The issue is filter accuracy. I have seen some of my favorite coworkers, who showed great pairing skills in the real world, melt down in typical algorithm interviews. I've also seen great performers in interview that were obvious terrible hires, from a purely technical perspective, within a month.
I've interviewed four digits worth of candidates, under different rubrics and different expected difficulty levels, and all the good I can say about the modern interview is that at least it tends to crib out the people that can't write code at all, at least if you are making sure nobody is feeding them answers. But can I say that interview performance with me, and how well I rated the person's work when they were hired and ended up working close enough to me, had much to do with each other? I don't think so.
That's why, when working at a small enough place I have some control over the interviewing process, I'll offer options to the candidate, and dedicate far more time to the process than I would in a large software firm. It's OK to just raise the technical bar enormously when you are offering the best salaries in the market, and you can expect to never run out of candidates. But when you are not competitive, you have to do something to find great candidates that don't look wonderful in a FAANG style interview format, but will be very good in practice anyway.
I've interviewed four digits worth of candidates, under different rubrics and different expected difficulty levels, and all the good I can say about the modern interview is that at least it tends to crib out the people that can't write code at all, at least if you are making sure nobody is feeding them answers. But can I say that interview performance with me, and how well I rated the person's work when they were hired and ended up working close enough to me, had much to do with each other? I don't think so.
That's why, when working at a small enough place I have some control over the interviewing process, I'll offer options to the candidate, and dedicate far more time to the process than I would in a large software firm. It's OK to just raise the technical bar enormously when you are offering the best salaries in the market, and you can expect to never run out of candidates. But when you are not competitive, you have to do something to find great candidates that don't look wonderful in a FAANG style interview format, but will be very good in practice anyway.