The truth is that companies have almost no clue how to differentiate high performing hires from low performing ones. There are a ton of 'tests' that claim to take the top X % of developers, but there is little evidence that their metric correlates to success. (The same way that top 3% SAT doesn't highly correlate to high performance in college/industry).

These tests ARE good for general measures of capability though, and should be used to weed out the wildly unqualified. For example, an SAT score of 1000 is way different from 2000.

But then again, if you don't use a test to determine quality, you have to fallback on personal judgement, and that doesn't work any better.