And asking people tricky questions doesn't tell you about people's problem solving skills either, it just tells you that person is a good talker. Justifying tricky tree traversal interview questions by comparing it the DOM??? That is a stretch. In any case, beyond _solid_ coding skills, the thing that makes someone a great member of a dev isn't coding skills, it is a whole pile of professional best practices, social skills, and general passion for continuous improvement. It is simply not possible to test for those things in an interview.
The reality is that interviews are broken. Not because of this guys subjective perception that it is so but because for employers you just aren't getting the SNR to justify typical interviews. People who cling to that approach anyway are largely, IMO, motivated by ego and cargo cult (lack of understanding and creativity). In addition, in that context, interviews are also broken because in being worthless, they are demeaning to the candidate who isn't great at tap dancing on your command.
What does work is past performance and actually working together. Both are problematic data to get at so I don't think there are easy answers here. We do a resume review to see if they even claim to have the expertise we're interested in, a very short and simple "gut check" coding exercise (not tricky, just checking they can actually write decent code and tests), a 30min phone conversations where we check that both parties are aligned on what we're looking for, contract to hire, then exercise extreme discipline in parting ways with people that aren't great before converting to W2. Our SNR is pretty good. A lot of people don't want to contract to hire so this system has cons. YMMV of course.
The reality is that interviews are broken. Not because of this guys subjective perception that it is so but because for employers you just aren't getting the SNR to justify typical interviews. People who cling to that approach anyway are largely, IMO, motivated by ego and cargo cult (lack of understanding and creativity). In addition, in that context, interviews are also broken because in being worthless, they are demeaning to the candidate who isn't great at tap dancing on your command.
What does work is past performance and actually working together. Both are problematic data to get at so I don't think there are easy answers here. We do a resume review to see if they even claim to have the expertise we're interested in, a very short and simple "gut check" coding exercise (not tricky, just checking they can actually write decent code and tests), a 30min phone conversations where we check that both parties are aligned on what we're looking for, contract to hire, then exercise extreme discipline in parting ways with people that aren't great before converting to W2. Our SNR is pretty good. A lot of people don't want to contract to hire so this system has cons. YMMV of course.