Hacker News Reader

Comment by wongarsu | Hacker News Reader

At my startup we did a round of "come in for an interview, pass FizzBuzz on a whiteboard, if you passed FizzBuzz you can come back for a 4 hour trial where you solve a small task representative of normal work in our normal work environment". We only hired students for 10-20 hours per week, so fairly low risk for both sides.

My main takeaway from that was that the actual talking was close to useless for judging candidates. One of the best people we hired looked horrible in the interview, while some others were great at talking about themselves but were at best mediocre otherwise.

View on HN

In my experience interviews are less effective with students because they rarely have relevant experience to talk about. At the mid level and above tests and technical questions become less useful, but they can talk about meaningful past projects. I don’t know of anything else that’s predictive without weeding great people out.

PS: It’s odd, but I suspect a random process may actually work better because the great people actively looking for jobs are likely to be bad at getting them.

When I was nearing graduation and doing interviews, the biggest question that stumped me was a "Where do you see yourself in five years". I was a few months away from finishing my masters, so between that and looking for a job, I was pretty overwhelmed. I answered basically by saying "I honestly don't know where I see my self in the next 6 months with everything going on. I am just focused on graduating and landing a job". I didn't end up getting that job, so I am not sure if they liked my answer, but it was truthful, so I don't feel bad about it.

Its kind of a BS question. But I think it holds water, at least a little. You went through a lot of school -- why? What career path were you trying to open that would not otherwise have been available, and how do you see yourself progressing in it? Truly not knowing is a sign that you haven't looked into the career and progression much. It means the hiring manager has less of a signal on if you'll be happy at the company in 1, 3, 5 years, because you can't tell them what you are looking for.

I was asked a similar question when I was applying for Medical School, and told my non-answer was BS. It frustrated me at the time. 10 years after quitting medicine, I can't say that person's viewpoint was wrong.

>PS: It’s odd, but I suspect a random process may actually work better because the great people actively looking for jobs are likely to be bad at getting them.

Somebody should almost build a toy startup around that random process eg Chatroulette for hiring. that might make sense.

However you would either need a great process for quickly terminating bad hires or keep this strategy very secret. Otherwise word spreads and your hiring pool quickly turns from "random" to "every freeloader in the industry".

Yeah, you would probably add some weighted sampling to it eg overweight those candidates who have lots of Github contributions.

'open-source contributions' might be what job postings are trying to capture with this ... presumably the longer a good developer is unemployed the more likely they are to contribute to open source

> because the great people actively looking for jobs are likely to be bad at getting them.

Why do you estimate it as "likely" (i.e. more probable than being average/good at getting the jobs)?

I assume they are mostly unrelated skills, my mental assumptions work something like:

Suppose the top 10% of developers break down: 1/3 are noticeably worse than average and take 6 months to find a job, 1/3 are average and take 3 months to find a job, and 1/3 are above average and take 1 month to find a job. Each month 3000 people or 100 good 100 average and 100 bad start looking for work and the number of people looking for a dev job is 10,000.

Our of that 10,000: 100 are great developers skilled at finding jobs so 1%. 300 are great developers that are average at finding jobs so 3%. And 600 are great developers that are bad at finding jobs so 6%. That’s still 10% of applicants but most of them don’t look as hot in interviews.

Interesting! I've never before considered the dimension of time, which will lead to accumulation of brilliant interview flunkers.

The only assumption that remains to be tested then is that our current practices do not identify developers that belong to, say, top 10%.

Yes! You aren't the first to make that observation.

Compare https://sockpuppet.org/blog/2015/03/06/the-hiring-post/

>> great at talking about themselves but >> were at best mediocre otherwise

Happens all the time. There's simply no way to judge intrinsic motivation one may or may not have for the job, and without that even a smart person will do the bare minimum to not get fired. Which is not much at all, at most places.

There are more reasons than that to perform good, I guess that's what the classic "where do you see yourself in 5 years" is trying to judge. But in general this is a problem, yes.

More importantly, the fact that FizzBuzz is used to the point of being a cliche is already admitting that given just a CV and a verbal interview we are unable to reliable judge if a person can even program at all.

I like how it's always that the hire is to blame for their skill level in the organization once they are hired. How come we don't ask whether companies know how to hire? Like, publicizing firing rates, turnover, that kind of thing.

I mean, I know why, but it's never a part of the rubric. "We stopped using talk interviews and our retention went way up. We used to see 1/3 of our hires quit within two months of their first salary review." This is probably a more reliable approach the more a company promotes from within, and my sense is that many-if-not-most startups do not.

Reply on news.ycombinator.com