Many companies have embraced Continuous Improvement as a means of ensuring that a product or process is performing satisfactorily. Yet many also fail to reap its benefits because the quality measurements they still employ are too crude for today’s competitive marketplace.

Take for example the way in which the performance of an engineering part is measured (see exhibit). As long as the product falls within a given specification range, performance is judged to be satisfactory. This range includes parts functioning at the lower edges of their specification limit as well as parts which are bang on target. Given this, in the exhibit, are the "good" parts A and B equally acceptable, or is part B, in fact, more like the rejected part C?

Such questions become important in the context of parts or processes which have to perform in unison. The door and door frame of an automobile, for example, can be made out of many small stampings welded together. If each part, passed as "good," only barely meets its specification, the overall result is likely to be an ill-fitting door. That is because tolerances can—and do—add up. In complicated products the cumulative effect is a loss of...