Analyzing and Interpreting Test Results - Software Testing: Testing Across the Entire Software Development Life Cycle - page 195

Project A drops below the prior project curve after the peak because the better
testing activity has left fewer defects to discover after the peak. The Project B curve
stays above the prior project curve after the peak for just the opposite reason that the
Project A curve dropped below. Project B testing is not fi nding as many defects as
quickly as the prior project. Therefore, Project B will continue to fi nd a large number
of defects long after the prior project defect discoveries peaked and began to decline.
Recall that the prior project curve in Figure 12.9 extended beyond project com-
pletion to “zero defects” on the x -axis predicted 487 customer-discovered defects.
The HelpDesk reported 70 customer-discovered defects, and for this reason, a pre-
diction ratio of 7 to 1 will be our fi rst guess for the next project.
Extending the Project A curve through the project completion date to “zero de-
fects,” we fi nd the area under the extended curve to predict two customer-discovered
defects. Applying the prior project prediction ratio of 7 to 1, we predict only one or
two customer-discovered defects to be reported to the HelpDesk after Project A is
completed. Applying our cost factors to get a grasp of the positive economic impli-
cations of this prediction, Project A testing results implies a $952,000 cost savings
to the company as a result of investing in better testing for the next project. The
negative economic implication is that even with improved testing, the development
organization can still expect to spend $28,000 on customer-discovered defects from
Project A. Here are the calculations behind these statements:

Prior project:
70 customer defects predicted * $14,000

$980,000 in correction costs

Project A:
2 customer defects predicted * $14,000

$ 28,000 in correction costs

Possible Project A savings to company $952,000 in correction costs

Extending the Project B curve through the project completion date to “zero de-
fects,” we fi nd the area under the extended curve to predict 1377 customer-discovered
defects. Applying the prior project prediction ratio of 7 to 1, we predict 196 customer-
discovered defects to be reported to the HelpDesk after Project B is completed. Apply-
ing our cost factors to get a grasp of the positive economic implications of this predic-
tion, Project B testing results implies a $1,764,000 additional cost to the company over
the cost of the prior project customer corrections. This is due to the lack of emphasis
on improving development and testing for the next project. The total cost of correct-
ing customer-discovered defects rises to $2,744,000. There is no positive economic
implication from Project B. Here are the calculations behind these statements:

Project B
196 customer defects predicted * $14,000

$2,744,000 in correction costs

Prior project:
70 customer defects predicted * $14,000

$980,000 in correction costs

Possible Project B overrun to company $1,764,000 in additional correction

costs