Sounds like a reasonable technique to me. Assumes 3D mental imaging skills that some people might not have, and which probably aren't applicable to most DBA/dev tasks, but that itself as a barrier would be easy enough to false-negative out of the test.
Just to be clear, I'm looking for a successful completion of the task or even how close they get. I'm looking at how they aproach it it. To be honest, I always thought it was impossible to complete in the timescale although one guy did get awfully close once. He completed a model but a couple of the bricks were miss-orientated. Damn close though. I put him forward but the budget got cut back just before we made an offer. I think that sucked and I hope he found something else decent.
20min seems long to me
I'd have thought so too if I hadn't tried it out on one of my team members first with no duration to see how long was reasonable. It took him over an hour but he did end up getting it exactly right. At the 10 minute stage most people are still establishing which bricks have actually been used. 20 is actually surprisingly tight.
Eventually I tried something that made him stop. He told me they'd had this problem and couldn't solve it, going through almost everything I'd done until that last thing. Was an interesting experience
That does sound like you got fleeced for some free consultancy.:-D By the way, I've got a position available for a particularly talented c# programmer with a deep understanding of the data binding model. It's imperitive that you know how to implement the iBindingList interface to give a source that will remember the persistance state of it's members, allow filtering and sorting and handle any generic type. This question will be purely to test your knowledge at interview and is in no way connected to the fact that I've been banging away at this for weeks now! There's definitely be a job at the end of it. Honest. You can trust me.;-)
The first thing I'd do is ask for specs.
That would definitely win you points as would asking for a picture. An insistence on ALWAYS having a spec (not sure you're saying you always need one though) would lose you points. To me a good developer needs to be able to investigate a problem themselves when the situation demands it. As often as not the customer genuinely cannot spec what they want because they don't fully understand the problem themselves and on those occassions I want someone with a much more rounded skill set than just the ability to follow a set of instructions. Any developer who ever tells me they don't feel they should ever have to speak to a user gets counted out immediately. Specs when possible, flexibilty when not.