The method I prefer is to look at our ability over time to deliver. If we cannot sustainably deliver continuously then the code is crap. If we can and everything breaks then the tests are crap.