Thinking about it a bit more, I think it’s more like the metrics used to get in front of a human (the automated/hr part) aren’t well matched to the actual goals. We end up interviewing a lot of people who are good on paper according to the first sort, but actual good hires within that aren’t as common as we’d like. But none of the engineers ever know about any of the people who were disqualified due to having an unimpressive resume…

So in the end, the initial sort does indeed end up wasting time and money, but no one’s gotten around to making a good solution for this yet. The alternative so far is to interview a bunch more people, which is also really expensive anyway.

Basically, we have no efficient way to find people who are bad on paper but are actually quite skilled.