Interviewing is a random event for many companies. You as a candidate could fail miserably on one day and get great feedback the other day. It depends on so many factors in larger companies. For example in one of my previous employers, they would send a random engineer (juniors included) to interview a candidate and give a mark from 1-10. There were engineers who would not pass their own interview and questions, and on the other hand, there were others who would just see that you can code anything and you are cool to work with and would let you in. And anything in between. There were times where juniors wanted to be smart and asked some puzzles they got from the net, and when later I asked them to solve them they failed. So many interviews depend on so many random events and circumstances and.

The biggest takeaway I got from all this experience is that as a candidate I should not be intimidated by anything and should just be cool and myself, because there is so much randomness that trying to find reason in it will just make things worse. And ALWAYS negotiate as much as you can because your biggest raise is when you get hired.

Many companies instruct candidates should study for interviews with specific books or courses (e.g. “Cracking the Code Interview”).

That indicates the idea is not that someone who is a competent professional should pass without studying. The passively hostile whiteboard coding exams are more like a hazing ritual than an actual job performance test.

I prefer that over being ambushed by unexpected topics. Here, you measure ability to learn, at minimum. Unfortunately, you measure also how much free time candidate has. Which is still better then having candidate to guess what is popular now or which if competing preferences your company has.

Apparently there are already companies that take it a step further. They test sample of their employees with some fancy IQ test that measures them across several dimensions. From those results profile of desirable new hire is build and all candidates get measured by the same test.

On reddit in threads about interviews I've already seen couple people discussing those strange tests where in many questions there are no obvious correct/wrong answers, like for example all presented shapes can in their own way fit in presented pattern.

If there is no obvious right/wrong answer, I wonder if they are limiting their pool taking people who all think alike (kinda like in-breeding) or are they using it diversify the ideas and knowledge etc. (like diversifying gene pool).

Different answers probably give points to different traits the test measures, the ambiguity is just irritating for test takers coming from more classical tests where you know (or at least suspect) what is expected from you.

This whole test setup is not cheap so one can at least hope they considered the issue of diversity? The other thing is did they actually tested which teams perform better in specific problem domain that interests them?

>That's a fascinating idea that's never crossed my mind. E.g., how many current employees can pass whiteboard coding exams?

I once confronted a colleague who had refused a candidate because he couldn't answer a technical question. When I heard the question, I asked him "How many people in the current team can answer this?" (Hint: Only one). His response: "Yeah, but we told him to read the first 6 chapters of textbook X!"

So we denied him a job because he didn't do his homework.

Testing the change is one thing, but if the original people were hired and asked something very similar, you have an existing group selected based on that characteristic and so will likely do well.

e.g. if you have always done a whiteboard interview, then you've selected for people that do well in that type of scenario. Changing the question but keeping the format will likely just prove that those people are still suited to that type of task, but not that an otherwise good hire will be filtered out.

I'm not against whiteboard interviews or similar, but just pointing out that selection bias is an important factor here if you're judging effectiveness by comparing it to results for existing employees.