Planning poker

One person creates the room and shares the invite link; everyone else opens it and types a name. Put the story or ticket title in the story field, and each person picks a card in private. The table only shows who has voted, not what they voted, so nobody anchors on the first number said out loud. That anchoring is the whole reason planning poker exists: estimates given after hearing someone else's number drift towards it. When the votes are in, hit Reveal, and if the numbers disagree, have the highest and lowest voters explain what they saw before voting again.

Rooms are meant for a session, not for records. The room lives on the server only while people use it and is deleted after two hours of inactivity. Names and votes are never stored beyond that.

Creating a room puts a six-character code in the address bar and nowhere else. The alphabet it is drawn from leaves out I, L, O, zero and one, so a code read aloud over a call or copied off a screen has no ambiguous characters in it. Anyone with the code can type it into the "Have a code?" box and reach the same table, which is worth knowing before you paste an invite link into a channel wider than the team.

Up to fifty people can be in one room; the fifty-first is turned away rather than silently dropped. Your display name, avatar and a random participant id are kept in this browser so that a reload puts you back in your own seat instead of creating a second one. They stay on your device, and clearing site data for ChronoClock removes them.

If the connection drops, the page retries five times with a widening gap before it gives up and shows "connection failed"; reloading starts the attempts again. Someone who closes the tab without a card on the table leaves the seat list, while anyone who has already voted stays visible so the round is not silently short of a vote.

Revealing turns three figures over. The average uses only the numeric cards, so a question mark or a T-shirt size is left out of it and a table that voted entirely in T-shirt sizes shows a dash rather than a number. The median uses the same numeric cards and is the more honest summary of the two, because one person's 34 cannot drag it. Agreement reads Yes only when every card on the table is identical, with at least two votes cast; anything else reads Discuss, and a round holding a single card shows a dash until somebody else plays one.

A worked round: the story is "add CSV export to the reports page" and five people vote 3, 5, 5, 8 and 13. The average comes out at 6.8, the median at 5, and Agreement says Discuss. The number to act on is not the average but the gap: the person on 3 had assumed the export already existed elsewhere, and the person on 13 knew the report is paginated. After both explain, the second round lands on 5, 5, 8, 8, 8. The average is still 6.8 and the median has moved to 8, but the spread has collapsed from ten points to three, and that is the part that tells you the team is now estimating the same piece of work.

A reveal is not a commitment. It is a prompt to talk, and the new round button clears every card so the same story can be re-estimated as many times as it needs.

Switching the deck mid-round does not clear the cards already played, so a number cast on the Fibonacci deck is still on the table when the T-shirt cards appear. Pick the deck before anyone votes, or press New round straight after changing it.

Reveal, New round, the deck and the story are open to everyone at the table rather than to a moderator. That suits a team of five and goes wrong in a room of thirty, where it pays to say out loud who is driving. Someone who only wants to watch should press Spectate: spectators are excluded from the vote count, so the round no longer waits for them.

Estimating in hours instead of points is the other common slip. The cards are a comparison between stories, not a schedule, and the Fibonacci gaps exist to stop the table arguing about whether something is an 8 or a 9. Turning points into dates is a separate job, and the project estimator below is where it belongs.