Agile teams estimate in Fibonacci story points — 1, 2, 3, 5, 8, 13, 21, 34 — because uncertainty grows with size, and the widening gaps stop teams pretending otherwise.
Nobody can reliably tell a 7 from an 8. The Fibonacci gaps force a real choice between clearly different sizes, which makes estimates faster and more honest.
A large card is a signal, not a number: 21 and 34 usually mean the story should be split before it enters a sprint.
The sequence is not chosen for its mathematics. It is chosen because the spacing matches how confidence actually decays.
Between 1 and 2 you are choosing between two things you understand well, and the difference is real. Between 21 and 34 you are choosing between two things you barely understand, and a finer scale would only invite false precision. The widening gaps encode that: the bigger the work, the coarser the honest answer.
This is also why a team that argues over 8 versus 13 for ten minutes is usually arguing about scope, not size. On a Fibonacci deck adjacent cards are far apart on purpose, so a genuine split between them means the two voters are picturing different work.
1, 2, 3, 5, 8, 13, 21, 34, and a ? card.
The ? is not a zero and not a joke — it means "I cannot size this from what I have been told", which is a legitimate and useful answer. A table of question marks says the item needs work before it needs an estimate, and Point Poker deliberately does not treat a unanimous ? as consensus.
Plenty of teams use a modified sequence instead — 0, ½, 1, 2, 3, 5, 8, 13, 20, 40, 100 is the common variant, rounding the large end for readability. Either works. What matters is that the team uses one deck consistently, because story points only mean anything relative to the team's own past estimates.
The sequence does not produce good estimates on its own. These habits do.
Most teams use 1, 2, 3, 5, 8, 13, 21 and 34, where each number is the sum of the two before it. Many use a modified version — 0, ½, 1, 2, 3, 5, 8, 13, 20, 40, 100 — which rounds the large end. Point Poker opens with 1 to 34 plus a ? card.
It means the team thinks the item is large and is less confident about it than about anything smaller. In most teams a 13 is the last size still worth putting in a sprint, and a 21 or above is a prompt to split the story rather than an estimate to plan around.
Only if you have real work that small. A 0 is useful for an item that is genuinely free because it rides along with another story, and a ½ for trivial changes. Both tend to cause more debate than they save, which is why the default deck here starts at 1.
Ask both sides what they are picturing — a two-card split almost always means someone is including work the other person is not. Then re-vote and record whichever card the team lands on. Do not record a 4: it is not on the deck, and the moment your history contains numbers the deck cannot produce, comparing sprints stops working.
The deck is the template — there is nothing to download or configure. Open a room and the Fibonacci sequence is already the active deck, with T-shirt sizes and powers of 2 available if the team prefers those. If you want the results as a file afterwards, each session exports to CSV.