Evidence · Methods

The method is the product. The findings are downstream of it.

Anyone can publish a number. The question a reader should be able to answer in ten seconds is: what kind of claim is this, who says so, and when was it true?

The three tags

[A]

Verified against a primary source.

Earns the tag: A primary source we can name, retrieve and cite with a date: a statute, an audited account, a filing, a micro-data file we hold.

Never: A number reported by a newspaper quoting an official. That is a [B] about the newspaper's sourcing, not an [A].

[B]

An organisation's own claim, reported as such.

Earns the tag: An organisation's own claim about itself, reported as such and attributed by name — company projections, ministry press releases, investor decks.

Never: Aggregation into a headline total without the tag travelling with it. A sum of [B]s is a [B], never an [A].

[C]

An estimate, including our own arithmetic.

Earns the tag: An estimate, including our own arithmetic, with the formula shown inline and the assumptions stated in the same paragraph.

Never: Rounding into a cleaner number, or reappearing in a later paper as an input without its tag. That is laundering, and it is the failure mode this standard exists to prevent.

Standing rules

Six rules that outrank any deadline.

01No figure without a tag, a date and a source→

A number that cannot carry all three does not appear on this website, in a paper, or in a slide shown at a convening. There is no verbal exception for launches.

02No silent amendment→

Every change to a published figure produces a version bump and a log entry stating what was wrong and what it now says. The previous version stays retrievable.

03The tag travels with the number→

When a figure is reused in a later publication, its original tag comes with it. A [C] cannot become an [A] by changing document.

04One definition, frozen across years→

A comparable series is worth more than a precise one. Where a definition must change, the whole series is restated and the break is marked in the chart.

05Colour never carries meaning alone→

Every chart series is distinguished by shape and dash pattern as well as colour, and every tag has a text label, so the evidence survives being printed in greyscale.

06Methods publish before findings→

If we cannot describe how a number was produced well enough for a critic to rebuild it, the finding waits. This has delayed two publications so far.

Playbooks

The rules exist as documents, not as culture.

Culture leaves with the person who held it. Each rule above traces to a numbered playbook with an owner, a review date and a change history.

P02 · Independence and conflicts of interest: The unamendable clause, the register, and who may stop a publication.
P06 · Funding, disclosure and concentration caps: Bands, caps, the refusal log, and the annual disclosure format.
P09 · The confidence-tag standard: What earns [A], what is only ever [B], and why [C] must never be laundered.
P11 · Corrections and versioning: Discovery to publication, the severity ladder, and the no-silent-amendment rule.
P14 · Programme charter and kill gates: Named owner, stated kill criteria, quarterly review by the Telos Council.
P18 · Partnership and endorsement limits: Why there is no logo wall and no testimonial on this website.

Test it

The standard is only worth what a critic can do with it.

If you can rebuild one of our series and get a different answer, we want the working. It goes in the log either way.

Corrections log