Many Observed Incorrect Implementations of CMMI High Maturity
At the CMMI conference a number of difficulties with high maturity, CMMI levels 4 & 5 quantified management. Some included incorrect use of standard statistical techniques for example:
- Too few data points
- Updating control charts too infrequently
- Bad formulas in control level calculations
- Trying to statistically manage inappropriately
- Using aggregated data
- Presenting baselines as stable when they were not
- Lack of quantification of quality and process performance objectives
- Lack of management use of information
- Using trivial cases to show CMMI levels 4 & 5
They also discussed how many organizations just wanted a number.. that they didnt really have a commitment to process. And how that is (finally) changing.
I was thrilled to hear that practice is changing. Over the years we have been disturbed by the trend to get a small project or organization certified, then do business as usual throughout the company.. that is process not being used.
This is the reason SEER for Software provides approximate CMMI level as an output rather than an input. When CMM first came out people asked for knowledge bases or parameters the woudl change the estimate just based on the CMM level. We actually built some draft knowledge bases based on the characteristics of organizations with higher maturity but never released them because we realized that organization CMM or CMMI level does not generally impact the individual projects on a per sey basis. We realized that people, process and technology inputs to SEER actually predict the effort, schedule and effective CMMI level best.
SEER estimates any CMMI level well. It used the root causes of improvement (people, process, technology, reuse, etc.) to generate such estimates. The other reason this is so powerful is because management and other stakeholders can see and understand the specifics of why their estimates change.
Related posts
Comments
Leave a Reply
