The Denied-Party Screening Data Model: Ingesting OFAC SDN and the BIS Entity List
What changed: GingerControl does not offer denied-party screening: the claims about screening, the retired AI Integration service, the old API name, the free audit and Compliance Radar as private beta are corrected, and the data-model guidance stands on its own.
The denied-party screening data model is the record schema, ingestion pipeline, refresh cadence, and fuzzy-match key design that turns the raw OFAC Specially Designated Nationals (SDN) List and the BIS Entity List into a queryable, audit-defensible system of record. It is not a search box; it is the one-to-many table structure (primary party, aliases, addresses), the daily refresh job, and the match logic that every order, vendor, and consignee is scored against. GingerControl is a trade compliance AI platform that helps importers, exporters, and customs brokers classify products, simulate tariff costs, and track policy changes; it does not screen counterparties, and this article maps the data model a screening system needs.
In this post
- Why does the screening data model matter more than the screening tool itself?
- When the notice arrives: what the enforcement reviewer asks for first
- What fields does the OFAC SDN List actually contain?
- What fields does the BIS Entity List contain, and how is it structured?
- How do you ingest and refresh these lists, and on what cadence?
- What are the right fuzzy-match keys, and why does exact matching fail?
- How do you make screening a shared system of record across the enterprise?
- Frequently asked questions
- Wiring screening into your system of record
- References
Why does the screening data model matter more than the screening tool itself?
Because the data model is what fails when an enforcement notice arrives. A clean tool query against a stale or misjoined dataset will return "no hit" on a party that was added to the BIS Entity List last Tuesday, or on an alias your schema dropped on ingest. The data model, not the UI, determines whether your screening is a real control or a checkbox a BIS or OFAC reviewer will dismantle.
TL;DR: The notice has already landed. A shipment is detained, an administrative subpoena cites a counterparty you "screened," or your bank flags a wire to an entity you cleared last quarter. The first question from counsel is not "do you screen?" It is "show me the exact list version you screened against, the fields you matched on, and the timestamp." For an export compliance team running 5,000 to 50,000 counterparties across order entry, vendor onboarding, and consignee checks, the denied-party screening data model is the difference between producing that record in an hour and discovering your screening never ingested aliases at all. GingerControl does not sell denied-party screening; this guide lays out the data model to demand from whichever screening platform or in-house build you use to bring OFAC SDN, the BIS Entity List, and the broader Consolidated Screening List into one screening system of record. Unlike a standalone screening widget that returns a yes/no against an opaque dataset, a sound data model preserves the matched fields, the list version, and the timestamp as an audit-ready record. This article is research and program guidance, not legal advice; the importer or exporter and their counsel decide and file.
When the notice arrives: what the enforcement reviewer asks for first
Picture the sequence most teams live through. A container is held at the port, or a BIS outreach letter references a transaction with a party now on the Entity List, or your treasury team gets an OFAC inquiry about a blocked payment. The clock starts. And the questions that follow are all data-model questions, not policy questions:
- Which list version did you screen against, and when? (Not "we use a screening tool", the actual file date.)
- Did you screen the party plus its aliases and addresses, or just the exact name on the purchase order?
- Can you show the match score and the fields it matched on for the cleared decision?
- Did your screening account for the OFAC 50 Percent Rule, where an entity owned in aggregate by blocked persons is itself blocked even though it never appears on the SDN List?
If your screening is a person pasting names into a public search page, none of those answers exist as a record. The screening happened, but it left no system of record. The program-design and governance side of this, match-threshold tuning, escalation, and who owns the false-positive queue, is its own discipline; we cover that in the denied-party and sanctions screening program design guide and the foundational restricted-party and denied-party screening guide. This article is the layer underneath both: the actual fields and schema.
The pattern underneath all of this is that denied-party screening fails at the data layer, not the policy layer. The OFAC SDN List is four joined files, not one. When a program never ingests the ALT alias file or the ADD address file as separate one-to-many tables, a party can clear on the primary name while matching a dropped alias. The screening was technically run and substantively useless, and the enforcement record will show exactly that gap.
What fields does the OFAC SDN List actually contain?
This is where most teams are surprised. The SDN List is not a flat spreadsheet of names. In the legacy delimited (CSV) format, it is a set of joined files in a one-to-many relationship, keyed on a unique entry number (ent_num), which is the primary key that links a listed party to all of its addresses and aliases.
