Planning Poker
The estimate is the smaller half of what this produces. The valuable half is finding out, before the sprint, that two people were picturing different work.
How a round works
- Read the story together. Whoever brought it explains what it is and who wants it. Questions now, not after the cards are out.
- Everyone picks privately. No thinking out loud, no "this feels like a five".
- Reveal at once. This is the mechanism — see below.
- Ask the outliers first. The highest and the lowest explain what they were seeing, in that order, before anyone else speaks.
- Re-vote if the discussion changed anything. Otherwise take the number and move on.
Why the simultaneous reveal matters
If estimates arrive one at a time, the first number becomes the anchor and the rest cluster around it. If the tech lead says eight first, it is an eight. Everyone reveals at once so that eight independent judgements exist before any of them influence the others.
The same logic explains why estimation should not be anonymous, even though retrospective cards often should be. The whole mechanism depends on being able to ask the person who played the 13 what they were seeing. Hide the names and you keep the disagreement but throw away the only thing that resolves it.
Choosing a deck
| Deck | Cards | Best for |
|---|---|---|
| Fibonacci | 0 1 2 3 5 8 13 21 34 ? ☕ | Almost every team. Start here unless you have a reason not to. |
| Modified Fibonacci | 0 ½ 1 2 3 5 8 13 20 40 100 ? ☕ | Teams that size epics as well as stories, where 21-versus-20 is false precision. |
| T-shirt sizes | XS S M L XL XXL ? ☕ | A first pass over a new epic, or a team whose points keep being read as hours. |
| Powers of two | 0 1 2 4 8 16 32 64 ? ☕ | Teams who forecast with capacity maths rather than velocity. |
| Linear 1–10 | 1 2 3 4 5 6 7 8 9 10 ? ☕ | Voting on confidence, risk or complexity — not for story points. |
The gaps widen deliberately. You can distinguish a 2 from a 3; nobody can honestly distinguish a 21 from a 22. A deck that offers that precision invites a conversation that cannot produce a real answer.
Two cards do work that numbers cannot. ? means "I do not understand this well enough to size it" — always worth more discussion than a guess. ☕ means "I need a break", and a team that never plays it is probably estimating badly by hour two.
Reading the spread
- Everyone agrees. Commit it and move on. A room that talks through an agreement it already has is how a two-hour session becomes three.
- One card apart. Ordinary variance between people who agree. Take the higher card. Do not spend ten minutes closing a gap that will not change the sprint.
- Several cards apart. Stop. This is the signal you came for. Two people are imagining different work — different scope, a dependency one of them knows about, a definition of "done" that was never written down. Whichever is right, the story is missing a line, and adding that line is the actual output.
Keeping the session moving
- Timebox the discussion, not the estimate. Two minutes per outlier is plenty.
- Park what is not ready. A story nobody can size is not an estimation problem, it is a refinement problem. See the Definition of Ready.
- Split anything too big. A story that cannot finish inside one sprint has no honest estimate — the number will be a guess at an epic, and the team gets held to it anyway.
- Write questions down as they come up. Anything the room could not answer belongs on a list with a name against it, not in someone's memory.
What not to do
- Do not convert points to hours. The moment they map to time they stop being relative and start being a commitment, and estimates inflate to protect people.
- Do not average the cards. "It came to 6.4" throws away the disagreement, which was the useful part.
- Do not compare velocity between teams. Points are calibrated within a team and mean nothing outside it.
- Do not let one person estimate for the room. If the answer was already known, the meeting was not needed.
Common questions
What is Planning Poker?
Everyone privately picks a card for the relative size of a story, then all reveal at once. The simultaneous reveal stops the most senior voice anchoring everyone else.
Why use Fibonacci numbers for story points?
The widening gaps match how estimation accuracy behaves. You can tell a 2 from a 3, but nobody can honestly tell a 21 from a 22.
What does it mean if estimates are far apart?
Usually that two people are picturing different work, which means the story is missing a line. That discovery is worth more than the number.
Should story points map to hours?
No. Points measure size relative to other stories; velocity turns that into a forecast.
Next: Definition of Ready — the checklist that decides whether a story can honestly be sized at all.
Run it, rather than read about it. Free, and nobody needs an account — share a six-character code and start.
Start a session