When you pitch your idea to a VC and he is expressly concerned with numbers:
1) How do you best determine projected revenues?
2) Do most VC's take into consideration projected revenues or do they view them as "fluff" numbers with no ground to stand on?
3) How do you prove that your projections are accurate?
4) Do you give accurate numbers or conservative (so they are more believable)?
I agree with Rand. As part of our discussion with investors, we show the projections we had shown investors in prior funding rounds -- and how we mapped to those projections (we almost always beat them). This makes the projections (and you) much more credible.
If you don't have the revenue history yet, then you need to have realistic projections with the expectation that the VCs are going to discount them. The reason they discount is very simple -- on average, most startups don't hit their target revenue numbers so they're factor that risk in.
This is going to be a fairly personal perspective (and thus may not be directly applicable to situations that aren't very similar), but my feeling is that VCs really trust projections when you've shown them the previous projections you've made over a number of months/years against the actuals. For example, in our recent pitch process, we showed 2009's estimated revenues/expenses/etc. alongside our actuals and this certainly helped to show that we're serious when we say we A) always estimate conservatively and B) always beat projections.
When you can show things like visits, conversion rates, retention, recitivism, etc. in a systematic way over time, then make projections that estimate slowing down your growth (i.e. "We're not even going to improve next year as much as we did this year") that makes for, IMO, a compelling, believable story.