I can't take credit for this tip. Robert Hoekman (rhjr.net) inspired me to really think about whether or not we are gaining truly valuable information by spending days, weeks, months studying "real users". Thinking about attempts at studying users I have done before made me realize that, most of the time, the assumptions we made without users were largely correct. And while it is valuable to have your assumptions validated, it isn't always necessary before starting development. Notice that I say "most" of the time. Well, the reality is, get ready... we don't always get it right, even when we do extensive user testing.
Why? Well when you are testing users, regardless of how "real world" you try to make it, the user always knows that they are being tested. Many people are self-conscious having someone (or two people usually) looking over their shoulder. There are many other reasons as well.
Now before all the UCD practicioners stand up and start disagreeing with me, let me make one last point that is important. In order for this approach to work you have to make realistic assumptions in which you are highly confident. So how do you make confident assumptions? That is part of Tip #2. Stay tuned.
If ignoring your users sounds like an approach you are interested in, check out http://rhjr.net/theblog/2007/06/06/why-we-should-ignore-users-the-podcast/