Re: The Famous Ford Study of Mazda Transmissions - Can the characteristic be measured
I think a lot of people are further polluted with a perception of the "quadratic loss function" model that indicates the closer you are to the target the less "loss" there is. It is an interesting broad brush model, but you have to be careful. It does not accommodate the cost of always being just at the target. That includes (as Ford found) measurement systems that can discern that closely, as well as process controls capable of better discerning adjustment. It also assumes one facet of functionality - there are many, so the true function will be multi-modal.
If you get back to the core of variation - the total variance equation - you will see that the output is always multi-modal. Always. But, you can make many of the factors statistically insignificant - often at more of a cost (again, see Ford's dilemma with their air gages). So, real decisions need to overlay these factors to determine specifications: loss (functional), risk, impact on how to control factors of variance in the total variance equation and cost.
One other nice detail...did you see how Ford measured those bores? Did they take one diameter, as one would do with traditional X-bar R charting, to describe them accurately? No. They took the diameter and roundness, as you would do in X hi/lo-R charting. That is how you correctly control precision machining! Thanks Ford!