The problem with startup hiring isn't that you need to think harder about it, it's that it's an inherently low-information situation. The factors that make a good employee often depend as much on what the startup is doing and how that squares with the employee's goals rather than any innate personnel factors.
For example, I'm leaving my day job. My reasons for this are complex and I'm not going to air all of them in a public forum, but they boil down to "It's not a good fit for me anymore." Thing is, there was no possible way we could know that when I was hired, because I was a good fit back then. My job description now is totally different than it was back then - in fact, when I was working with a coworker to come up with a job advertisement for my position, I realized that I would not have applied for my position as it currently exists. I've also grown as a developer - when I took this job, I thought Java was a decent-if-not-great language, I thought dynamic languages were unsuitable for real work, I had only passing familiarity with Haskell, I didn't really know JavaScript (though I listed it on my resume), and I basically didn't know what I wanted out of my career. I even mentioned during my interview that "Yeah, I don't know if this is something I really want to do, but I can't know that until I give it a try - after all, when I took my last programming job, I didn't know if programming was something I wanted to do, and that worked out great." Come to think of it, I still don't know what I want out of my career, I just think that this particular startup I'm working on is more likely to be it than my current job.
In my experience, the best indicator of job success isn't raw programming talent, it's how much the programmer believes in the product. That's why cofounders usually need to be friends before they can work together. Oftentimes, you'll have to take a leap of faith that you're on the right track, and you need to trust your cofounder to do that. I don't see how you can test for that with a third-party service. You can take the absolute best programmer in the world, stick him in a project he thinks is boring, and he'll suck at it.