Estimation argumentsPlanning poker that catches false consensus

Planning poker is good at one thing: when the cards differ, everyone notices. A 2 next to a 13 starts a conversation, and that conversation is where the real value of estimation lives.

The opposite case gets no attention at all. When every card shows 5, the facilitator locks the estimate and moves on. But a story point blends several things — volume of work, uncertainty, dependencies, coordination — and the same total can come from very different mixes.

Take a task called Migrate billing to a new provider. Alexey and Kira see mostly code: a new API client, a mapping layer, tests. Marina and Semyon have done a migration like this before and remember that the payment provider’s compliance review took two weeks of back-and-forth. Everyone plays 5. Nobody asks why. In the middle of the sprint the task is blocked, and the team discovers it had two different plans all along.

That situation has a name: false consensus (SmartPlanning calls it apparent consensus). It is more dangerous than an obvious disagreement, because nothing in the meeting signals it.

With arguments, the same meeting looks different: the cards still match, but the “Communication” row lights up red and the estimate doesn’t get the green celebration. One question — “Marina, why a 9?” — and the compliance review is on the table before the sprint, not in the middle of it.

An argument is one dimension of risk that your team cares about: code volume, communication, uncertainty, testing and so on. The facilitator sets up to 10 of them once per room, and from then on every vote has two parts.

1. Pick a card as usual — that’s your estimate.
2. Rate the arguments from 1 to 10 right under the deck, where 10 means high risk. Rating is optional: skip any argument you have no opinion on, and click a selected value again to clear it.
3. Reveal. Scores travel with your card and stay hidden until the reveal, so nobody anchors on the senior engineer’s “9”.
4. Read the heat map — one row per argument, one column per person, with the average and the spread (σ) of every row.

Three verdicts instead of one number

After every reveal SmartPlanning gives the round a verdict. It looks at the cards first and at the arguments second.

Full consensus

At least two votes, all the same card, no coffee card, and no contentious argument. The cards turn green with a short burst of confetti — lock it and move on.

Apparent consensus

The cards match, but at least one argument is contentious. No celebration: the result is highlighted in yellow and a banner names the argument to discuss.

No consensus

The cards differ. The heat map shows where the difference comes from, so the discussion starts from the cause, not from the numbers.

For every argument SmartPlanning takes the scores that were actually given — people who skipped it are left out — and computes the average and the standard deviation (σ). σ answers a simple question: how far, on average, are people from the team’s average?

Here is the “Communication” row from the example above, step by step:

3.0 is above 2.5, so the argument is contentious — and the verdict becomes apparent consensus even though every card says 5. The average alone (5.5, “medium risk”) would have hidden the problem completely: it’s the average of two people who see almost no risk and two who see a lot.

Why 2.5?

On a 1–10 scale a σ of 2.5 means people are, typically, about five points apart — one half of the room says “low risk”, the other says “high risk”. Smaller differences (“6 versus 7”) are normal noise and don’t deserve a meeting. The threshold is the same for every room, so a “contentious” label means the same thing in every team that uses it.

Notice the third row: one strong outlier among three people may stay just under the threshold. The verdict is a safety net, not a replacement for looking — a lone red cell on the heat map is still worth a ten-second question.

You can add up to ten arguments, but three to five work best. Every extra row costs each person a few seconds per task, and a heat map with ten rows is hard to read. Start with the presets — Code volume, Communication, Uncertainty, Testing, Dependencies, Refactoring — or type your own names.

A good test for an argument: think of the last task that blew up. Was it because of the amount of code? Waiting for another team? A part of the system nobody understood? Those are your arguments. Some starting sets:

Calibrate the scale once

“7 out of 10” means nothing until the team agrees what it means. Spend five minutes in your first session and write down anchors for each argument, for example:

Uncertainty 1–3

We’ve done exactly this before.

Uncertainty 4–7

Known approach, new place in the code.

Uncertainty 8–10

We’d need a spike to even start.

Keep the direction consistent: 10 always means more risk. An argument called “Test coverage” invites people to answer “how good is coverage” (high = good) — rename it “Testing effort” so a high score means trouble, like every other row.

A plain spread of cards gives you a vague prompt: “why 2, why 13?”. A contentious argument gives you a precise one. Use this script for every apparent consensus and every wide spread:

1. Find the red row. Read the argument’s name out loud — that’s the topic of the next two minutes.
2. Ask the highest score first, then the lowest — on that argument, not on the card. “Marina, you rated communication 9 — what do you expect to happen?”
3. Decide what changed. Usually one of three things: someone knew a fact the others didn’t (share it and revote), the task is really two tasks (split it), or nobody actually knows (park it and plan a spike).
4. Revote only if the understanding changed. If the new information doesn’t move anyone’s card, lock the estimate — but write the risk into the task description so it isn’t lost.

Arguments before the meeting

Arguments work the same way in pre-estimation. When the team estimates asynchronously, a task whose cards match but whose arguments disagree is marked Discuss instead of Consensus in the facilitator’s review — exactly the tasks worth bringing to the live meeting. And when you export the session, the file includes the average score of every argument for every task.

Classic planning poker vs. planning poker with arguments