Planning poker is a conversation design
Planning poker is often introduced as an estimation technique, but the better way to understand it is as a structure for conversation. The hidden vote stops the first confident answer from becoming the anchor. The reveal gives the team a fast map of disagreement. The follow-up discussion turns that disagreement into useful information about risk, missing context, and different implementation paths.
A useful round has five beats
A practical round is simple: explain the backlog item, let everyone vote privately, reveal together, ask the high and low voters what they saw, then either estimate again or send the item back for refinement. The facilitator should keep the rhythm moving. If the conversation drifts into design detail, dependency discovery, or acceptance criteria repair, that is a signal that the story needs more refinement before another vote.
What the number should and should not mean
A story point is a relative estimate, not a promise of hours from a specific developer. The number should summarize the team's view of size, uncertainty, and risk compared with other work they know. It should not hide questions about deadlines, staffing, or availability. Those questions are real, but mixing them into points makes both conversations worse.
When to stop estimating and refine
The most useful planning poker outcome is sometimes no estimate. If the spread is wide because the item is too large, unclear, or dependent on unknown work, the team has learned something important. Split the item, add acceptance criteria, or clarify the technical path. A clean estimate after that work is more valuable than forcing consensus on a story nobody understands.