A scenario that cannot be wrong is not a scenario. It is a description.
Before a scenario goes active, Palanor asks for one more thing: what would have to be observably true for it to have happened, and where that truth would be read. This is the resolution criterion, and it is written at the moment you state the scenario — never afterward, when the answer is already known.
The test
A criterion passes if someone who could see the future could settle it — happened, or did not happen — without asking you a single follow-up question.
If settling it would need a conversation, a judgement call, or a definition, it does not pass. Three things are required:
- **Observable.** It names a number, a filing, an index, or a published figure. Not a feeling.
- **Dated.** It settles on or before the scenario's horizon.
- **Attributable.** It names exactly where the truth gets read.
Some examples, in both directions:
- Does not pass: "AI meaningfully compresses research margins."
- Passes: "Gartner, Forrester and IDC report combined subscription revenue below FY2025 levels in their FY2027 annual filings."
- Does not pass: "The capital window closes."
- Passes: "US IPO count in H1 2027 is below 60, per the Renaissance Capital IPO index."
The second version of each is not more precise for its own sake. It is the version that can be checked by someone who does not already know what you meant.
Numen drafts three, you choose one
Nobody writes a good criterion from a blank field. Numen reads your hypothesis and your signal map and offers three ways the scenario could be proven wrong — one that reads a company disclosure, one that reads a third-party index, one that reads a metric you already track internally.
They differ on purpose. You are choosing between genuinely different ways of being wrong, not between three wordings of the same test. Take one, edit it into your own language, or write your own.
Each draft carries a short "Why this settles" note. That note is the standard being taught: read a few and the shape of a good criterion becomes obvious.
When Numen declines
Sometimes Numen will say the scenario cannot be resolved as written. That is not an error, and it is not Numen being unhelpful.
It means the hypothesis names nothing countable — no market, metric, company, population or price. "Things get harder" is a mood, not a forecast. When this happens, sharpen the hypothesis: name the thing that would change, by how much, and where it would show up. The scenario gets better, not just the criterion.
Draft is not active
A draft may be unfalsifiable. That is where thinking happens, and Palanor will not interrupt it.
An active scenario is different: it is one your organisation is steering by. Steering by a claim that nothing could disprove is the exact failure Palanor exists to correct, so **Save draft** always works, and **Activate** waits until the scenario can be proven wrong.
Criteria are locked, not edited
Once a scenario goes active, its criteria, its source, and its settlement date stop being editable — along with the probability recorded at that moment.
This is deliberate. A criterion that can be revised after the fact is not a claim, it is a commentary. If a criterion turns out to be badly formed, the scenario is voided in public and restated as a new one. Nothing is quietly rewritten.
The probability is frozen for the same reason. Your scenario's live probability keeps moving as the lattice recomputes; the number it is judged against must not.
What happens at the horizon
When the settlement date arrives, the scenario resolves — happened, did not happen, or ambiguous.
Ambiguous is a real outcome, not a failure state. If a criterion turns out not to settle cleanly, it is recorded as ambiguous, carries no score, and is kept on the record. A criterion that failed to settle is a lesson about how it was written, and hiding it would waste the lesson.
Related: [How Probability Is Calculated](/help/scenarios/how-probability-is-calculated) · [Building Your First Scenario](/help/scenarios/building-your-first-scenario)