The four SDN files and their fields
| File | Role | Key fields |
|---|---|---|
| SDN.CSV | Primary party record (one row per listing) | ent_num (UID), SDN_Name, SDN_Type (individual / entity / vessel), Program (sanctions program), Title, Call_Sign, Vess_type, Tonnage, GRT, Vess_flag, Vess_owner, Remarks |
| ADD.CSV | Addresses (many per party) | Ent_num (foreign key), Add_num, Address, City/State/Province/Postal Code, Country, Add_remarks |
| ALT.CSV | Aliases (many per party) | ent_num (foreign key), alt_num, alt_type (aka / fka / nka), alt_name, alt_remarks |
| SDN_COMMENTS.CSV | Remarks overflow beyond the 1000-character limit | ent_num, comment text |
Two design facts from OFAC's own data specification and file-format documentation govern any correct ingest:
- The relationship is one-to-many. Primary SDN record
117is associated with every address row carrying117in theEnt_numcolumn of ADD.CSV, and every alias row carrying117in ALT.CSV. If your ingest flattens this into a single "name" column, you lose the aliases and addresses, which is precisely where most evasion-driven matches live. - Null is encoded as
-0-, and the CSV uses comma delimiters with carriage-return record separators. A naive parser that treats-0-as a literal value, or that mishandles commas inside theRemarksfield, corrupts the record.
OFAC also publishes SDN_ADVANCED.XML, an advanced sanctions data model built on an internationally designed standard that breaks non-western names into discrete name parts, which materially improves transliteration matching. For an export compliance team screening counterparties across Cyrillic, Arabic, and CJK source documents, ingesting the advanced XML name-part structure rather than the flat name string is the single highest-leverage schema decision you can make.
GingerControl does not offer denied-party screening. Its Export Control Classifier classifies items under the EAR (one ECCN, or EAR99); screening the parties to a transaction against OFAC SDN, the BIS Entity List, the Denied Persons List and the Unverified List is a separate job, done by dedicated screening platforms such as Descartes Visual Compliance or Thomson Reuters.
What fields does the BIS Entity List contain, and how is it structured?
The BIS Entity List is a different animal. It is not a flat-file feed maintained on a daily mechanical schedule; it is a regulatory list codified in Supplement No. 4 to Part 744 of the EAR and amended through the Federal Register. Per 15 CFR 744.16, each listing carries five columns that your schema must capture as distinct fields, not collapse into "name":
| Column | What it holds | Why it matters to the data model |
|---|---|---|
| Country | Country associated with the listed entity or address | Drives geographic risk rules and joins to country-of-origin data |
| Entity | The entity name and address (often multiple addresses) | The match target; aliases and addresses are part of the listing |
| License Requirement | The specific license requirement for that entity | This is entity-specific; two entities can have different requirements |
| License Review Policy | The review presumption (e.g., presumption of denial) | Determines the outcome even if a license is sought |
| Federal Register Citation | The notice that added or modified the listing | Your provenance anchor for the audit record |
Two structural facts change how you ingest it:
- Additions, removals, and modifications are made by the End-User Review Committee (ERC), composed of the Departments of Commerce (Chair), State, Defense, Energy, and, where appropriate, Treasury. There is no fixed daily cadence; changes land when the Federal Register publishes them. Your refresh job must be event-driven against the Federal Register, not a once-a-day cron that assumes the list only changes overnight.
- The License Requirement and License Review Policy are per-entity fields. A screening data model that stores only "on the Entity List: yes/no" throws away the two fields that actually tell your team what the consequence is. The right schema stores the requirement and the review policy alongside the match, so the decision record explains itself.
GingerControl's free Policy Briefings email you when a U.S. trade rule changes, BIS included, and its Compliance Radar checks new trade notices against your HTS codes; neither screens your counterparties, so the event-driven refresh job in your screening pipeline is what catches an Entity List Federal Register notice the day it lands rather than at your next quarterly review.
How do you ingest and refresh these lists, and on what cadence?
Different lists, different clocks. This is the part program pages never specify, and it is where "we screen daily" quietly becomes false.
| Source | Authoritative refresh cadence | Ingestion implication |
|---|---|---|
| OFAC SDN List | Updated as changes occur (no fixed schedule); OFAC publishes file-format downloads | Poll for file-version changes; do not assume a fixed time |
| BIS Entity List | Amended via the Federal Register by the ERC (event-driven) | Trigger ingest on Federal Register publication, not a daily clock |
| Consolidated Screening List (CSL), trade.gov | Updated automatically every day at 5:00 AM EST/EDT | A reliable daily anchor that aggregates the others |
The Consolidated Screening List is the pragmatic backbone for most teams: it consolidates the export screening lists of the Departments of Commerce (BIS), State, and Treasury (OFAC) into a single feed, including the BIS Denied Persons List, Unverified List, Entity List, and Military End-User List; the State Nonproliferation Sanctions and AECA Debarred List; and the OFAC SDN List and several Non-SDN lists. It is offered in CSV, TSV, and JSON, the delimited download carries 27 columns with the field names in the first row, and a source column tells you which agency list each row came from. It also exposes an API with built-in fuzzy name search.
