planning-poker

Each estimator privately picks a card from a Fibonacci-like scale, then everyone reveals simultaneously to prevent anchoring. Outliers explain what they saw, the group discusses, and revotes once; two rounds is the usual budget before treating non-convergence as a scope problem.

Use planning poker when the team runs fixed-length iterations and individual estimates would plausibly diverge significantly. For obvious items, solo work, or continuous delivery with no iteration to plan against, a single senior estimate or t-shirt size is the right-sized tool.

Repeated failure to converge usually means the story was never ready — its scope is fragmented or undefined. Planning poker cannot fix a badly-scoped input; send the item back through a Definition of Ready check instead of adding more estimation rounds.

Four named modes: anchoring (a senior voice states a number early), poker on badly-scoped stories, false-fast convergence (playing safe numbers to avoid friction), and the empty ritual (the number produced never influences prioritization or planning).

No — points, ideal hours, or t-shirt sizes are all equivalent choices. What matters is applying the same relative scale consistently sprint over sprint, since the real deliverable is a long-run velocity signal, not per-item accuracy.