Planning Poker

What is Planning Poker?

Planning poker is a consensus-based estimation game in which team members privately pick a card for a backlog item's size, reveal together and discuss the gaps. Revealing at once keeps one voice from anchoring the group.

Updated November 2026 · 3 min read


Planning poker is a team estimation technique in which each person privately chooses a card showing their estimate for a backlog item, everyone reveals their cards at the same time, and the team discusses any differences before voting again. The simultaneous reveal stops the first or most senior voice from anchoring everyone else, and the discussion of outliers surfaces assumptions and risks the team would otherwise miss.

Planning poker was described by James Grenning in 2002 and popularized by Mike Cohn in his book Agile Estimating and Planning (2005). It is a lightweight variant of Wideband Delphi, a consensus-based estimation method associated with Barry Boehm. Teams usually play with cards showing a modified Fibonacci sequence and estimate in story points, though the technique works with any scale.

What you need to play planning poker

  • The estimators. The people who will do the work, typically developers and testers. The product owner explains items but usually does not vote.
  • A deck per person. Common values are 0, ½, 1, 2, 3, 5, 8, 13, 20, 40 and 100, plus a "?" card for "I cannot estimate this" and often a coffee-cup card for "I need a break."
  • Refined backlog items. Stories with enough description and acceptance criteria to discuss.
  • Reference stories. One or two previously completed items the team agrees are, say, a 2 and a 5.

How to run a planning poker session

  1. Present the item. The product owner reads the story and acceptance criteria.
  2. Ask questions. Estimators clarify scope, edge cases and dependencies. Keep this short.
  3. Choose privately. Each estimator selects a card without showing it.
  4. Reveal together. Everyone turns their card over at the same moment.
  5. Discuss the extremes. The people with the highest and lowest estimates explain their reasoning.
  6. Vote again. Repeat until the estimates converge. Two or three rounds is usually enough.
  7. Record and move on. If the team cannot converge, take the higher value, split the item, or flag it for a spike.

A hypothetical session

Consider a hypothetical team of five estimating the story "Allow users to sign in with their company's single sign-on." In the first round, the cards show 3, 5, 5, 5 and 13. The developer who played 13 explains that the company's largest customer uses an identity provider the product has never supported, which would need extra testing. The developer who played 3 had assumed only the two common providers. The product owner confirms the large customer is in scope. In the second round, the cards show 8, 8, 8, 13 and 8, and the team agrees on 8, noting the new identity provider in the acceptance criteria. Without the reveal, the 13 might never have been voiced, and the story would have been underestimated.

Running planning poker remotely

Distributed teams use dedicated planning poker web apps or estimation features built into work-tracking tools. The same rules apply: hide votes until everyone has chosen, reveal at once and discuss outliers. Cameras on and a shared view of the story help keep the discussion focused.

Common pitfalls

  • Averaging instead of discussing. The value of the technique is the conversation about differences, not the arithmetic.
  • Managers voting or overruling. Estimates should come from the people doing the work.
  • Estimating unrefined items. If the story is unclear, most votes will be "?". Refine it first.
  • Endless rounds. If estimates do not converge after three rounds, the item needs splitting or more information, not more votes.
  • Treating points as hours. Planning poker produces relative sizes; converting them to time commitments undermines honest estimates.

Planning poker is the most common way teams apply Fibonacci agile estimation, and it is usually done during backlog refinement or sprint planning. Its sibling technique, priority poker, uses the same mechanics to rank items by value instead of effort.

Frequently asked questions

How do you run a planning poker session?

The product owner presents a story and its acceptance criteria, estimators ask short clarifying questions, and each person picks a card in private. Everyone reveals at the same moment, the highest and lowest voters explain their thinking, and the team votes again. If estimates have not converged after about three rounds, split the item, take the higher value or schedule a spike.

What numbers are on planning poker cards?

Most planning poker decks follow a modified Fibonacci sequence: 0, ½, 1, 2, 3, 5, 8, 13, 20, 40 and 100. A question-mark card means the voter cannot estimate the item, and many decks include a coffee-cup card to ask for a break. Teams estimate in story points with these cards, though any relative scale works. See Fibonacci agile estimation.

Who should vote in planning poker?

The people who will build and test the work cast the cards, typically developers and testers. The product owner presents each item and answers questions but usually stays out of the vote. Managers should neither vote nor overrule the result, because estimates stay honest only when they come from the people responsible for doing the work.