) than it would be for a predicted 300-500
undiscovered defects (maximum estimated correction cost around $7million).
One way to predict the approximate number of expected undiscovered defects
is to extend the defect discovery curve in Figure 12.13 on out beyond the project
completion date as we did with Figure 12.9. Determine the approximate area under
the extended curve. The area under the extended curve represents a workable ap-
proximation. Figure 12.14 shows what the prior project, Project A, and Project B
curves might reveal about predicted customer-discovered defects.
Figure 12.14
Defect tracking log comparison
Search WWH ::




Custom Search