[viraptor]

I think it can work both ways. Tasks like that can be time-consuming, but would you rather take a day off for an on-site interview, or spend ~1h doing it at home? There's some line that shouldn't be crossed of course - I think companies shouldn't give the test to everyone - just those who are actually considered after some screening. But I've been on both sides: taking and preparing the test, and I think it's a good idea as a 2nd stage filter.

My test was completely open-ended. Trivial task (script to copy data from .csv to a sqlite, optionally notify on the queue at the end), but make it as production-ready as you can/want. I don't think I could learn as much about people's abilities during an interview, as I could from that task. I mean, the 1 person who made a proper package, a short .rst install & use documentation, and included crazy unicode test cases was immediately on the top of the list.

[pmiller2]

Those take home tasks are rarely 1 hour deals, so it's not a question of spend an hour at home or take a day off for an on-site. It's usually spend X hours at home and take a day off for an on-site.

Do you really think that "crazy unicode test case" person really only spent an hour on your test?

[viraptor]

> Those take home tasks are rarely 1 hour deals

This seems to be the common complaint, but really this is discussing the idea -vs- the implementation. Some examples here of checking specific things that were not in scope, or tasks that take whole day are really disappointing. And I agree there are loads of really crappy tests out there.

But yes, you can create a test that's easily doable in 1h. As for the crazy edge case tests: probably. It's still a reasonable thing to check for, but often people don't. (having a non-ascii name helps here) If you have simple enough and open-ended enough tasks, I don't think extra time helps the candidates in any way. I mean, either you have experience doing something or you don't. Even if you spent, for example, X hours researching "what does production-ready mean?" and related topics, I don't believe you'd produce as good result as someone with actual experience. (yeah, I could see that in the answer)

[pmiller2]

I will concede the point that if it literally takes an hour, and it means I don't have to write code in the interview with someone watching my every move at all, then, yes, I would gladly do such a test. I'm still skeptical that a test that requires writing production-ready code and tests would really take only 1h.

[dpark]

So you prioritized for people willing to sink many hours into your assignment. The need to sink many hours into a homework assignment is the criticism.

[viraptor]

Phone interview - 1h during work hours. On-site interview - ~3h total, but probably a whole day off work. Test - ~1h at any time you choose.

I'm not sure I follow your logic. They would spend just as many hours, at a more inconvenient time of day without the test. I did the task myself to a level I'd expect at the time to make sure it does not take "many hours". If that simple task would take people many hours, then I hope they didn't apply... (or they would probably fail very soon if hired at all)

[throw_away_777]

The candidate gets no benefit from doing your take home test. In an interview the candidate learns something about the company. Also the ~1h requirement just isn't realistic if you give an open-ended test - you penalize the people who actually spend 1h and reward those who spend more time.

[dpark]

As the other commenter pointed out, you don't control how many hours a candidate spends on the assignment. All you see is the end result, and the guy who spends 8 hours will generally end up with a more impressive result than the candidate who spends 1 hour, and this reflects time spent far more than candidate quality. If you give an "open ended" assignment, you are encouraging candidates to spend a ton of time and then lie and claim they spent very little time, because that looks impressive.

Do you really think "the 1 person who made a proper package, a short .rst install & use documentation, and included crazy unicode test cases" spent only an hour?