The point is that you'll only get applicants up to a certain level. Once an applicant has enough skill and knows it, and also values his time, he'll take one look at your 14 hour challenge and laugh.
So in the end you're only hurting yourself because you'll mostly get naive and easily manipulated people who are fine for the lower ranks but absolute poison once they move up. That's if they don't wise up and leave first, costing you time and effort to find and train replacements and scramble to rediscover and distribute the lost knowledge they left with.
But then again, such a long duration challenge is already very telling of the company culture. If you can't respect other peoples' time, then your company is already on shaky footing.
he'll take one look at your 14 hour challenge and laugh
These things shouldn't be designed to take 14 hours. They should be designed to take one, maybe two or three . . . if you know what you're doing.
One of the other posters took so long because he didn't know what he was doing. Well, perhaps that is an uncharitable way to say it, but you know what I mean. :)
The test I describe should take an hour or two - it's a test of "can you actually do these things your CV claims you can?" Which is nontrivial, but I enjoyed taking it (fix a broken thing!) and feel fine about giving it to others.
>"can you actually do these things your CV claims you can?"
Yeah but how many questions must you ask before you are certain? Otherwise you're rolling the dice that the candidate has simply seen that problem before but has terrible habits otherwise.
Reduce the time required and you risk more false positives. Increase the time required, and you risk the skilled ignoring you.
But the process described in the post above is clearly about proficiency, so a challenge is not supposed to last 14 hours. ;)
And I admit you have a point. The thing is, people really have well-looking CVs and they can't code BFS in Python. Probably balancing of 'technical grilling' is a hard to master skill in and of itself.
