Affinity estimation

The point of affinity estimation isn't that it's silent or fast. It's that you sort first and commit to numbers last.

Every other relative-effort technique asks for a number per item, then waits for the team to agree on it. Affinity flips the order: arrange the stack into "this is bigger than that" piles, no numbers, then label the piles at the end. The argument that would have happened over each story — "is this a 3 or a 5?" — never starts, because by the time the numbers arrive the relative order is already settled and the buckets are obvious.

That order — relative first, absolute last — is what makes it fast. It also defines where the technique works and where it doesn't.

Where the order pays off

First-pass backlog sizing, where rough relative order is enough; pre-refinement before stories reach Planning Poker, putting everyone in the same ballpark; and re-anchoring exercises against already-shipped stories. In each case the team needs a coarse-grained ordering more than a calibrated number, and the sort produces both.

Where the order breaks

Sprint-level estimation. The reveal, the disagreement, the question that surfaces hidden scope — that's the whole point of Planning Poker, and affinity estimation skips it deliberately. For the stories the team is about to commit to, the silence is the failure mode, not the feature.

Affinity vs magic estimation

Both flip the same order. Affinity sorts by relative size against the wall and labels the buckets at the end. Magic estimation sorts into named buckets from the start. Affinity is more flexible; magic is more decisive. The choice is between argue-about-the-line and argue-about-the-bucket — pick the argument you'd rather your team have.

If you reach for affinity, commit to the order. Don't let someone shout a number across the room during the sort.

See magic estimation for the cousin technique; other estimation techniques for how both fit.