Steven Covey discusses two types of costs in the production of products. There are the costs associated with the production of the product itself, and then there are costs associated with improving the way that you produce those products, which he describes as the means of production. For example if you are a farmer, the costs of production include the operational costs of labor and consumables necessary to grow and harvest your crop. However, before the farmer plants his (or her) crops, he may need to improve the yield he expects at harvest time, by considering the methods he has used in the past, and alternative methods available to him in the current season. Yield is typically enhanced by utilizing more effective means of production - better seed, better equipment, and so forth - but not all of these new methods will be affordable, or can be adopted at the same time.

Joseph Juran introduced the concept of the cost of quality as a way to 'quantify the size of the quality problem in language that will have impact on upper management'. This concept means that we need to learn how to deal with quality as an economic, rather than emotional, issue. What this really means is that we must trading off short-term pain for longer-term gains. While the long-term benefits of improved product quality and reduced testing will become quite evident given enough time, they require investments up front, usually which requires more resources, discipline, and diligence than may otherwise have been planned for. As a result, for quality gates to be used consistently, it is essential that ways are established to demonstrate positive feedback from these investments in the short term.

The costs of quality have three discrete components:

• failure costs - the costs of diagnosing in-service defects, making the necessary repairs to those units, and getting back into operation
• appraisal costs - the costs of evaluating products to determining their quality
• prevention costs - the costs associated with identifying the causes of defects and the actions taken to prevent them in the future

Examples of types of failure costs includes the time you spend in troubleshooting problems, warranty claims, and the effort it takes to correct and re-release a product that was expected to be ready to enter into service.

Examples of different types of appraisal costs include the time that is spent in conducting reviews and inspections.

Prototypes may be considered as an example of a prevention cost, since nearly all reasons for developing prototypes can trace back to avoiding future defects.

Consideration of all of the above require a coherent operational model with respect to several key concepts:

• defects, or things which cause changes to a product or its definition or design. Note that all changes which are associated with the same problem would constitute the same defect
•  

 One of the problems with CMMI appraisals is that they tend to drive bad behavior. This is because they strive to achieve "consistency" (or "objectivity") and I can fully understand why the SEI would feel that such an approach may lead to far greater "repeatability" - but I'm not sure that they realized the unintended consequences of that approach, as they take away "professional judgment".

Some CMMI practices are written as "good practices to be performed" more than "easy practices to be appraised."  As aresult, an appraisal's focus on direct artifacts can be somewhat of a crap shoot. For example, how many times have you seen organizations put people through a 2-hour DAR (or PP or PMC or TS or...) training sessionso they can establish artifacts that will satisfy an appraisal team?  The training practice says, "... as needed," but because of the lack of specificity of that criteria, and concerns about the consequences of not meeting it on failure to achieve their target maturity level, lots of unnecessary training is conducted.

So the organization adopts a defensive posture by investing time in developing and providing "training" that is, at best, "orientation."   I can sit through a two-hour "training session" on Software Architecture - but that doesn't make me an architect!  Conversely, there are a number of software architects that would be wasting their time sitting through a two-hour session that they themselves would be more than capable of teaching.

I firmly believe that most of the model's practices are capable of providing tremendous value to most organizations - but it's the fear of an appraisal that drives bizarre behaviors in a given organization's implementation of those practices.

I also believe, as you alluded to in your post, that the very experienced (and less experienced, but very insightful/pragmatic) lead appraisers have found a way to restore the balance between direct artifacts and professional judgment.