Standards

Methodology

AAT Intelligence distinguishes between documented facts, estimates, assumptions and analytical conclusions. This page sets out how that distinction is maintained, so a reader can check the work rather than take it on trust.

A claim, a calculation and a conclusion are three different things. Most technology coverage presents all three in the same voice. Every AAT Intelligence piece marks which is which.

The editorial process

Every piece moves through five stages, in order:

  1. Observe — identify a development where the public account and the underlying evidence may differ.
  2. Research — assemble the primary record: filings, technical documentation, regulatory instruments, published specifications, academic literature.
  3. Verify — establish which claims the record supports, which it contradicts, and which it leaves open.
  4. Analyze — apply a stated standard, usually the subject’s own published targets or a defined comparison set.
  5. Explain — publish the finding together with the reasoning, the sources and the limits.

Primary-source preference

Analysis is built from primary documents wherever one exists: regulatory filings, government instruments and gazettes, company disclosures, technical specifications, standards documents, peer-reviewed papers and datasets published by the body that collected them.

Secondary reporting is used to locate a primary document, not to substitute for it. Where a figure originates in another publication and no primary source could be obtained, the piece says so and attributes it.

A press release is a primary source for what a party announced. It is not a source for whether the announced thing exists, is funded, or is under construction. AAT Intelligence treats those as separate questions throughout.

Announced, contracted, operational

These three states are reported interchangeably across the technology press and they are not the same. AAT Intelligence keeps them separate in every piece and in every dataset:

  • Announced — a party has stated an intention. No commitment is implied.
  • Contracted — a binding agreement, filing or procurement record exists.
  • Operational — the capacity, facility or system is in service and that service can be evidenced.

Where a piece can establish only one of the three, it says which one. Merging them into a single headline number is the most common error in infrastructure coverage and it is one this publication treats as disqualifying.

Verification and the verification note

Every analysis carries a verification note directly beneath its key takeaways. The note states what the piece is built on and, critically, what could not be established. A reader should be able to judge the strength of a finding before reading the argument for it.

Where a claim could not be independently verified, the text says so at the point the claim appears — not only in the note. Phrases such as “reportedly” are avoided in favour of naming who claimed what.

Calculations

Where a piece performs a calculation, it states the inputs, the operation and the source of each input. Readers should be able to reproduce any figure AAT Intelligence publishes using only what is on the page.

Normalisation is chosen before the result is known, and the choice is stated. Comparisons across countries or operators are presented on a stated basis — per capita, per researcher, per unit of installed capacity — rather than as absolute totals, which are rarely comparable.

Estimates and uncertainty

An estimate produced by AAT Intelligence is labelled as an AAT Intelligence estimate wherever it appears, including in tables and charts. Estimates carry the method used to derive them and, where the inputs support it, a range rather than a point figure.

Where the available evidence does not support a conclusion, the piece says the question is open. An analysis that reaches no verdict is preferable to one that manufactures a verdict the record cannot carry.

Choice of comparison set

Comparative work states its peer set and why that set was chosen, before the comparison. Peer sets are selected for analytical usefulness, not rhetorical effect: a comparison whose result is obvious in advance carries no information.

Anonymous and confidential sources

AAT Intelligence prefers documents to conversations. Where a source speaks on condition of anonymity, the piece states the reason anonymity was granted and characterises the source’s position precisely enough for a reader to weigh it.

A claim resting on a single anonymous source is not published as an established fact. No source is granted anonymity to make an attack on a named individual.

Original reporting standards

The Original reporting category is reserved for information AAT Intelligence produced: a document obtained, an interview conducted, a finding established through direct examination. Aggregating or summarising other publications’ reporting never qualifies.

Where a piece reports something that has not appeared elsewhere, the basis for it is described as specifically as the source protection allows.

Use of AI tools

AI tools may be used for research support: locating documents, summarising long filings for triage, checking arithmetic, and drafting assistance on non-substantive passages.

They are not used to generate factual claims, and no figure enters a published piece on the authority of a model output. Every fact in an AAT Intelligence piece traces to a human-verified source. Analytical conclusions are the author’s.

Corrections

Errors are corrected promptly and visibly. A correction states what was wrong, what is now correct, and when the change was made. Substantive corrections are noted on the piece itself and listed on the corrections page. Pieces are not silently edited after publication; typographical fixes are the only exception.

Conflicts of interest

Any financial interest, professional relationship or prior affiliation bearing on a subject is disclosed within the piece itself, not only on this page.

Where the founder has previously worked with or for an organisation under examination, that relationship is stated in the piece. Prior affiliation does not preclude coverage; concealing it would.

Commercial relationships

Editorial conclusions are not for sale. Should AAT Intelligence introduce sponsorship, commissioned research or institutional subscriptions:

  • Sponsored material is labelled as sponsored and is visually distinct from editorial work.
  • Sponsors receive no advance sight of editorial content and no right of review.
  • Commissioned research states who commissioned it, within the work itself.
  • No commercial relationship confers influence over whether a subject is covered, or how.

Data methodology

Datasets published under AAT Data carry a collection date, a per-row source, an explicit separation of announced from contracted from operational values, a statement of coverage limits, and a downloadable CSV matching what is shown on the page.

Where a dataset is revised, the revision is dated and the prior version remains available.

Independence

AAT Intelligence is independent and unaffiliated. It examines specifications, evidence and stated commitments; it does not advocate for or against any organisation, government or programme. A finding that a stated target has not been met is an observation about the record, not a position on the party that set it.

Questions

Questions about the method behind a particular piece are welcome and will be answered. Write to contact@aatintelligence.com, or see editorial standards for the conduct rules that sit alongside this page.