Thanks to Jim McHale for finding the follow up to Kent Schneider’s story.
Steve Brady and Sherri Turner tell us more about projects A and B, (they reverse the project names) and then about project C
Project A was their baseline.
Project B introduced developer design packages, used PSP, and peer inspected 20% of the “critical code”
- used historical data for estimates
- implemented PSP during the contract proposal
- introduced 100% peer inspections
- implemented other CMMI level 4/5 practices
The defects/KLOC delivered to test in project C were a factor of 14 lower than those of project B, and a factor of 50 lower than project A.
Project C had 0% cost overrun, project B, a 32% overrun, and project A a 47% overrun. They then discuss the ROI
Training for project B cost ~$33,600. Peer reviews cost $70,000, for a total of $103,600
The returns on project B were
- test was completed at 25%
- Projet B was accepted 7 months before project A
- SPI improved from 0.746 to 2.657
- reduced cost overruns by 12 months over project A
Project C inspected 100% of the code the return was
- prevented 110 critical defects
- improved cost and schedule performance
- 4 people tested 700 + cases in 6 days (only 0.36 defects/KLOC entering test!)
They calculated that the additional cost of $11,160 saved $148,500
Read the rest of the report. Their bottom line was
“we can’t afford to manage projects without process”