First: software is nearly always a dicey proposition. There are so many variables that it's just impossible to predict much about it with any regularity. The best approaches that I've seen so far accept this, and say, "Trial and error? So be it: we will do trial and error as best as it can possibly be done." YC and the lean startup model (aim low, accept that most things will fail) are good examples.
The other approach is the "tried, no error" method, which cares intensely about the past. Hiring a web developer? Portfolio first, examples of use of specific technologies, examples of specific features. Find no one who meets your exact qualifications, and proceed to either a)raise salaries a lot, and constantly (banks, hedge funds), or b) complain to Congress about a "talent shortage" (everyone else).
I think YC is the better model, and is a sort of company-less version of michaelochurch's "open allocation." The key insight is: if you want 10x engineers to ride up on their white horses and save you, you can't presume to tell them what to do. They're cars, not faster horses. Beating a car with a whip won't make it go faster---and this is a feature, not a bug.
To get the best out of developers, you have to accept that they bring risk, and manage around that, rather than blindly trying to reduce it. That's why I think Apple is really fucking smart to have a huge cash cushion like they do: with that cushion., it frees you to go for long shots. Apple is nearly riskless in its financials, and so can afford to be very risky in its product development process.
If I were starting a software company, and knew exactly what I wanted to build, I'd hire people with proven track records, pay them a bazillion dollars each, and give them zero autonomy. Or, I'd hire a bunch of college dropouts who've obssessed over some kernel of good ideas (Tufte, Hickey, FRP, ML), pay them nothing, and let them run wild (this is basically YC's business model).
If you're really serious about this---and I think it's worth it to be---I'd read everything on michaelochurch's blog. He exaggerates, has a chip on his shoulder, uses weird analogies, and has a better understanding of these issues than anyone I can think of.