Three amigos
Three people, one story, before it is estimated: the product owner who knows what it is for, the developer who knows what it will cost, and the quality engineer who knows how it will be judged.
Why exactly these three
Section titled “Why exactly these three”Because they fail in different directions. Left alone, the product owner specifies an outcome without a mechanism, the developer builds a mechanism without checking the outcome, and the quality engineer discovers the gap between them after both have moved on.
Each of the three asks a question the other two do not naturally ask:
- What is this for? — the business intent, and what happens if it is missing.
- What does this cost? — the implementation, its edge cases, what it touches.
- How would we know it works? — the observable behaviour, including the unhappy paths nobody wrote down.
How it runs
Section titled “How it runs”Fifteen to thirty minutes per story. Someone reads the story aloud. The other two interrogate it. The output is not a document — it is a story that has either become clear or become visibly ambiguous.
An ambiguous story does not get estimated. It gets an open question and an owner.
The output
Section titled “The output”Either the story is ready for example mapping and then ATDD, or it has a list of questions that must be answered first. Both outcomes are successes; only “we discussed it and everyone nodded” is a failure.
Common failure modes
Section titled “Common failure modes”- Four amigos, then six. Once it is a meeting with a distribution list, it stops being a conversation. Add a domain expert when the domain is the uncertainty — and only then.
- The developer presenting. If one person explains and two listen, it is a handover with extra steps.
- Running it after estimation. The purpose is to make the estimate meaningful, which is why it belongs upstream of grooming. Afterwards there is nothing left to change.