I take a not-stupid approach to goals and predictions for big projects.

They don’t need to be good.  There are too many variables to reasonably expect good road maps.  If you think it’s good, you might have too many blind-spots.

But it can’t be bad.  If it’s bad, you’re nowhere near ready.  It doesn’t have to be good, but it has to be not-stupid.  If you can get it to not stupid, you can proceed and adjust as more information is revealed.