04 / METHODOLOGY
How we work.
Every analysis separates what is verified from judgment, and states where the evidence is weak.
Five epistemic labels
Every claim is assigned a label indicating its analytical status.
Probability bands
Fixed bands: very low 5–20 · low 20–40 · medium 40–60 · high 60–80 · very high 80–95. They are J&H judgments, not statistical projections. Confidence measures the strength of the evidence, not the likelihood that the scenario materialises.
Sources and asymmetries
We state where the evidence comes from and which positions were not consulted. Media outlets aligned with a particular actor are used only for documentable data, with their editorial line declared; foreign outlets are used for coverage, never for verification. Each source carries its level: primary, partly primary, secondary, referred or pending cross-check.
What we leave out
- Unverified leaks.
- Social-media posts without independent verification.
- Advocacy figures used as data (kept as attributed claims).
- Influence-campaign content: actively searched for, with the result declared.
- The framing or qualification of any source; the underlying data is used.
Limits
Every document states its cut-off date. We do not attribute facts or intentions to actors. “Unclassified” describes the document; “restricted use” describes its circulation.
Use of artificial intelligence
Example, Ceuta 2026: the analysis was carried out with the support of artificial-intelligence tools for searching, organising and checking open sources, under the editorial direction of the responsible person, who set the criteria, corrected the content and approved the final version.