On the other hand, when you (metaphorically) want the same consultants to go beyond "twoallbeefpattiesspecialsaucelettucecheesepicklesonionsonasesameseedbun", like a McBurger flipper trying to make basil aioli, it usually doesn't turn out well.
Where they fail (overrun budget and time) is for more complex stuff. Software complexity requires exponentially better programmers to cope with it. Each time you're building a complex piece of software, it's like you're having to invent the burger again and explain it to a computer.
Couple that with the fact that each company will want their particular meal tailored very specifically to their tastes, so that there's little reuse of recipes possible, and it's obvious why a lot of projects fail when the burger flippers are asked to design a new burger.
Every time you go cheap, or with the "big consultants" that joel describes the expected outcome is massive time overruns and massive budget overruns.
The closest thing to the big mac model I can think of is installation of existing software, with customization for the environment, but even there you get the 3 year long SAP projects that cost tons.
What prevents software from being put into a manual? I'm hoping the answer is something beyond "requirements don't match up exactly to previous problems... ever", but I fear that's all it is.