And yet in the politics of a company, an estimate becomes fact, and then if you don't meet it they are angry.
- Pad.
- Communicate risks.
- Communicate assumptions.
Then when assumptions are wrong, immediately communicate that this will change the estimate. When risks hit, communicate that a risk hit, and whether or not you can absorb the impact in the planning. If not, communicate the expected delay. Pad that too.
Communicate, communicate, pad, and communicate.
(Padding, BTW, may feel dishonest, but it isn't. It is making the assumption that a certain share of you risks will strike and making allowances for that.)
If you still get in trouble for missing estimates, pad more and communicate in writing. Then when people start blaming, you can point to the emails sent. To their/your boss if necessary.
This. It helps to realize that when someone asks "how long will X take". What they are actually asking is "When can I rely on X being done by?". If you say 2 days and it takes one, no problem. But if you say one and it takes 2, they've already told the client it'll be done tomorrow, arranged for the machinery to be in place, and now everything is jammed up based on your delay.
Estimating is as much art as science and requires understanding the context around the estimate and how it will be used.
This stuff is fluid but too many technical people feel as if it's not.
Estimate, multiply by 4, give the number.