What is Planning Poker?

By kkrawczyk · Published July 5, 2026 · Updated July 5, 2026 · Markdown version

TL;DR — Key takeaways

• Planning poker is a consensus-based estimation technique where agile team members privately pick a card representing a story's effort, then reveal simultaneously to avoid anchoring bias.
• It was created by James Grenning in 2002 and popularized by Mike Cohn's book Agile Estimating and Planning (2005).
• The simultaneous reveal is the whole point: it surfaces genuine disagreement instead of everyone echoing the first number spoken.
• Most teams use a Fibonacci-style card deck because uncertainty grows with size — the gaps between cards reflect that.
• You can run planning poker with physical cards or free online tools — online works better for remote and async teams.

Planning poker (also called Scrum poker) is a consensus-based technique agile teams use to estimate the relative effort of user stories. Each estimator holds a deck of cards with numbers on them; for every story, everyone privately selects a card, and all cards are revealed at the same time. The team discusses differences and re-votes until it converges on an estimate.

The technique was created by James Grenning in 2002 while he was consulting on an Extreme Programming team whose estimation meetings had devolved into long, dominated-by-one-voice discussions. Mike Cohn later popularized it in his 2005 book Agile Estimating and Planning, and it has since become the most widely used estimation ritual in Scrum and other agile frameworks.

How does planning poker work?

Planning poker follows a simple loop: present a story, discuss it briefly, vote simultaneously, reveal, discuss the outliers, and re-vote until the team agrees. A typical session works like this:

1. Present the story. The product owner or facilitator reads the user story aloud and answers clarifying questions about scope and acceptance criteria.
2. Discuss briefly. The team talks through risks, unknowns, and assumptions. Timebox this — two or three minutes is usually enough before a first vote.
3. Vote simultaneously. Every estimator privately picks the card that represents their view of the story's effort. Nobody sees anyone else's card yet.
4. Reveal together. All cards are flipped at the same moment.
5. Discuss outliers. The highest and lowest voters explain their reasoning. This is where the real value lives: the outlier usually knows something the rest of the team doesn't — a hidden dependency, a forgotten edge case, or a simpler approach.
6. Re-vote. The team votes again with the new information. Most stories converge within two or three rounds; if one doesn't, that's a signal the story is too vague or too large and should be split or clarified.

Why use planning poker?

Planning poker produces better estimates than open discussion because it forces every opinion onto the table before anyone can be influenced. Three effects do the heavy lifting:

It prevents anchoring bias. In an open discussion, the first number spoken becomes an anchor — subsequent estimates cluster around it, even when it's wrong. Psychologists Amos Tversky and Daniel Kahneman documented this effect in their classic 1974 paper Judgment under Uncertainty. Simultaneous reveal removes the anchor entirely: everyone commits to a number before hearing anyone else's.

It engages the whole team. Every estimator must actively choose a card for every story. Nobody can zone out or defer to the loudest voice. The people who will actually build the feature — not just the most senior person in the room — shape the estimate.

It builds shared understanding. The discussion triggered by divergent votes is often more valuable than the estimate itself. When one developer votes 2 and another votes 8, the conversation that follows uncovers assumptions that would otherwise surface mid-sprint as a blown estimate.

Relative estimation also plays to a human strength. People are demonstrably poor at absolute estimation ("this will take 14 hours") but much better at comparison ("this is about twice as big as that"). Steve McConnell's Software Estimation: Demystifying the Black Art covers the research behind this in depth. Story points and planning poker lean into comparison instead of fighting it — see our guide to story points explained for the full picture.

Planning poker card values

Most planning poker decks use a non-linear number sequence, and the non-linearity is deliberate: uncertainty grows with size. You can meaningfully distinguish a 2-point story from a 3-point story, but nobody can honestly distinguish 19 points from 21 — so the deck simply doesn't offer those choices.

Common decks include:

Choosing between numeric scales and t-shirt sizes is its own decision with real trade-offs — our comparison of Fibonacci vs t-shirt sizing walks through when each one wins.

Many decks also include a ? card ("I have no idea — we need to talk") and a coffee card ("I need a break"). Both are useful signals, not jokes.

Planning poker online vs. physical cards

Physical card decks work well when the whole team shares a room, but online planning poker has become the default for a simple reason: most agile teams are now at least partially remote. An online tool gives you:

• No setup. Share a link; everyone joins from a browser. No printing, no shuffling, no forgotten decks.
• A true simultaneous reveal. The tool holds every vote hidden until the reveal, which is actually harder to guarantee with physical cards and eager teammates.
• Automatic statistics. Average, spread, and vote distribution appear instantly, so the facilitator doesn't do mental math.
• Async support. Team members in other time zones can vote when they're online — see our guide to async planning poker.

Our own tool, poker but for planning, is free, requires no sign-up, and supports Fibonacci, modified Fibonacci, t-shirt, and powers-of-two decks. Create a room, share the link, and vote.

Planning poker best practices

A few habits separate productive estimation sessions from painful ones:

• Timebox the discussion. Two to three minutes of discussion before the first vote. If a story needs twenty minutes of debate, it isn't ready to estimate — send it back for refinement.
• Use a facilitator. One person presents stories, keeps time, and makes sure outliers get heard. The facilitator shouldn't dominate the estimate.
• Break down large stories. Anything landing at the top of your scale (13+, XL+) is a candidate for splitting. Large estimates are mostly a measure of ignorance, not size.
• Estimate against a reference story. Keep one or two well-understood, already-completed stories as a baseline ("our login page rebuild was a 5").
• Don't average and move on. If votes are 3, 3, 5, and 13, the 13 isn't noise to be averaged away — it's the most important data point in the round.
• Let velocity do the forecasting. Estimates in points, tracked against actual velocity over several sprints, give better forecasts than any amount of up-front precision.

Frequently asked questions

How long should a planning poker session take?

A well-run planning poker session takes about one to two hours and covers ten to twenty stories — roughly five minutes per story on average. If sessions consistently run longer, the stories entering the session are usually under-refined; fix the backlog refinement process rather than extending the meeting.

Can planning poker be done asynchronously?

Yes. Async planning poker lets team members vote at their own convenience within a voting window, and the facilitator reveals the results once everyone has voted. It works especially well for teams spread across time zones. Our guide to async planning poker covers the workflow and its trade-offs in detail.

What's the difference between story points and hours?

Story points measure the relative effort, complexity, and uncertainty of a piece of work; hours attempt to measure its absolute duration. Points sidestep the human weakness for absolute estimation and absorb per-person speed differences — a senior and a junior developer can agree a story is "a 5" even if it would take them different amounts of time. Our guide to story points explains the distinction and why it matters.

Do you need special cards for planning poker?

No. Any agreed set of values works — physical cards, fingers, sticky notes, or a free online tool. What matters is that everyone commits to an estimate privately and reveals simultaneously. An online planning poker tool simply automates the hiding, revealing, and bookkeeping.