But the CSL is an aid, not a substitute for the primary lists. Two gaps your data model must close on its own:
- The OFAC 50 Percent Rule. An entity owned 50 percent or more, in the aggregate, directly or indirectly, by one or more blocked persons is itself blocked, even though OFAC does not publish those derivative entities on the SDN List. No list ingest catches them. Your data model needs an ownership/beneficial-ownership layer and a due-diligence workflow, because the name will never appear in SDN.CSV.
- Version pinning. Whatever you ingest, you must store the list version and timestamp with every screening decision. The cleared decision is only defensible if you can reproduce the exact dataset it ran against.
Quotable insight: A screening program that ingests only the Consolidated Screening List will pass an internal audit and fail an OFAC inquiry. The CSL refreshes at 5:00 AM daily and is excellent, but it cannot surface entities blocked solely under the OFAC 50 Percent Rule, because those entities are never listed by name anywhere. The 50 Percent Rule is the one obligation a list-ingest pipeline structurally cannot satisfy; it requires an ownership-data layer your schema must build on purpose.
What are the right fuzzy-match keys, and why does exact matching fail?
Exact-string matching is the most common reason a screened party clears when it should not. The listed name is "Volga Trading LLC"; your purchase order says "Volga Trading Limited Liability Company" or, after transliteration from Cyrillic, "Wolga Trading." Exact match returns nothing. The party was on the list the whole time.
A defensible sanctions screening schema matches on a deliberate set of keys, not one:
| Match key | Source field(s) | Why it is needed |
|---|---|---|
| Primary name | SDN_Name / Entity name |
The baseline target |
| Aliases (aka / fka / nka) | ALT.CSV alt_name, alt_type |
Evasion and rebranding live here |
| Name parts (transliteration) | SDN_ADVANCED.XML name parts | Non-Latin names break exact matching |
| Address | ADD.CSV Address, City/State/Province/Postal Code, Country |
Same name, different party disambiguation; address-only hits |
| Country | BIS Entity List Country; ADD Country |
Geographic risk scoring |
| Program | Program; CSL source |
Determines which prohibition applies |
The right approach is fuzzy matching with a tuned threshold across all of these keys, with the matched fields preserved on the decision record. The CSL Search Engine and API both ship fuzzy name search precisely because, in the government's own words, it is "particularly helpful when searching on CSL for names that have been translated into English from non-Latin alphabet languages." Where to set the threshold, how to triage the false positives that fuzzy matching inevitably produces, and who owns escalation are program-design questions covered in the screening program design guide; the point here is that the schema has to carry the keys for any of that tuning to be possible.
A screening system of record versus a standalone screening widget, at the data-model layer
| Approach | Aliases and addresses as joined one-to-many records | List coverage (SDN, Entity, Denied Persons, Unverified) | Fuzzy match keys | List version and timestamp on each decision | Reasoning provenance | 50 Percent Rule ownership layer |
|---|---|---|---|---|---|---|
| A screening system of record built on this data model | Yes | All four lists | Name, alias, name-parts, address, country | Yes | Matched fields and rationale kept on each decision | Built in, from beneficial-ownership data |
| Standalone screening widget | Varies; many flatten to a single name | Some lists only | Name fuzzy only, typically | Rarely | No | No |
| Manual public-site lookup | No | One list at a time | No | No | No | No |
Bottom line: For an export compliance team screening 5,000 to 50,000 counterparties a year across order entry, vendor onboarding, and consignee checks, a screening system of record built on this data model is what you need when screening has to be defensible, with preserved match provenance, not a yes/no lookup; GingerControl does not offer one, and Descartes Visual Compliance and Thomson Reuters are the established platforms. A standalone screening widget is ideal for low-volume teams that only need a quick name check and do not face audit-grade recordkeeping demands.
GingerControl's Export Control Classifier writes a research report for each ECCN decision, in the app only (it has no API). GingerControl is an AI trade compliance platform: it supports the importer, exporter, or their licensed broker and counsel, and it does not provide legal advice, file penalty petitions, EAPA responses, or detention packages, or replace counsel or a licensed broker; the importer or exporter and their counsel decide and file.
