Google kinda-sorta did a study there, but the "study" was fundamentally flawed. They didn't look at people who didn't get hired (duh). So the most they could sa

Google kinda-sorta did a study there, but the "study" was fundamentally flawed. They didn't look at people who didn't get hired (duh). So the most they could say is that once people passed the (very high, risk averse) bar, how much better they did beyond that didn't have a strong enough correlation with job performance to come out in a small-ish number of data points. To me, it's pretty clear that it basically works. It's not perfect, but yeah it works. Virtually every top tech company has been formed with these practices. So unless you're going to argue that they could have hired random engineers, then it must be effective. (Note: this doesn't mean that it's perfect or that there's nothing better or that you can't improve it. It just means that it is reasonably effective.) I [author here] have done a lot of interviews with companies going through acquisitions, and that's offered me the ability to talk with their managers about their job performance. My assessment is generally accurate. There are certain things it can't evaluate, of course, like work ethic. Work ethic is really important, but there's really no way to interview for that.