Earlier quoted context omitted.
No, asking questions in a face-to-face interview is not the best way to figure out if a person is going to do what you want them to. It is not a very effective way of doing that at all. To be meaningful at all, it requires an interviewer with many of the same skills as a psychologist, and a candidate who can represent themselves effectively in an extraordinarily hostile situation. Even if you have both, you're still…
How long does a work-sample test take? I'm typically hiring people to work on on-going projects over a period of months or years; the fact that candidate A does better on a 2-hour or 2-day task than candidate B may not give me that much information about things like how effective they will be at sharing their own knowledge with the rest of the team (or taking guidance from more senior members of the team), developing…
Work sample testing has been extremely powerful for us. In particular: it has enabled us to screen candidates for aptitude instead of experience, and the result has been a series of hires with no prior experience in our (very specialized) field who have completely blown us the fuck away, like the former line-of-biz insurance company .NET guy who, a month or so in, looked at Rails for the first time and found a serious security hole (which now has a CVE), or the other line-of-biz .NET guy who now runs our crypto review board and co-wrote our crypto challenges, or the math major we hired out of college who ripped through the MSP430 CTF in a couple days (and also has a Rails CVE).
We picked up a guy who had done a couple months of Android development after college, and then moved to what I think (from his resume) was a devops job --- I never asked him in the whole course of our hiring process what that job was about, so I don't know. Six months later, he implemented a CRI elliptic curve DSA exploit that involved BKZ lattice reduction and an FFT filter step. The lattice reduction steps on that attack took something like SIX HOURS to run, and when they finished, you still had to get the FFT step working to figure out if you got the BKZ step right.† Months later and I'm blown away just thinking about how he could have gotten that working. In case you're wondering, he kicks ass on the more boring parts of our work too.
Some other things to know about that guy:
* No prior crypto experience
* No prior software security experience
* Less than 2.5 years dev experience across both jobs
* From flyover country
* Physically uncomfortable during the interview (something noted by interviewers on his feedback)
How compelling do you think his "greatest achievement so far" story would have been? We weren't dumb enough to ask, so we'll have to guess.
Before we did work-sample testing, overcommunication about hiring process, demoted the importance of phone screens, and standardized face-to-face testing, we never would have hired these people. Any process that filters outs Seans, Alexes, Peters, and Ryans over things like "passion" or "what is your greatest achievement" is a stupid fucking process. But there it is, enshrined as the default process for our whole industry. And here's a thread full of people arguing on its behalf.
Yes, I think someone is going to make a lot of money correcting that market inefficiency.
† If you're wondering, number of other people on our team who could have helped him debug lattice reduction C++ code: zero.