Get started

The method

Magic Estimation is also known as silent grouping, affinity estimation or swimlane sizing.

Why sort instead of estimating one by one

People are better at comparing than at guessing absolute sizes. Saying "this is bigger than that" is quick; agreeing that something is "exactly 5" is slow. When the whole team sorts every issue into size columns, most issues land in the same place for everyone, and only a few need a conversation.

How a session runs

  • Silent. Everyone sorts on their own, without discussing. Placements stay hidden until the reveal, so nobody anchors on the most senior voice.
  • Relative. Compare issues with each other and with work you have done before, not with hours.
  • Outliers first. After the reveal, talk only about the issues where the votes are far apart. Often someone knows something the others did not.
  • Done. Keep the median for the rest and write the values to Jira.

Tips

  • Agree on one or two reference issues before you start, for example "the login form was a 3".
  • Give the sorting a time box. If an issue is unclear, put your best guess and say so in the discussion.
  • Keep sessions to one sprint or one epic, so the issues are comparable.

Asynchronous works too

People do not have to sort at the same time. Open the session on Monday, reveal it in Wednesday's refinement.