The Personal Software Process: An “ego-centered” improvement paradigm

Author(s): Daniel M. Roy
Venue: Proceedings of the Software Engineering Process Group Conference
Date: 1996


This study presents results obtained from an instructor after giving a course to engineers at the Lawrence Livermore National Laboratory. The data presented indicates that reviews are two to four times more effective than tests at removing defects. The paper also claims that the average number of defects found was reduced by a factor of six and that overall productivity slightly improved by the end of the PSP training.

The validity of the results presented in this paper should be questioned as they indicate significantly higher levels of improvement over other studies. The average Errors/KLOC was reported to be 18 while most other studies report defect ratios between 44 and 70 errors per thousand lines of code. Also, the improvement is significantly higher as other studies report a reduction of defect between 45% and 80%. It should also be noted that the sample size for the data was very small since of the 18 students that participated in the training, only 7 completed the class.