← Back to articles

Why Planning Poker Works (and When It Doesn't)

ESTIMATION · 5 MIN READ

Planning poker looks like a novelty at first — grown adults holding up numbered cards instead of just saying a number out loud. But the card-flip mechanic isn't decoration. It's solving a specific, well-documented problem: the first number said out loud in a group anchors everyone else's estimate, whether they realize it or not.

The anchoring problem

If a senior engineer says "I think this is a 5" before anyone else speaks, the next few answers cluster around 5 even from people who privately thought it was closer to a 13. Nobody's being dishonest — anchoring is just how people estimate under uncertainty in a group. Planning poker breaks this by forcing everyone to commit to a number simultaneously, face-down, before any number is spoken.

The reveal is the actual point of the exercise. When a team flips 3, 3, 13, 5, the wide spread isn't a failure — it's the estimate doing its job. It means someone knows something the others don't: hidden complexity, a dependency, a piece of legacy code nobody else has touched. The follow-up conversation, not the number itself, is where the estimate actually gets more accurate.

A tight spread (5, 5, 5, 3) usually means real shared understanding. A wide spread means talk before you average anything.

Where it breaks down

Planning poker assumes the group actually understands the story well enough to estimate it, and that's not always true. A few common failure modes:

A smaller tweak that helps

Include a "?" card in the deck for genuine unknowns, and a coffee-cup or break card for when someone needs a pause. Teams that skip these end up with people guessing numbers for stories they don't understand yet, just because there's no honest way to say "I don't have enough information."

Written by The Agile Shop.