[timrobinson]

+1 I use the same approach: repeatedly split the task into smaller tasks until you've got a list of tasks all the same size. I don't think I've estimated down to less than half a day, however: I've never been comfortable coming up with fixed hours-per-day number for the team. (As long as you can get enough work done in a 24-hour period, and I can get hold of you when I'm in the office, I don't care when you do it.)

However I'm not entirely comfortable with applying fudge factors like "multiply all estimates by four". I'd rather try to gauge the confidence that we have in our estimate: either by planning poker, or by coming up with separate worst case, best case and expected estimates.