Planning poker is combination of analogy, expert opinion, and disaggregation in a fun way so that it will result quick and reliable estimates. All the team members are included in planning poker. On any agile project, you will have typically ten team members or less. If it does, the team can be split in twos. Then estimation is done independently by each team. The PO participates in planning poker but he or she doesn’t estimate.
At the beginning, each team member is handed deck of cards. All the cards are marked with a valid estimation number. Each member will be given a deck that reads a number series. The most popular of these estimation numbers are Fibonacci numbers. (1,2,3,5 ,8,13,2, 34,55 and so on). The cards are prepared before the planning poker meeting.
Then a moderator describes each of the user stories or theme that team is planning to estimate. Though generally the product owner acts as a Moderator, anyone can be a moderator. No special privilege or role is associated with the moderator. The product owner will answer all questions that the team members have.
The goal of estimation is to be somewhere on the left of the actual effort line. Important thing to remember is that this process is not about deriving an estimate that will resist all future inspection.
After all the queries are resolved, each team member selects a card that represents their estimation. Each estimator has to make a selection before Cards can be visible to everyone. Cards are kept private until everyone has estimated. Then the cards are turned over at the same time.
Then, all cards are instantaneously spun over and displayed so that all estimators can see each estimate. Chances are that these estimates will differ significantly. In that case, the high and low estimators will explain their estimates. The focus of this process is not to attack these estimators but to learn on what basis these estimations were assigned.
After this discussion, each team member will re-estimate by selecting a card. The earlier mentioned process will be followed again. Chances are that the estimates will meet by the subsequent round. Continue to repeat the process until all the estimators converge on a single estimate that can be used for the story. Very rarely it takes more than three rounds. Continue this process until estimates are moving closer together and they everyone converges on a single estimate.