Methodology
How ScenarioPilot works
ScenarioPilot runs a strategic foresight cycle: it scans for weak signals, combines them into drivers of change, builds four scenarios on the two most critical uncertainties, and attaches strategies and indicators to each. This page describes what it does today, where a person decides, and how its quality is measured. Anything not yet live is listed under Roadmap at the end.
Where does the evidence come from?
From public sources searched in parallel for each scan. One model call first breaks your theme into research angles across the STEEP+V categories (social, technological, economic, environmental, political and values), which widens the set of search queries. Duplicates are removed, a relevance check drops items that are off the theme, and every remaining item is tagged with how many independent sources report the same thing: two or more count as corroborated.
| Source type | Where it comes from |
|---|---|
| News | News search services |
| Web | Web search |
| Research | Scholarly publication records |
| Patents | Patent search |
| Open data | Public dataset catalogues |
How is a weak signal scored?
Each signal gets three scores from 0 to 100: relevance to the theme, potential impact, and novelty. Each signal is also fact-checked: is the claim in its source, and is it specific enough to measure, with an actor, a number or a date? A signal that fails the check stays visible, marked with the reason, so you can judge it yourself, and the drivers built on it inherit that status.
How are drivers and scenario axes chosen?
A driver is a force of change that combines two to four signals, drawn from at least two STEEP+V categories wherever the evidence allows. Its impact (how large the effect) and its uncertainty (how unpredictable over the horizon) are rated separately, as high, medium or low. The two scenario axes are the critical uncertainties: drivers rated highest on both impact and uncertainty, chosen so that neither axis predicts the other, and built on drivers with verified evidence wherever there is one. Their four combinations are the four worlds, each with a narrative, its implications and its early warning signs.
What comes with each strategy?
Each world gets two strategies. Each one has a kind (no-regret, hedge, option or bet), an accountable role, the observable trigger that would activate it, and its opportunity and risk. What a strategy relies on but the evidence does not establish, such as a budget, a team or a market fact, is written down as an assumption, with how far the evidence supports it, how much depends on it, and the test that would confirm or break it.
How do you know which world is arriving?
Through indicators: each one is read from outside your organisation, has a threshold, and takes its starting value from the evidence, or says plainly that none was found. When you log a reading that crosses a threshold, a trigger opens for the world that indicator belongs to, and you decide what to do with it.
Where does a person decide?
At every step that matters. You set the theme. You can run the cycle one stage at a time and choose which signals go into the drivers. Every trigger stays open until you acknowledge or close it. A project is seen only by you and the people you invite. A public link exists only if you create one, and anyone who has that link can open what it shares.
How is quality measured?
With a fixed set of five test themes, run through the live pipeline. Each run gets a score from 0 to 100 from the weights below, plus integrity checks that must stay green, and separate checks on the decision layer: strategies, assumptions and indicators. A change to a prompt or to the pipeline is kept only if the average score goes up and every check still passes.
| What is measured | Weight |
|---|---|
| Signal count against a target of about 30 | 30% |
| Spread across the six STEEP+V categories | 25% |
| Specific, non-generic signal titles | 20% |
| Mean relevance to the theme | 15% |
| Share of signals that pass the fact-check | 10% |
What does ScenarioPilot not do?
- It does not put probabilities on scenarios. The four worlds are plausible futures to plan against, not a forecast to bet on.
- It does not decide for you. Strategies, assumptions and indicators are proposals; you choose what to act on.
- It does not take your company documents. Document upload stays switched off until the providers’ terms are checked.
- It does not watch indicators on its own. A trigger opens when a reading you log crosses a threshold.
What is on the roadmap?
These are planned or in testing. None of them is a live feature today.
- A collaborative track: stakeholder rounds that vote on drivers and axes, with a facilitator’s approval at each step. It is with testers now, not open to every project.
- Testing each strategy against all four worlds, so you can see which moves hold up in more than one future.
- Indicators read automatically from external data, instead of readings you log.
- Official and regulatory sources searched in every scan.
- Measuring, theme by theme, how many of the developments an expert says must be found the scan actually finds.