How to Run a Planning Poker Session

A practical, step-by-step facilitation guide for in-person and remote teams.

Running planning poker well is mostly about good facilitation. Here's a simple six-step process that keeps sessions fast, fair and focused — whether your team is in one room or spread across time zones.

Before the session

• Groom the backlog. Each item you plan to estimate should have a clear description and acceptance criteria. Vague items produce vague estimates.
• Invite the right people. Only those who will actually do the work should vote. Stakeholders can join as spectators.
• Pick a scale. Fibonacci is the default; see our guide to cards and scales to choose.
• Set a reference item. Agree on one small, well-understood item and call it a "2" (or a "3"). Everything else is estimated relative to it.

The six-step loop

1. Present the item. The product owner reads it aloud and answers questions.
2. Discuss briefly. Clarify scope and edge cases — but time-box it to a minute or two.
3. Vote privately. Everyone selects a card without showing it, so no one is anchored.
4. Reveal together. All cards flip at once.
5. Discuss outliers. The highest and lowest voters explain their thinking. This is where hidden complexity surfaces.
6. Re-vote. Repeat until the team converges. Record the agreed estimate and move to the next item.

Running planning poker remotely

For distributed teams, a shared online tool replaces the physical deck. Everyone joins the same room from a link, picks a card on their own screen, and the cards reveal simultaneously for all — preserving the anti-anchoring benefit. Look for these features:

• Instant rooms with a shareable link and no sign-up.
• A synchronized reveal so nobody sees votes early.
• Spectator mode for facilitators and stakeholders.
• Custom scales and a visible average or consensus after the reveal.

Poker Sizing does all of this for free.

Facilitation tips

• Keep it moving. If an item triggers a long debate, split it or park it rather than stalling the session.
• Estimate effort, not time. Points capture complexity, uncertainty and volume — not hours. See story points explained.
• Watch for big spreads. A 2 next to a 13 is a signal the item isn't understood the same way by everyone.
• Don't average the votes. Discuss and converge; averaging hides disagreement.
• Re-estimate rarely. Once an item is sized and understood, avoid re-litigating it.