Therein lies the problem. How much is an avoided bug really worth? A > where really a >= is needed is theoretically going to get caught with pair programming, but how do you make that a metric you can track when we still don't have any way of measuring developer productivity. We never got past the fact that measuring lines of code output is dumb, to any other sort of metric. Even if we had such a metric (using magic, perhaps), would programmers actually welcome it?