How to run planning poker

Five steps for a session

1. Agree on a scale

   Choose a deck such as 1, 2, 3, 5, 8, 13. Pick a familiar task as a reference and agree on its estimate. Estimate other tasks relative to that example.

2. Discuss the task

   Review the expected outcome, acceptance criteria, dependencies and unknowns. Before voting, everyone should understand which work the task includes.

3. Choose cards independently

   Each participant picks an estimate without following colleagues. Hidden cards help reduce the influence of the first estimate someone voices.

4. Reveal and discuss

   Reveal estimates and ask the people with the lowest and highest cards to explain their reasoning. Differences may reveal a missed scenario, technical risk or a different understanding of scope.

5. Refine and repeat

   Clarify requirements, split a large task if useful and run another round. Agree on a final estimate together: an arithmetic average is not a substitute for discussion.

Example: why estimates of 3 and 8 differ

The team estimates password recovery. One developer chooses 3, considering the form and email delivery. Another chooses 8, including token expiry, rate limiting and tests. After discussing these assumptions, the team clarifies scope and votes again. The value is in the requirements uncovered, rather than the number itself.

Why use Fibonacci cards?

The gaps between numbers grow as task size increases, helping avoid false precision when uncertainty is high. Teams can use other scales; a shared understanding of the values matters most. Splitting a large task is often more useful than debating two large numbers.

Common estimation mistakes

• Converting every estimate into a fixed number of hours.
• Treating the average of the cards as the team’s final decision.
• Estimating without acceptance criteria or discussing risks.
• Comparing story points across teams or using them to measure an individual’s performance.