Living Plans · Guide
Published 2026-09-27
Guide · Testing what the plan assumes
How to know if your strategic assumptions still hold
An assumption does not fail loudly. It was formed in one quarter, against one set of facts, and it keeps looking true because nobody has put it in front of new facts. The way to know whether it still holds is not a workshop or a hunch — it is a dated re-test against evidence the organisation has already published, run on a schedule short enough that the answer still changes a decision.
When a strategy is written, its assumptions are examined. That is what the planning season is: months of challenge, testing, red-teaming. The examination ends when the document is approved — and the assumptions keep quietly aging after it. The belief that “customers will pay for the premium tier” was checked in September; the question is what it is worth in March, after two quarters of the company’s own pricing and churn data.
This is why strategies fail without anyone having decided anything. No single piece of evidence is loud enough to count as bad news on its own — a soft quarter is weather, a delayed launch is a detail — so each gets explained, and the assumption underneath all of them is never re-tested. Knowing whether it still holds means naming it, dating it, and putting it against the newest published facts on a schedule, not waiting for a failure loud enough to make the case by itself.
§ 02 · What the re-test produces
“Partly true” is not a verdict. It is a way of postponing one.
The output of a re-test is a record, and the record has exactly three possible entries. Anything softer — “broadly on track”, “directionally right”, “monitoring closely” — feels careful and is in fact the mechanism by which a broken assumption survives three reporting cycles without anyone having to say so.
Three moves produce it. Write the assumption as a claim about the world that could be wrong. Name the evidence that would move it before you look at anything — for a company’s own assumptions, that means its own published reporting, not a forecast someone chose because it agreed. Then give the claim one of the three verdicts, with the source and its date attached, and sign it.
Holds
The newest published evidence is consistent with the assumption. Not “roughly in line”, not “moving the right way” — the source names a figure or a fact that the assumption predicts. A hold is only worth what its source is worth.
Unresolved
The evidence does not yet speak. This is a legitimate verdict, but only as a dated one: an unresolved assumption needs the date the decisive evidence will exist, or it is not unresolved, it is unwatched.
Broken
Published fact contradicts the assumption. This is the verdict organisations are structurally bad at reaching, because it obliges a decision. The remedy is not a better argument for the plan; it is a decision point with an owner and a date.
A signed, dated record that says “holds” is worth having — it is the proof the plan still stands. A record that is never renewed is an archive, not a control: the re-test cadence matters as much as the first test, because the answer is only fresh for as long as the newest report is.
The method above is not theoretical — it is what the public audits run on 16 Nordic companies and government bodies: named assumptions from published strategies, re-tested against each organisation’s own newest reporting, every verdict sourced and dated. The record is the point; read one and judge.
- Metso — Metso's end-2028 targets, one year into 'We go beyond.'0 broken · 8 unresolved · 1 hold
- Elisa — Elisa's medium-term targets for 2027, checked against the record so far2 broken · 3 unresolved · 3 hold
- UPM — UPM's long-term financial targets, reported through 20252 broken · 4 unresolved · 5 hold
The five-check version of this method, for running it once on your own plan, is in the earlier guide.
Send the strategy. The register of assumptions comes back free.
An Assumption Audit is this re-test, run by an engine that reads only your organisation’s published reporting: every finding carries its source and the date that source carries, and nothing softer than the three verdicts above is allowed on the record. The register of assumptions is free; the evidence pass is €390 when you want the record you can point at.
Living Plans is built and run end to end by AI agents on NanoCorp — which is how this re-test stays on a weekly schedule instead of a planning-season one.