What we refuse, and what we guarantee
A classification is only worth buying if it cannot be bought. This page exists so that a client, a court or a journalist can check what the constraints are before deciding whether to believe anything produced here.
Work that will not be accepted
- Commissions to characterise an organisation for use against it in an election campaign, an advertising campaign or a lobbying push.
- Any engagement where the client specifies the conclusion, or where payment is conditional on a particular finding.
- Work for an organisation seeking a favourable coding of itself.
- Anything requiring an opinion on a legal finding — persecution, terrorism designation, apartheid — rather than on political position.
The first of these is the important one. The moment a campaign can commission a characterisation of its opponent, every other client loses the reason to trust the output — and those other clients are the ones whose work actually depends on the analysis being independent. Refusing that work is not a moral flourish. It is the thing that makes the rest of it worth anything.
Enquiries are asked to state the intended use at the point of briefing. An enquiry that declines to say, or whose stated use changes after a quote has been given, will not be taken forward.
What is guaranteed
| Turnaround | If a stated deadline is missed for any reason within our control, 25% of the fee is refunded. If it is missed by more than half the quoted turnaround again, the fee is refunded in full. |
| Revisions | One round of revision is included, within the original scope. Changes that alter the question being answered are quoted separately. |
| Confidentiality | Commissioned work is never published, and the subject is never added to the public atlas, without your written consent. The fact that a commission took place is itself confidential. |
| Evidence | Every score carries a rationale and at least one artefact you or anyone else can go and check. If you dispute a score, say so and it will be re-examined against the evidence. |
| Independence | The conclusion is not negotiable and is not for sale. If the measurement does not support the position you hoped for, the report says so. |
What the instrument cannot do
These limits are not commercial caveats. They are published on the method page, they predate every instruction, and they appear in the reports themselves.
It is not objective
No coding scheme is. A number rendered in monospace still came from a judgement. What you get is an inspectable claim with the evidence attached, not a certainty.
It is not peer reviewed
It is a synthesis with published evidence, not an academic dataset. Where you need an authoritative figure, cite Chapel Hill, the Manifesto Project or V-Dem.
It has one coder
So it cannot report inter-coder reliability, which is the standard evidence that a scheme works in other hands. This is stated on the method page and there is an instrument on the site that lets you test the codebook yourself.
It does not settle legal or moral questions
Whether conduct meets a legal definition is for a tribunal on the evidence. A coordinate is a description, and distance measures similarity, not merit.
Every one of these limits is already public. They were not written for this page.
Method and limits