No Estimates: what I answer when asked how long
An estimate commits nobody. A shippable slice commits the person shipping it.
The question always turns up at the same moment, near the end of the first conversation: "so how long does it take?". The honest answer is that I do not know, and that the person who answers with a number does not know either. Which does not mean you get to refuse to answer.
The question behind the question
Nobody asks for a duration out of curiosity. People ask because they need to know whether they can commit a budget, warn a customer, hold a quarter. The real question is: from when can I count on it?
That is a much better question, and it has an answer.
What an estimate actually costs
An estimate commits nobody. It is made before anyone knows anything, it is forgotten the moment it is missed, and it turns the first piece of bad news into a negotiation instead of a piece of information. The time spent producing it is time not spent reducing the uncertainty it claims to measure.
Worse: it moves the conversation. Once a number is on the table, people discuss the number. They stop discussing what needs to be built.
What I offer instead
A shippable slice. The smallest one that is genuinely useful to somebody, in production, with a date.
The difference is not cosmetic. On the strong authentication of a neobank, the estimate phrasing gives you: "strong authentication, six weeks". The slice phrasing gives you: "the second factor, on one journey, in production, Friday". The first is a promise. The second is checkable on Friday.
After three slices you no longer need to estimate: you have an observed pace. An observed pace projects, and it projects far better than a hunch formed before the project started.
The budget objection
"We cannot commit spending without knowing where it stops." That is fair, and it is not an argument against the method. A budget is capped in euros and in weeks, independently of what anyone believes can be fitted inside it. You know what you spend at most, and you find out as you go what that buys, with the option to stop at every slice.
That is the exact reverse of an estimate, where you believe you know what you get and you find out what you spend.
The corollary
This only works if bad news travels the day it is known. A slice that slips is announced immediately, not at the next review. Otherwise you have replaced one wrong estimate with a series of wrong slices, and you have lost on the trade.
That is the uncomfortable part, and it is also the only part that makes the rest credible.