Estimating confidence in the completion date for an IT project can be hard. When I've been asked to review a troubled IT project, I've found that a very useful technique to determine how far along the project actually is -- and when it's actually likely to be done -- is to ask those involved at different levels what their "$1,000 confidence level" is -- that is, what date they would be willing to bet $1,000 out of their own pocket that the project will actually be done by. I find that even the most optimistic engineer suddenly turns cautious and starts thinking of all the things that could go wrong.
I then ask what their $10,000 confidence level is. Those answers tend to be the most accurate.