Methodology
We operate in one of the most disinformation-saturated environments in the world. Our answer is process, not confidence theater: every event is decomposed into atomic claims, every claim is scored against its evidence, and every published item carries its uncertainty label. We never present a conflict party's statement as established fact — regardless of which party it is.
Source classes
Published legal acts (EUR-Lex, OFAC, OFSI), official registers (Prozorro, TED), regulator and sensor data (ENTSO-E, NASA FIRMS). Facts about their own content — e.g. a sanctions listing — are treated as established.
Global news agencies and outlets with strong editorial standards and correction practices (AP, BBC, FT).
Reputable regional media and independent outlets (Kyiv Independent, Meduza, The Moscow Times, Notes from Poland). Solid, but treated as one voice, not final confirmation.
Government communiqués of conflict parties, state agencies, well-tracked OSINT channels. Usable with attribution, never alone.
Anonymous Telegram/X accounts, unverified footage. Can only open an event in 'claimed' status — never publish on their own.
Claim statuses
Confidence is computed deterministically from source quality, evidence strength, independent confirmation (tracing the origin graph — ten portals quoting one Telegram post count as a single source), temporal and geographic verification, claim specificity and the source's historical accuracy, minus red-flag penalties. Raw scores are internal; you see the labels:
Authoritative source or at least two independent strong confirmations; no serious contradictions.
Strong evidence with at least one independent confirmation; some details still uncertain.
Consistent information from a credible source; no full independent confirmation yet.
Reported by a specific party; insufficient evidence to confirm.
Credible sources present conflicting versions.
Confirmed wrong date, place, material or interpretation.
The source withdrew the information or our correction removed it.
AI and human review
AI handles collection, deduplication, claim decomposition and drafting. Publication of high-risk content — casualty figures, attribution of weapons or responsibility, legal interpretations of sanctions, market-moving statements about specific companies — always requires human approval. Every automated decision stores an auditable explanation.
Safety policy
We do not publish precise troop positions, movement routes, planned transport timings or any material that could serve targeting. Content matching these criteria is blocked from automatic publication.
Sanctioned media policy
We do not redistribute or monetize content from sanctioned Russian state media (RT, Sputnik and related entities). Where their narratives are analytically relevant, we reference them as objects of analysis with clear attribution, without amplifying the content itself.
Corrections
When we get something wrong, the item is marked False / Misleading or Retracted — visibly, in place, with the original preserved for transparency. Retracted events remain accessible with a prominent notice. To report an error, email corrections@easterneuropeimpact.com.
What we are not
Eastern Europe Impact provides risk information for business decision-makers. It is not legal advice, not investment advice, and not a substitute for sanctions counsel. Always verify against the authoritative legal acts before acting.