Method and trust

How we decide what to believe, and how we protect what you tell us.

An assessment is only useful if you know how it was reached. This page explains how we grade information and the safeguards around our work.

Grading every report

We use the Admiralty system, the two-part grading scale used by NATO and many intelligence services. Every item gets a letter for the source and a number for the information, judged separately.

How reliable is the source?

  1. ACompletely reliable
  2. BUsually reliable
  3. CFairly reliable
  4. DNot usually reliable
  5. EUnreliable
  6. FReliability cannot be judged

How credible is the information?

  1. 1Confirmed by other sources
  2. 2Probably true
  3. 3Possibly true
  4. 4Doubtful
  5. 5Improbable
  6. 6Truth cannot be judged

Example. A report graded B2 comes from a source that is usually reliable, and the information is probably true but not yet confirmed independently. You would treat it seriously, and watch for confirmation before acting on it alone.

Ten stories, one source

A claim repeated by many outlets is not necessarily confirmed. Often they are all copying the same original.

PRISM IE compares reports to tell the difference between outlets repeating the same text and outlets that reached the same conclusion independently. Only independent sources count towards confirmation. The same checks help us spot coordinated campaigns, where many accounts push identical wording.

Our safeguards

Public sources only

We work with publicly available information. We do not access private accounts, intercept communications or use deception to obtain data.

Analysts check the machine

Automated grades are reviewed by a trained analyst before they reach a client. Where we are uncertain, the assessment says so.

Separate environments

Each client’s work runs in its own environment with its own audit trail. Government and corporate work are kept fully apart.

On your servers if needed

For clients who cannot send requirements outside their own network, the platform runs on their infrastructure with no outside connection.

Data protection

We sign a data processing agreement with every client and handle personal data under the data protection law that applies to the engagement. Read our privacy notice.

Discretion

We do not name clients without their written consent, and we do not discuss one client’s work with another.

Questions about our method?

We are happy to walk your team through it in detail.