Analyzing and Interpreting Test Results - Software Testing: Testing Across the Entire Software Development Life Cycle - page 190

12.6.2 Leveraging Prior Project Defect History Before
the Next Project Starts

This brings us to a second useful piece of defect prediction information that be-
comes available between the time the prior development project was completed and
the time the next development project starts. This second piece of information is a
HelpDesk log of customer-discovered defects. The number of customer-discovered
defects will correlate somehow with the Area D prediction. This correlation may
be 1 to 1 or 100 to 1 or some other number in between. For the purposes of dem-
onstration, say there were 70 customer-discovered defects found in the Figure 12.9
release and reported to the HelpDesk. The customer defect discovery correlation is
calculated by

No. predicted
No. actual

No. predicted per actual

For example:

487 Predicted from curve
70 Actual from the defect log

6.9 predicted defects per actual defect

The goal of the exercise is to be able to predict fairly accurately the number of
customer-discovered defects that can be expected from the next development proj-
ect as it concludes. The value of this prediction is the proactive economic implica-
tions for other parts of the organization like the HelpDesk and the software support
groups. If HelpDesk and Support know to prepare for an estimated 70 customer de-
fects in the next 6 months instead of 7,000 customer defects, then the HelpDesk and
Support can be smarter about near-term staffi ng and hours of operation. Remember
that this correlation of project defect discovery to customer defects is just an obser-
vation that can be useful.
Another interesting implication is the cost of correction represented by the
customer-discovered defects. Chapter 1 places the average cost of correcting a
customer-discovered defect at $14,000 per defect. If our customers found 70 de-
fects in the last release of our software, the cost of correction to our software
organization is

70 customer defects $14,000 per correction $980,000 in correction costs

When the customer defect logging system is established, consider requesting the
inclusion of some elements from the development defect log such as a severity code
when the call is received and a code earmark when the correction has been made by
Support. This additional information will allow you to analyze the customer defects
in the same way that we suggested analyzing the defect tracking log early in this
chapter.
Figure 12.11 represents the HelpDesk log that corresponds to the customer de-
fect discovery predictions in Figure 12.9.