Guide

How to run a sprint planning poker session

This guide is for the person who facilitates the session: a Scrum Master, an engineering lead, or whoever keeps the meeting moving. It assumes your team already knows the basics. If not, start with What is planning poker?.

Before the session

Prepare the backlog. Put the stories in priority order, each with a short description and acceptance criteria. Estimating a story nobody understands wastes everyone's time, so anything vague should be refined first.
Invite the people who do the work. Developers, testers and designers who will build the story should vote. The product owner should attend to answer questions, but usually does not vote.
Pick a scale. Fibonacci works for most teams. See Estimation scales if you want to compare options.
Set a reference story. Choose a finished story that everyone remembers and agree on its size. Every new story is then compared with it.

Set up the room

1. Sign in as a guest with a name, or with an email magic link, and create a room. Give it a name and choose its scale.

2. Share the room URL in your team chat or calendar invite. The URL is permanent, so you can reuse the same room for the next sprint and keep the history of earlier rounds.

3. Choose the room settings that suit the session:

- Auto reveal votes reveals the cards as soon as every voting player has voted, so nobody has to wait for a click.

- Show countdown adds a short countdown before the cards flip.

- Lock room means only the room creator can reveal votes, undo a reveal, start new rounds and rename them. Use it if you want to be the only one driving the session.

Run the session

1. Present one story. Read it out, let the team ask questions, and name the round after the story so the room history is readable later. Rounds can be renamed from the room.

2. Vote. Each player picks a card. Cards that have been voted on show face down, so you can see who is still deciding without seeing what they chose.

3. Reveal. With auto reveal on, the cards flip when the last player has voted. Otherwise, use the reveal button. If someone needs another minute, wait for them: revealing early defeats the purpose.

4. Discuss the outliers. Look at the highest and lowest cards first and ask each person what they were thinking. The room shows the average and the agreement level, but the conversation is what improves the estimate.

5. Vote again or settle. If the gap closed, record the estimate in your tracker. If not, undo the reveal, vote again, and if it still does not converge, park the story and split it or get more information.

6. Start a new round for the next story.

Keep a rough time limit for each story, such as five minutes. If a story is taking longer, it usually needs to be refined outside the meeting.

Use the facilitator tools

Viewer mode. If you are facilitating and do not want to vote, turn on viewer mode from your name in the room header. You stay in the room and can still undo a reveal and start new rounds, and auto reveal does not wait for you.
Hidden keyboard voting. If you share your screen, a click on your card would show your estimate to everyone watching. Turn on hidden vote mode in the room, and each card shows a letter beneath it. Press that letter to vote, and press it again to withdraw the vote.
The question mark and coffee cards. Every room has a card for "I cannot estimate this yet" and one for "I need a break". Neither counts toward the average, and a question mark is a signal that the story needs more detail.
Room history. Open the history from the room to see previous rounds with their average and agreement, which helps when you write up the outcome of the session.

Common problems

One loud voice sets the number. Hidden votes fix most of this. If it still happens, ask the most senior person to explain their card last.
Everyone always votes the same. That is fine for small stories. If it is true of large stories too, the team may be anchoring on a reference that is too far away, or the scale may be too coarse.
Estimates keep getting bigger in the discussion. Compare with the reference story, and split anything larger than the biggest card you are willing to accept in a sprint.
The session is dragging. Timebox each story, and estimate only what the sprint needs.