Skip to main content

Analytical Standards

Intelligence Tradecraft, Applied

Most geopolitical commentary doesn't distinguish between an outcome that is 30% probable and one that is 80% probable. That gap matters in decisions. Beurling applies professional intelligence tradecraft to every report: calibrated probability language, evidence lineage, daily delta selection, forecast tracking, and compact challenge review. This page documents the standards behind every assessment.

ICD 203 — Intelligence Community Directive

ICD 203 is the analytical standards directive issued by the US Director of National Intelligence, mandating structured analytic techniques across all 17 US intelligence agencies — CIA, DIA, NSA, and others. It was created in the aftermath of 9/11 and the Iraq WMD failure: both were failures of analytical process, not raw information. Equivalent standards govern the UK Joint Intelligence Committee, NATO, and Five Eyes partners.

The directive requires four things from every key judgment:

  • Explicit confidence levels attached to every assessment
  • Alternative hypotheses considered and documented
  • Key assumptions stated and flagged for review
  • Observable warning indicators identified in advance

Beurling applies these requirements to every report cycle. Where the evidence base is thin, we say so. Where a judgment rests on assumptions that could be wrong, we name them.

7-Level Likelihood Scale

Every key judgment uses standardised probability language. "Likely" means the same thing in every report. Probability and confidence are two separate dimensions — a judgment can be highly probable but rest on thin sourcing.

Term Range
Almost certainly >90%
Very likely 75–90%
Likely 60–75%
Roughly even odds 35–60%
Unlikely 15–35%
Very unlikely 5–15%
Remote <5%

Confidence levels (independent of probability)

High confidence

Multiple independent corroborated sources; minimal collection gaps

Moderate confidence

Some corroboration; reliable sources; incomplete picture

Low confidence

Thin sourcing; unverified reporting; significant collection gaps

Evidence Lineage and Review

Beurling does not promise a fixed six-pass diagnostic bundle every cycle. The methodology is built around source provenance, daily change detection, forecast calibration, and a single compact challenge review that checks the assessment before synthesis.

Evidence lineage

Every judgment is tied back to the source trail that produced it: publication, source family, extracted claim, and the downstream assessment that used it. That lineage stays visible so readers can trace why a conclusion moved.

Daily delta

Each run prioritises what changed since the prior cycle: new sources, new claims, forecast movement, resolved assumptions, and meaningful gaps. Repeated wires only matter when they change the evidence state.

Forecast tracking & calibration

Standing questions turn the assessment into a calibration loop. Probabilities are updated deterministically where possible, and resolution history is retained so Brier scoring and replay can measure whether the system is well calibrated over time.

Compact challenge review

Instead of a six-stage ceremony, each report goes through one focused challenge layer that names the fragile assumptions, strongest counterevidence, missing support, and any confidence cap that follows from the evidence.

Source Standards

Every report discloses its sources, classified by publisher family. Source families carry different default reliability assessments that inform how heavily any single source drives a judgment.

Official / Government

Statements from ministries, defence departments, heads of state. High reliability for declared positions; requires corroboration for capability or intent claims — official statements are also information operations.

Established Press

Wire services and established outlets with editorial standards and correction policies. High reliability for factual reporting; assessed with attention to national framing and sourcing transparency.

Think Tanks & Research

Policy institutes, university research centres, and analytical organisations. High reliability for structural analysis; assessed against known institutional biases and funding sources.

Regional & Local Press

Local and regional outlets with proximity to events. Variable reliability; high value for ground-truth detail unavailable to international press; cross-referenced before use in key judgments.

Adversarial / State Media

State-affiliated media and outlets with documented disinformation mandates. Included deliberately — adversarial framing is analytically significant. Flagged and assessed against independent corroboration.

Financial & Market Data

Pricing data, market movements, financial disclosures. High reliability for revealed preferences and economic signals; integrated in topics where markets serve as leading indicators for political developments.

Source count and family diversity are reported for each assessment. A judgment resting on a single source family carries lower confidence than one corroborated across three or more independent families.

Judgment Tracking

Every key judgment is assigned a stable identifier at publication. When events develop, judgments are formally resolved with one of three outcomes:

  • Confirmed The predicted development occurred as assessed.
  • Falsified The prediction did not materialise.
  • Superseded The analytical framing was updated before the event resolved. Excluded from directional accuracy calculations.

Subscribers on the relevant topic receive an email notification when a judgment resolves. The scored outcome remains attached to the underlying assessment record so it can be reviewed alongside the original analysis.

See the standards in practice

Every Beurling report applies these standards: likelihood estimates with numeric ranges, confidence levels, source lineage, forecast calibration, and a compact challenge review integrated into the assessment.

Beurling is an independent intelligence platform. All analysis is open-source — no classified sources are used or claimed. For questions about methodology or to report an analytical error, use the contact page.