Part of the problem is that an imperfect measure often isn't better than no measure, unless you're very careful about how the measurement is used. You end up wi

Part of the problem is that an imperfect measure often isn't better than no measure, unless you're very careful about how the measurement is used. You end up with people devoting huge amounts of resources to gaming the metrics, which produces worse outcomes than even not evaluating people at all would.