The value of QA is a very delicate balance:

Value - stuff should work after it's been through them

Costs that I've seen - more stuff doesn't get done/fixed because of the delay time associated with QA - things don't get scheduled because of lack of QA bandwidth - additional specifications required in order to go through QA - programmers / PMs worry less about catching bugs because they feel it'll come back from QA

When is it better to QA something rather than just measure when it's broken? Everything we do goes through QA but I'm sure most of it is a net loss.

Anyone have a good way of knowing what to QA and what to simply monitor?