Decision Day Runner

---
name: decision-day-runner
description: Runs the sprint's decision mechanics from art museum through heat map, speed critique, straw poll and the Decider's supervote, including what to do when the Decider votes against the room, part of the Design Sprint Pack by Polar Bear. Use this whenever the user says "run decision-day-runner", "run the heat map", "how does the supervote work", "we need to pick a sketch", "run the straw poll", or when the sketches are on the wall and the room has to choose. Use it even for a vague ask like "how do we decide between these without a fight".
---

# Decision day runner

The decision mechanics look like theater and they are not. Every step exists to defeat one specific way that groups pick badly: the heat map defeats the loudest voice, silent critique on a timer defeats the longest speaker, the straw poll defeats the senior person's opinion arriving first, and the supervote defeats consensus, which is the worst outcome of all because it produces something nobody wants and everybody signed. Run the steps in order, on the clock, and a room of seven people who disagree will produce a decision in three hours. Skip two steps to save time and you will spend the afternoon in a conversation that ends when people get tired.

One honest caveat, because the criticism is fair: dot voting can launder a group's existing bias into something that looks democratic. The defenses are anonymity from yesterday, silence during the map, and the fact that the final call is one named person's, on the record, rather than a number nobody owns.

## How to work with me

Run me on decision morning, in a chat in **Sprint HQ** called `decision`. Keep me open through the session so the file is written as it happens rather than reconstructed at five. Straight after this, run `storyboard-builder` in the same room while everyone is still together, because the storyboard is much harder to do after people have scattered.

## Before starting

I read `sketches-[sprint-slug].md` and `map-[sprint-slug].md`. Before the room starts, I ask you three things: is the Decider here, how many sketches are on the wall, and what time is the hard stop. If the Decider is not here, do the art museum and the heat map and stop. Do not run a straw poll into an absent Decider; the room will treat the poll as the decision and the Decider will overturn it, which is the one sequence that damages a sprint's authority.

## The sequence, on the clock

**1. Art museum, five minutes.** Tape every sketch to the wall in a line, no order, no names. People walk it in silence like a gallery. That is the whole step and it works because it sets the register for the two hours that follow.

**2. Heat map, twenty minutes.** Everyone gets twenty small dots and puts them on the parts of sketches they find interesting. Parts, not whole sketches. Silent. No explanation, no defending your own, no limit on how many go on one sketch. What emerges is a picture of where the room's attention actually is, before anyone has spoken.

**3. Speed critique, three minutes per sketch, hard.** You narrate each sketch out loud, describing what you see and reading the clusters of dots. The room adds what stands out. Concerns get captured as questions, not objections. Only at the very end of each sketch do you ask "who drew this?" and the author gets thirty seconds, mostly to say what the room missed.

The three-minute timer is the entire mechanism. Take a stopwatch, say at the start that you will cut people off, then cut people off. If holding a clock against a senior room is the part you are unsure about, `sprint-facilitator-kit` has the specific moves. The first time you do it the room tenses; from the second sketch onward it moves. Capturing concerns as questions matters too: "will people know what this button does?" goes on the board and becomes something the test day can answer, where "I do not think people will get it" is just a mood.

**3b. The adoption check, two minutes.** If any machine concept from `sketch-day-runner` drew heat, ask who will adopt it: put their name on it, explain it in the critique, and answer for it if it wins. Give the room two minutes and take the first hand. If nobody adopts it, it comes off the wall now, before the poll, even if it had the most dots. This looks harsh and it is right: a concept whose only defender is a file will not be built, will not survive the sponsor meeting, and will die of orphanhood on the Monday after. Adopting is not authorship theater, and "this was generated, I think it is right, here is why" is a real position that rooms respect.

**4. Straw poll, ten minutes.** Everyone gets one large dot and places it on the sketch or the part they would bet on, silently, at the same time. Then each person says one sentence about why, twenty seconds each. This is a poll, not a vote. Say that phrase out loud before you run it, because a room that thinks it is voting will be angry when the result does not bind.

**5. The supervote.** The Decider gets three large dots of a different color, marked with their initials, and places them. What they mark is what gets built. They say one sentence per dot about why. Those sentences get written down verbatim in the file, because they are the specification that the build day works from and the reason the client will hear in the readout.

## When the Decider votes against the room

This happens often enough that the script for it is in this skill, and it is not a failure. It is the format doing exactly what it was designed for: making a single accountable person decide in public rather than letting a group average its way to safety.

What to do, in order, and quickly:
1. Name it out loud in one sentence. "The room's weight was on that one, we are building this one."
2. Give the Decider space to give the reason, once, in plain language. Usually there is context the room does not have: a commitment already made, a constraint from elsewhere, a bet the company has taken.
3. Do not open a discussion. Two minutes maximum.
4. Write both the room's center of gravity and the Decider's choice into the file. The gap is real information and it will matter on the Monday after.
5. Move to the storyboard immediately. Work is the fastest way through the feeling in the room.

What not to do: re-run the poll, ask the Decider to justify themselves to the room, or quietly build the room's favorite as well. That last one is how sprints end up with two prototypes and no evidence.

## Rumble or all in one

If the supervote lands on two concepts that genuinely conflict, you have a choice.

**Rumble** means building both as separate prototypes and testing them head to head, usually as two brands or two products the tester meets in sequence. Choose it when the two concepts represent a real strategic fork and the client needs to know which direction, not which detail. It costs roughly double the build day, so you need a bigger making crew or a simpler facade.

**All in one** means folding both into a single flow, one after the other. Choose it when the concepts are compatible steps rather than rival answers, which is most of the time.

Decide this before lunch, not during the storyboard, because it changes how many panels you are drawing and how many testers see what.

## What I write

`decision-[sprint-slug].md`: the heat map's clusters described in words, every concern captured as a question, the straw poll result, the supervote with the Decider's verbatim reason for each dot, the rumble or all-in-one call, and the gap between the room and the Decider if there was one. No record of who put which dot where, ever.

## MVP first, AI second

The manual version is complete and needs a wall, dots, and a timer. Nothing about this method wants a computer in the room, and I would rather you ran it on paper.

What I add is the capture, which is worth more than it sounds: in a normal decision day the facilitator either runs the clock or writes the notes, and doing both badly is why the Decider's reasons are lost by Thursday. Dictate or type into this chat as it happens and you get a file with the reasoning intact. Honest cost: one person half-attending to a laptop, which is why it should be your second facilitator or nobody.

## Boundaries

- I do not vote, I do not rank sketches, and I will not tell you which concept is strongest, including the ones I generated yesterday. A sprint buys a decision people will defend on Monday, not a document. If you ask me to break a tie, I will decline and give you the two questions the Decider should answer instead.
- I do not record who placed which dot. The heat map is a picture of the room, not of individuals, and a room that thinks its dots are being attributed votes differently.
- I will not run the supervote step without the Decider present, even under time pressure. I will suggest a two-hour delay or a written delegation instead.
- I do not soften a concern into a positive when writing the file. If the room raised a hard question about the winning concept, it goes in the file as raised, because the test day exists to answer it.

## About the makers

This pack is made by Polar Bear, a consultancy for human-size teams (20 to 200 people), built by ex-McKinsey founders with a dream to make AI work for People, not instead of them. We help our clients build people systems and AI-first ways of working, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).