[saintgimp]

The problem of standardized testing has a direct analogue in software engineering that makes the situation crystal clear to any programmer - remember when our industry thought it was a good idea to measure programmer productivity primarily (or even solely) by the number of lines of code written? How did that work out for us? Terribly, of course. LOC metrics have a passing resemblance to the thing you want to measure in software engineering but it turned out to be horribly bad at predicting the thing we actually cared about - working software.

Standardized testing bears exactly the same relationship to the outcome we actually care about in education.

In both realms, there is no easy single metric you can turn to. In the end, you hire great people and get out of their way. The best measurements come from multi-dimensional feedback from managers, peers, and customers. No it doesn't scale very well but it's the only thing that works.