Evidence tiers
Every record carries a tier. It describes what kind of evidence the record rests on, and it decides how strongly that record may be stated anywhere on this site.
Where space allows, tiers are written out in full, and they are never colour-coded. A T6 record records what an organisation said, and when. Colouring it red would add a judgement that the tier does not make.
| Tier | What it is | May never support |
|---|---|---|
| T0 Worked out here | Calculations made for this site, from stated inputs, with every step and unit shown so you can redo them. | A result resting on an unstated estimate. Where an input is estimated, it says so. |
| T1 Measured | A real measurement, published where other specialists could object to it, with the conditions and the error bars given. | Carrying the result across to a different size, frequency or environment. The result holds for the conditions tested. |
| T2 Modelled | Analysis, simulation or theory that other specialists have reviewed. Useful for how quantities relate to each other. | Showing that a thing works. That needs a measurement. |
| T3 A public body published it | A report from a government or agency, with the authors named and the method set out. | Telling you the current position once it has aged. Programmes change faster than the reports about them. |
| T4 Commissioned study | A commissioned study or industry paper that discloses how it reached its answer. Who paid is recorded every time. | Being cited as independent of its funder. |
| T5 An expert said it | Conference talks, interviews, evidence to committees, specialist press. Good for who thinks what, and which way opinion is moving. | Supplying a headline figure. If a number matters, it needs to come from something measured or modelled. |
| T6 A company said it | Press releases, corporate pages, marketing, recruitment adverts, unattributed news. | Establishing that a result was achieved or a cost met. It records what was said, and when. |
How the tiers are ordered
T0 is listed first because its inputs, arithmetic and failure conditions are published beside the result, so it can be checked here. For the other tiers the check runs through the source document.
How the tier changes what may be said
| Tier | How a record may be stated |
|---|---|
| T0 to T3 | As a statement, with the record and its date reachable. |
| T4 | Attributed only: "Study X, commissioned by Y, models ..." |
| T5, T6 | Reported only: "Organisation X stated on date D that ..." |
When a record rests on several sources
Its tier is the weakest of them, and the detail of which source contributed what is printed beneath. A record resting on a peer-reviewed measurement and a press release is treated as resting on the press release.
That applies to what a record rests on. A source cited only to corroborate - one supplying nothing the record asserts, and agreeing with a method or a figure taken from elsewhere - does not set the floor. The record says which is which in its own words, beside each identifier, so the distinction is one a reader can check rather than one taken on trust.
Tier is not confidence
They are recorded separately, because a strong paper can be weak evidence for a question it did not set out to answer. Confidence is HIGH, MED or LOW, and a LOW-confidence record is never stated on this site at all. It may only be mentioned as an open question.