Multi-Jurisdiction Sanctions Coverage & Data-Quality Analysis — Case Study A

A reproducible analytical dataset, pipeline and dashboard for practising sanctions-screening coverage analysis, cross-list data quality, and financial-crime data governance across the four primary sanctions regimes — OFAC SDN, EU FSF, the UK Sanctions List, and the UN Security Council Consolidated List. Designed to be analysed in SQL, Python (pandas), Power BI or Tableau, and read as an intelligence brief.

ℹ️ Real where it counts, modelled where it must be — and labelled throughout. Per-regime designation counts, update cadence and version strings are sourced live from OpenSanctions (v.20260517074701-kjc) and are independently citable. The entity-level cross-list resolution, data-quality flags and designation-velocity fields are modelled and calibrated to those live marginals — the exact entity-resolution join requires the licensed OpenSanctions bulk export. No real designated-person names or PII are reproduced; target keys are synthetic. Data © OpenSanctions Datenbanken GmbH, licensed CC BY-NC 4.0.

---

Project overview

A fintech screening or financial-crime data-governance team needs to know:

1. Are we screening the right lists? (coverage)
2. How good is the data inside them? (quality)
3. What would single-source screening miss? (blind spots)
4. Where is risk concentrated? (jurisdictions, typologies, transliteration)
5. Which programmes are moving fastest? (velocity → refresh cadence)

This repository models the inputs to that workflow. It is sized for a portfolio — small enough to explain end-to-end in an interview, real enough that the analytical findings stand up to scrutiny.

---

Data model

Four live regimes are consolidated by a deterministic pipeline into a single resolved analytical table at unique-target grain, where each list-membership column reconciles exactly to that regime's live target count:

OFAC SDN  (19,678) ┐
EU FSF     (5,900) ├─► build_case_study_a.py ─► consolidated_sanctions_4regime.csv  (24,183 rows × 18 cols)
UK List    (5,986) │                            analysis_stats.json                 (computed metrics)
UN SC        (999) ┘
        32,563 raw designation records  ─►  24,183 unique targets after resolution

File	Rows	Grain
consolidated_sanctions_4regime.csv	24,183	One row per unique designated target across the four regimes
analysis_stats.json	—	Pre-computed metrics (coverage, overlap, data-quality, velocity) emitted by the pipeline

Schema — consolidated_sanctions_4regime.csv

Identity

• target_id — synthetic stable key (NK-000000…). Not a real OpenSanctions identifier.

List membership — drives coverage & overlap analysis

• on_ofac_sdn, on_eu_fsf, on_uk_fcdo, on_un_sc — boolean; is the target designated on that regime's list
• list_count — integer 1–4; number of the four lists carrying the target
• coverage_class — single-list / multi-list
• sole_list — for single-list targets, which regime carries them; otherwise multi

Classification

• schema — entity type: Person / Organization / Vessel / CryptoWallet / Aircraft / Security
• typology — programme theme: Russia/Ukraine · Counter-terrorism · Narcotics/TCO/Trafficking · Iran/DPRK proliferation · Cyber · Human rights/Corruption · Other regional
• country — primary jurisdiction of risk
• added_year — year first designated (2018–2026); drives the velocity analysis

Data-quality flags — drive the remediation / governance analysis

• dq_missing_dob, dq_name_translit, dq_mononym, dq_cross_source_conflict, dq_stale — boolean defect flags
• dq_any — boolean; target carries ≥1 material defect

---

What's in the data

The dataset is calibrated so the analytical patterns match the Case Study A findings:

• 32,563 raw designation records resolve to 24,183 unique targets
• 77.5% of targets are single-list-only; 22.5% appear on 2+ lists; 650 are on all four
• Best-case single-list gap = 18.6% — the share missed even by the largest list (OFAC SDN); UN-only screening misses 95.9%
• 12.0% of targets carry ≥1 data-quality defect — cross-source conflict (5.1%), transliteration risk (3.3%), missing DOB (2.8%), mononym (1.0%), stale record (0.4%)
• Entity-type mix: Person 12,072 · Organization 9,573 · Vessel 1,507 · CryptoWallet 702 · Aircraft 322 · Security 7
• Fastest-growing typologies (2022→2025): Cyber +144%, Narcotics/TCO/Trafficking +103%
• Transliteration risk concentrates on Russia, Iran, DPRK and other non-Latin-script jurisdictions

---

How to use

# Python (pandas)
import pandas as pd
df = pd.read_csv("consolidated_sanctions_4regime.csv")

df["coverage_class"].value_counts(normalize=True)      # single- vs multi-list share
1 - df["on_ofac_sdn"].mean()                           # best single-list coverage gap (~0.186)
df.groupby("schema")["dq_any"].mean().mul(100).round(1)# data-quality defect rate by entity type

-- SQL (SQLite / Postgres / BigQuery — CSV-loadable)
-- Single-list blind spots: how much of the universe sits on exactly one list?
SELECT coverage_class,
       COUNT(*)                                          AS targets,
       ROUND(100.0 * COUNT(*) / SUM(COUNT(*)) OVER (), 1) AS pct
FROM consolidated_sanctions_4regime
GROUP BY coverage_class;

-- Remediation backlog: defect rate by programme typology
SELECT typology,
       ROUND(100.0 * AVG(CASE WHEN dq_any THEN 1 ELSE 0 END), 1) AS defect_rate_pct
FROM consolidated_sanctions_4regime
GROUP BY typology
ORDER BY defect_rate_pct DESC;

Load consolidated_sanctions_4regime.csv into Power BI / Tableau and slice by coverage_class, typology, country, added_year, and the dq_* flags. The boolean on_* columns drive coverage and overlap visuals directly.

---

Reproduce

pip install -r requirements.txt
python build_case_study_a.py      # regenerates consolidated_sanctions_4regime.csv + analysis_stats.json

The pipeline is deterministic (fixed random seed), so a re-run reproduces the dataset and every figure above exactly.

---

Methodology note

Per-regime target counts, entity-type splits, programme taxonomies, update cadence and version strings are sourced live from the OpenSanctions Consolidated Sanctions catalogue (May 2026) and are individually citable. The cross-list resolution layer (which targets sit on more than one list), the data-quality defect incidence, and the designation-velocity series are modelled — calibrated to the live marginals and to documented OpenSanctions resolution behaviour, and reported as planning estimates rather than a measured join. The build is seeded and regenerable. No real designated-entity names or PII are reproduced. Data © OpenSanctions Datenbanken GmbH, licensed CC BY-NC 4.0; commercial use of the underlying sanctions data requires a licence from OpenSanctions.

---

Companion artefacts (included in this repo)

• Sanctions_Intelligence_Dashboard.html — interactive Chart.js / Grid.js dashboard (coverage, overlap, velocity, data-quality, jurisdictions, entity-type mix)
• Sanctions_Intelligence_Report.html — narrative intelligence report version of the same analysis
• build_case_study_a.py — the reproducible pipeline that builds the dataset

About

Part of a financial-intelligence portfolio by Bukola Apampa (Financial Intelligence Research Analyst) demonstrating sanctions-screening coverage analysis, cross-list data quality and AML/financial-crime data governance in SQL, Python and Power BI. Companion to pep-screening-governance-synth.

Suggested topics: portfolio · sql · power-bi · pandas · data-quality · aml · sanctions · financial-crime · data-governance · opensanctions