: Watts S. HumphreyVenue
: SEI interactive,Software Engineering Institute, Carnegie Mellon UniversityDate
This paper presents a condensed version of a three part series publish in CrossTalk from February to April 1998. The results are very limited and present what seems to be the same data as published in “Results of Applying the Personal Sotware Process” (1997) Pat Ferguson, et. al.
The paper claims that engineers from Advanced Information Services (AIS) decreased their estimating error from 55% down to 27% after receiving PSP training. Additionally, compile and test defects decreased from 110 defects/KLOC to 20 defects/KLOC with lines of code productivity remaining relatively constant. The paper also presents results that seem to indicate that system testing time drastically decreased from a couple of months for projects to several days. Upon closer inspection it is seen that a 15,800 line Non-PSP program took 1.5 month to test and testing of a 11,700 line PSP program also took 1.5 months. While the authors want to convey that PSP has a huge benefit it is unclear how accurate these results are as results for Non-PSP test results are measured in terms of the number of requirements for the system while PSP results are documented by the number of lines of code of the project.
Overall, this paper is relatively insignificant. First, the paper isn’t published in an actually journal but was rather just released by the SEI. Second, few details are provided to allow other to replicate the work and comparisons are drawn between Non-PSP and PSP results despite the fact that there is not common measurement for the complexity and size of the program.