Kai Petersen MSE-2006:04, pp. 67. TEK/avd. för programvaruteknik, 2006.
Software Inspection is an effective and efficient method aiming at discovering faults within software artifacts early in the development lifecycle. The success of software inspections is highly dependent on reading techniques that guide the reviewer through the individual inspection. In other words, reading techniques help the reviewer during the inspection process. In this thesis a quite new reading technique, namely usage-based reading, is further evaluated. A reviewer who applies usage-based reading is guided by a set of prioritized use-cases. Hereby the use-cases are ranked according to their importance from the point of view of the user. Thus, the reviewers inspection effort is focused on the parts of the document that concern the most important functionality from the user’s perspective. The particular goal of this study is to figure out whether time-budgets assigned to each use-cases lead to improvements in inspection performance expressed in terms of efficiency, effectiveness and fault content. This concept is called time-controlled reading. The time-budget determines that a particular use-case must not be inspected longer than a certain number of minutes. Initially the assumption was made that time-budgets lead to performance improvements because the reading process can be better planned in forehand, that means most time is spent on the most important use-cases and all use-cases are utilized within the given inspection time. However, the result of this study contradicts that assumption. I found that both techniques are equally efficient and effective and that they find the same fault content. The reason for this is that the techniques are still quite similar and that the timebudgets assigned to the use-cases did not allow the subjects to thoroughly investigate the use-cases because they struggled with
the unknown application domain. This is at least the case for lower ranked use-cases with smaller time-frames. As a
consequence of this, one can claim that the results might have pointed in favor for time-controlled reading when people familiar with the application domain would have done the same experiment.