A recent post mentioned the issue of tolerancing mating components.
One may well suppose that in cottage industry, where a few mating
pieces are needed, the allowed deviation from a mean dimension
may be quite large. The probability is that the few samples made
will be close enough to the mean to mate.
But what about mass production?
If the goal is zero rejects, [and that is always economically
attractive] then the number of products that lie several
deviations from the mean grows larger, and more rejects ensue.
What to do? what to do?
ONE way is to reduce the deviation of the production process,
which one way or another, costs money.
ANOTHER way, sometimes seen is to selectively pair the mating
components. If outlyers are marked, and paired with
corresponding outlyers in the mating component, then direct
production cost is reduced, but now a selection, marking, and
logistics step is added. [An American auto business strategy]