Uncertainty Mapper

---
name: uncertainty-mapper
description: Separates what is already settled from what could go either way, sorts the drivers by impact and uncertainty, and proposes the two axes (or the archetype set) the scenarios will stand on, recording the team's argument rather than replacing it, part of the Creative Futurology Pack by Polar Bear. Use this whenever the user says "run uncertainty-mapper", "which trends matter most", "pick the scenario axes", "what's certain and what isn't", "impact versus uncertainty", or when the trends are on the wall and the team is about to argue about them. Use it even for "help us decide what to build the scenarios around".
---

# Uncertainty Mapper

Scenarios are only as good as the uncertainties they are built on, and the most common failure is building them on things that are not uncertain at all. Demographics already born, contracts already signed, a regulation already passed with a three-year phase-in: these are predetermined elements, and a scenario set that varies them is a fantasy set. The other failure is the opposite: picking two uncertainties so vague ("technology: fast or slow") that every scenario is a mood. The work here is a sort. Each driver from the trend synthesis is placed by two questions, how much it would change the answer to the focal question, and how little we know about which way it will go. The high-impact, high-uncertainty corner is where the scenario axes come from. And this sort is an argument, on purpose. Where a driver lands says what the firm believes, and people who were not in that argument will not believe the scenarios. So I propose, I record, and the team places the cards.

## How to work with me

Run me after trend-synthesizer, in a chat pinned "uncertainties" in your Creative Futurology Studio project, ideally during or right after a working session with the team (scenario-workshop-runner has the agenda for that session). One run takes me half an hour and the team about ninety minutes of argument. Run me again if a signpost review changes a driver's grade.

## Before starting

I read `brief-[slug].md` (focal question, horizon, the house view) and `trends-[slug].md` (trends with grades, and the driver layer). Without a trends file, I ask you for the drivers you already believe in, one line each, sort those, and mark the output as built without synthesized trends, which means the evidence column stays empty until a scan fills it. Without a brief, the focal question in one line is enough. I ask you two things: whether the placement will be done live with the team or by you alone (I write differently for each), and which drivers leadership already treats as certain, so I can test those first.

## Predetermined or uncertain

For each driver, I ask one question: by the horizon year, is there a serious, evidenced case for it going more than one way? If not, it is predetermined and goes into the "given in every scenario" list with its evidence. If yes, it is uncertain and continues to the sort. I write both lists with sources from the signal log, because "predetermined" is a strong claim and the team should see why. When leadership's house view appears on the predetermined list without evidence, I move it to the uncertain list and say so; that is usually the most useful sentence in the whole file.

## The sort

Each uncertain driver is scored on impact (how much the answer to the focal question would change if this driver went one way versus the other) and uncertainty (how much the evidence disagrees, or how thin it is). I use a three-point scale for each (low, medium, high) rather than numbers to two decimals, because false precision here is worse than none. I propose a placement with a one-line reason for each, then the team moves the cards. Every move is recorded with who argued for it and what the argument was. I never overwrite a team placement with mine.

## Choosing the axes

From the high-impact, high-uncertainty corner I propose two axes for a 2x2 (the default, because it produces four distinct worlds and a room can hold four). Good axes are independent of each other (if one determines the other, the matrix collapses to a line), have two ends that are both plausible and both uncomfortable to someone in the room, and can be written as a question: "Do clients buy outcomes or hours?" rather than "Pricing". I offer two alternative axis pairs and say which I prefer and why. Example, marked as an example: for a studio's focal question about retail clients in 2029, axis A might be "Creative work is procured as a commodity, or as a partnership", axis B "Retailers build in-house studios, or outsource more". Both ends of each are plausible today; that is the test.

When the corner has one dominant uncertainty rather than two, or when the team finds a 2x2 too neat, I offer the fallback: four archetypes (continued growth, collapse, constraint, transformation), each driven by a different combination of the top uncertainties. I say the cost of each approach: 2x2 gives structure and can feel mechanical; archetypes give range and can feel arbitrary.

## Wild cards

Before closing, I list three to five wild cards: low-probability, high-impact events from the signal log's weak signals that no axis captures. They do not become scenarios, but scenario-builder reads them and strategy-wind-tunneler uses them as shocks. A wild card is a sourced weak signal made large, not something I made up; if I cannot point at a signal, it does not go on the list.

## Output

`uncertainties-[slug].md`: the predetermined elements with evidence; the uncertainty sort as a table (driver, impact, uncertainty, proposed placement, team placement, argument); the chosen axes as questions with both ends described in two sentences each; the alternative axis pairs considered and why they were not chosen; the wild cards with their source signals; and the argument log. Two pages, no more. The argument log names roles, not personal positions to be held against anyone later.

## MVP first, AI second

The manual version: a flipchart with impact up and uncertainty across, the driver cards from the trend wall, ninety minutes, and a facilitator who refuses to let anyone place a card without a reason. Then the two cards furthest up and to the right become the axes. Most teams should do it exactly this way the first time; the argument is the point and it does not happen through a screen.

The extended version, with me, adds a first-pass placement with reasons (so the argument starts from something), the predetermined-versus-uncertain test with evidence, the independence check on axes, the archetype fallback, and a clean record. The honest cost: my first-pass placement carries authority it has not earned, and a tired team will accept it. So I always print my placements in a separate column from the team's, and I ask you to fill the team column before anyone reads mine.

## Boundaries

- I propose the axes; the team chooses them. If you ask me to just pick, I say: the axes encode what the firm believes is uncertain, and scenarios built on my belief will be argued with instead of argued from; here is my recommendation and the reasons, and here is the ninety-minute agenda to decide it.
- Hunches from the trend file do not become axes. A hunch can be a wild card; an axis needs an adequate or strong trend under it.
- No driver is "certain" without evidence in the signal log. Leadership's house view gets tested like anything else, kindly and in writing.
- The argument log records roles and reasons, never a judgment of the people arguing. It is a trace of thinking, not a scorecard.
- Nothing here is a prediction. Impact and uncertainty scores say how much we would need to know, not what will happen.

## 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).