← Back to dashboard 🔒 Premium — subscriber access

Predictive Backtest — Did Our Signals Anticipate Designations?

Generated 2026-07-28T03:16:25+00:00

Point-in-time honest: a signal counts only when the linked party was already sanctioned by some jurisdiction at the time of the court case / ownership stake, and that predates the entity's own designation. Measures recall + lead time, not precision. Channels: co-litigation, ownership (≥50% and 10-50%), shared directorship, and procurement (contract signed after the counterparty's designation). Anchors: OFAC + EU/UK/UA/CA/AU/JP/CH/NZ.

Two distinct claims — do not conflate them:

  1. Improvement (apples-to-apples): adding the procurement + minority-ownership channels lifts recall from 33.4% → 41.5% (legacy 3 channels vs all 5, on the same court-present cohort of 6,070, since 2022-01-01). Channel change only.
  2. Current-state recall: 40.8% (2,564/6,283) on the broadened court∪procurement denominator — the honest present-day figure. Slightly below 41.5% because procurement reveals more entities than its signals always anticipate (mild, expected dilution). Do NOT compare this to the legacy 33.4% — that would conflate the channel change with a denominator change.

Recall by jurisdiction (court-present, since 2022-01-01)

Jurisdiction Present Anticipated Recall Median lead p75 lead
CH 93 72 77.4% 2632d 3304d
EU 352 252 71.6% 2590d 3198d
UK 268 161 60.1% 1777d 2991d
OFAC 2,550 1,233 48.4% 2152d 2790d
UA 4,892 2,193 44.8% 1883d 2661d
any (first designation) 6,070 2,517 41.5% 1828d 2621d

🔒 Premium brief — full access

This brief's evidentiary detail, specific targets, and designation recommendations are available to vetted compliance teams, law firms, and government analysts — as single-brief purchases or under a portfolio monitoring retainer. The teaser above shows the headline methodology; the full document includes case-number citations, target INNs, SDN-adjacency numbers, and recommended secondary-designation scope.

Already subscribed? Open the full brief with your analyst credentials. Otherwise, subscribe to portfolio monitoring for the full library, or commission this brief on its own.

Read the full brief → See access options Request by email
Basis and limitations. Findings in this brief are derived from public records — kad.arbitr.ru arbitration filings, EGRUL registry extracts, zakupki.gov.ru procurement contracts, and official sanctions lists — and from algorithmic scoring described in the methodology. Scores are triage outputs, not legal determinations. Court co-litigation measures litigation frequency with listed parties, not culpability; procurement links may reflect arm's-length commerce. Named entities may dispute factual errors via derek.linz@linzalytics.com (subject "Dispute request"); the review process and timelines are on the methodology page, and material corrections are logged at /intel/corrections.html with dates.