When molding parts, the 'target' is making an acceptable part consistently. You get no extra points for making perfect parts with every dimension dead on the mean. Like the targets on the range, the object is to not miss the target (put every part within the specified tolerance). You get bragging rights if you hit the bulls-eye, but nothing more.
Asking for or insisting on a CpK is really an admission of a mistake from the designer. His mistake is the fear that 'within tolerance' might produce a bad part. Really? Does that sound to you like someone didn't do his or her homework?
Lean, CpK, 5S, Quality Ninjas, JIT, PPM-defectives, and all the other Statistical CorpSpeak methodologies are really designed to improve profits by avoiding waste. In reality, they have little to do with quality; it's all about money.
Consider that the legendary AK-47 rifle (the Kalashnikov) became legendary, and the best-selling firearm ever, not because of advanced metallurgy, innovative design, or iconic thinking. Its reputation came from loose tolerances. It is a highly reliable machine that will do what it is designed to do regardless of where or how any component is manufactured so long as it is within the design tolerances. Loose tolerances allow it to function under almost any conditions. Should we apply the statistical constraints of Lean, CpK, 5S, Quality Ninjas, JIT, and PPM-defectives to it, its numbers would scare the Black Belts white. Somehow these suckers just keep working-without all that stuff.
So, are these tools necessary? Here's the two-handed argument:
On one hand, if the goal (lean) is to reduce waste (in the most general sense), improve productivity etc., these tools are excellent guidelines. They do not, however, improve quality. The only thing they improve is consistency. This means that a part that is going to prematurely fail will do so with irritating regularity...with a tight CpK, manufactured using Lean principles, and following 5S methods. (Read: this is a design fault.)
On the other hand, there are people who make their living and build corporate kingdoms by generating charts, graphs, and PowerPoint presentations, and holding meetings in which highly esoteric terms like Students T tests, probabilistic certainty, and robust arguments are thrown about. Unfortunately, you have to pay them and support their efforts, but let's ask the ultimate consulting question: "What did they bring to the party?" Does the cost of all this fluff and sizzle have a net impact on profits? Did their effort generate the infamous 3X or 5X in additional profits measured against their salary and overhead?
What if the designer did his job by finishing the project with a 'due diligence' stack tolerance study and functional testing showing or simulating that any combination within his tolerance would produce a functional part that would last for its intended lifetime? Would we need all this statistical stuff? Nope, but they are still good tools to have when you need them.
These are interesting questions for management to ask. Take note that the answers can be embarrassing.
So, the next time you see your tolerancing chopped by two thirds by a CpK, ask the question: "Does all this improve reliability to the point of you being willing to pay for it?" Then sit down and politely listen to the answer and consider the source. If this is being shoved down your throat because it is the Management Philosophy Flavor of the Month, do what everyone else does: Nod your head knowingly and get on with your life. Next month there'll be another philosophy.
It's your money.
End of Rant.