Planning poker: complete guide
Planning poker is the most widely used estimation technique in Scrum and agile. Here is everything you need to know in 5 minutes.
What is planning poker?
Planning poker (also called Scrum poker) is a collaborative estimation technique where each team member secretly chooses a card representing their estimate of a user story or task. All cards are revealed simultaneously. Differences trigger a short discussion, then the team revotes until convergence.
The goal is not precision but discussion: uncovering hidden assumptions, unidentified risks and technical disagreements.
The rules in 5 steps
- The Product Owner presents a user story or task.
- The team asks clarifying questions.
- Each member secretly chooses a card (without influencing others).
- All cards are revealed simultaneously.
- If estimates converge: keep the value. Otherwise: discuss the extremes, then revote.
How long does a session take?
Budget 2 to 5 minutes per user story: a minute of presentation, one round of voting, and discussion only when the cards disagree. A one-hour refinement session therefore covers 12 to 20 stories, which is more than enough to feed a two-week sprint.
Past 90 minutes, estimate quality drops: fatigue pushes people to vote like their neighbour just to get it over with. Two 45-minute sessions beat one two-hour session. If a single story eats more than ten minutes, the problem is no longer estimation but framing: send it back to the backlog to be split.
Why the Fibonacci sequence?
The sequence 1, 2, 3, 5, 8, 13, 21, ? is the most common because it models growing uncertainty: the larger a task, the less precisely it can be estimated. The gap between values grows with size, which forces the team to take a clear position without pretending to illusory precision.
The ? card means "I can't estimate this task", a useful signal to detect a lack of clarity in the specification.
Comparison of card sequences
| Sequence | Values | Best for |
|---|---|---|
| Fibonacci | 1, 2, 3, 5, 8, 13, 21, ? | Experienced Scrum teams, story point estimation |
| T-shirt | XS, S, M, L, XL, XXL, ? | Rough estimates, non-technical teams, backlog sorting |
| Powers of 2 | 1, 2, 4, 8, 16, 32, 64, ? | Technical tasks with predictable doubling of complexity |
Planning poker for remote teams
In a room, the secrecy of the vote comes from the gesture: everyone lays a card face down. Remotely, you need a tool that keeps votes hidden until the reveal, otherwise the first number spoken on the call anchors the whole team. That is the one indispensable job of an online planning poker tool, and it requires no account, no install and no integration.
Three things worth checking before picking one: do votes stay masked until the reveal, does a participant who drops off get back into the session, and what happens to session content once the meeting ends. On that last point, the only answer that commits to nothing is that there is nothing to keep.
Asynchronous planning poker
Letting everyone vote on their own time, over two days, is tempting when time zones barely overlap. You get a number, but you lose what makes the technique worth doing: the discussion the gaps trigger. An estimate nobody argued about has surfaced no risk.
The compromise that works: one asynchronous round to sort, then a short synchronous session for the stories where cards disagree. In practice that is a third of the backlog, and the meeting drops to twenty minutes.
Four mistakes that skew estimates
- The most senior person votes first. The moment a number is said out loud before the reveal, everyone else lines up behind it. Simultaneous voting is not stagecraft, it is the entire mechanism.
- Converting points into days. "One point equals half a day" cancels the benefit of a relative scale and turns the estimate into a delivery commitment. Points compare stories to each other, they do not fill a calendar.
- Forcing consensus. A persistent disagreement is information: it signals a story that is misunderstood or badly split. Averaging it away removes the signal without fixing the problem.
- Estimating without the person who will do the work. An estimate produced by a subgroup commits people who were not there. It is the most common reason planned velocity and actual velocity drift apart.
Frequently asked questions
What is the difference between story points and days?
Story points measure relative effort, not time: the volume of work, the uncertainty and the complexity taken together. The scale is relative - a 5 is clearly bigger than a 2, without claiming an exact multiple of it. This abstraction lets teams compare tasks without getting blocked on individual durations, which vary by person and context.
How to handle estimation disagreements?
When estimates diverge (e.g., a 2 and a 13 in the same hand), ask the people who voted at the extremes to explain their choice. The discussion often reveals different assumptions or unidentified risks. Vote again until convergence, without forcing consensus.
Does planning poker work for remote teams?
Yes, that is precisely why CleanPoker exists. Create a session, share the URL, each participant votes on their own device. Results are revealed in real time via WebSocket, with no setup or account required.
What is the maximum number of participants?
CleanPoker has no technical limit. In practice, beyond 10 participants, discussions become difficult to manage. For large teams, prefer parallel sessions or sub-group estimation.