Manager: We (meaning you) need to do task A. How long will it take?
Me: Task A will take X days to do.
Manager: That seems awful long.
Me: How long do you think it should take?
Manager: It surely could not take any longer than Y days.
Me: Ok, it seems you have an answer to your question then.
Later:
Manager: It’s been Y days, why isn’t task A done yet?
The way we estimate on my team is to break tasks down into related subtasks that will take one day or less, then add up all the subtasks. It’s worked pretty well.
That pattern is also recommended in the book. Break down estimates into chunks of 5 hours or less.
With lots of smaller tasks, estimates tend to both be more accurate due to smaller scope, and some of the over/under inaccuracies will cancel out.