Many "traditionally-minded" companies have been using CMM or CMMI for decades. Many others have been wary of using CMMI and migrated towards lean/agile practices. Now, everyone is keen on agile and some agile outfits are even realizing the limitations of their approaches to install robust approaches that weather scaling and growth. In either case, both are struggling to make meaningful inroads towards the bringing the two together. They're not alone. In particular, traditional approaches towards CMMI make using agile approaches awkward, and, agile approaches make their traditional approaches to CMMI awkward. This is even more pronounced in "high maturity" CMMI. Agile or otherwise, "High Maturity" CMMI (Maturity Level 4 or 5) is often seen as "out of reach" for most organizations. If, for no other reason, than the perceived (and oft-reported) time and expense of establishing and maintaining the necessary processes for ratings in the Maturity Level 4 and 5 goals. Even for those who fully internalize the benefits of having high-confidence and predictable performance -- the anticipated benefits of CMMI high maturity -- the prospect of identifying, defining, baselining and modeling stable, capable, in-control, and "influencable" processes and associated input variables whose data indicate a central-tendency and whose p is null...... (You get my point) ... Are a bit intimidated by the work, rigor, and scrutiny of the brass ring of CMMI. For businesses serious about performance ML4/5 is where the pay-dirt is. If this is true, wouldn't it seem reasonable that it shouldn't take statistical geniuses or expensive consultants to come up with models that do a decent job of predicting performance? Wouldn't it seem reasonable that it shouldn't require months of poring over data to find *a* leading indicator of a process you can actually manipulate to improve your outcomes? Are software or services really all that special that SPC, baselines and models should be all that different than what's commonplace in manufacturing? Is it me, or does it seem like ML4/5 is way harder than it really ought to be for it to be all *that* useful for practical purposes? Well, as it turns out, ML4/5 can be *much simpler* than what most people experience. Using very simple data, no fancy models, and no statistical geniuses, ML4/5 data, baselines, and models can be had for any business, small or large, doing any kind of development or service work. Ironically, in fact, what makes "Simple High Maturity" possible are practices commonly found in lean, agile, and in particular, Kanban.