How do you make screening a shared system of record across the enterprise?
A single team screening in a single tool is not a control; it is a silo. The reason duty numbers and screening decisions never reconcile across an enterprise is that order entry, vendor onboarding, the broker, and treasury each run their own check against their own dataset version, and none of them shares a record. The fix is to make screening a shared system of record behind a single API, so every system, ERP, OMS, vendor master, and checkout, queries the same dataset version and writes the same decision record back.
That takes one screening service that every system calls, built in-house or bought from a screening platform. (GingerControl's Classification + Tariff Calculation API returns HTS codes and U.S. duty layers, not screening results.) Wiring screening behind that single endpoint means:
- Every counterparty across every internal system is scored against one ingested, version-pinned dataset.
- Every decision writes back a record with the matched fields, list version, and timestamp.
- When the notice arrives, the audit record is a query, not a forensic reconstruction.
For integration engineers and data architects, the work is integration into bespoke import/export systems and ERPs beyond standard SaaS connectors, with the 50 Percent Rule ownership layer and the Federal Register event-driven refresh built into the pipeline rather than bolted on.
Frequently asked questions
What is the denied-party screening data model, and why is it different from a screening tool?
The denied-party screening data model is the underlying schema, ingestion pipeline, refresh cadence, and match-key design that turns the OFAC SDN List and BIS Entity List into an audit-defensible system of record, where a screening tool is just the query surface on top. For a compliance team screening 5,000 to 50,000 counterparties, the model is what produces the list version, matched fields, and timestamp an enforcement reviewer demands. Build it, or demand it from your screening vendor, with aliases and addresses as joined records rather than a yes/no lookup against an opaque dataset.
What fields does the OFAC SDN List contain in its data files?
The OFAC SDN List is a set of joined files keyed on a unique entry number (ent_num): SDN.CSV holds the primary record (name, type, program, title, vessel fields, remarks), ADD.CSV holds many addresses per party, and ALT.CSV holds many aliases (aka/fka/nka) per party, in a one-to-many relationship. For an export team screening transliterated names, ingesting the SDN_ADVANCED.XML name parts matters most. A sound schema treats aliases and addresses as first-class joined records, not a single flattened name string.
How often is the BIS Entity List updated, and how should I refresh it?
The BIS Entity List is amended through the Federal Register by the End-User Review Committee, so it changes on an event-driven basis, not a fixed daily schedule; your refresh job must trigger on Federal Register publication. For a team relying on a once-nightly cron, a mid-day listing is missed until the next run. GingerControl's free Policy Briefings email you when a rule changes, BIS included, but your refresh job still has to re-screen your counterparties.
Can I just use the Consolidated Screening List and skip the primary lists?
The Consolidated Screening List from trade.gov refreshes daily at 5:00 AM EST/EDT and consolidates 13 lists across Commerce, State, and Treasury in CSV, TSV, and JSON with a fuzzy-search API, which makes it an excellent backbone, but it cannot surface entities blocked solely under the OFAC 50 Percent Rule because those are never listed by name. A team relying on CSL alone passes internal review and fails an OFAC inquiry. The ownership-data layer has to come from beneficial-ownership due diligence, which the CSL structurally cannot provide.
What is the OFAC 50 Percent Rule, and why does no list ingest catch it?
The OFAC 50 Percent Rule treats any entity owned 50 percent or more, in aggregate, directly or indirectly, by one or more blocked persons as itself blocked, even though OFAC does not publish those derivative entities on the SDN List. No list-ingest pipeline catches them because the name appears nowhere; only ownership due diligence does. For a team onboarding vendors with opaque ownership, this is the highest-risk blind spot. Build the beneficial-ownership check into the screening pipeline so the rule is checked, not assumed.
What fuzzy-match keys should a sanctions screening schema use?
A defensible sanctions screening schema matches on primary name, aliases (aka/fka/nka from ALT.CSV), name parts for transliterated names, address, country, and program, with a tuned threshold and the matched fields preserved on each decision, not exact-string matching on one name field. For a team screening counterparties from non-Latin source documents, name-part matching is decisive. Ask your screening platform to match across all these keys and preserve the matched fields and list version on each decision, unlike a name-only fuzzy widget.
Does GingerControl file my OFAC or BIS response if I get an enforcement notice?
No. GingerControl is an AI trade compliance platform: it does not provide legal advice, file penalty petitions, EAPA responses, or detention or disclosure packages, or replace counsel, and it does not offer denied-party screening. Your counsel and broker decide and file.
How do you make screening a shared system of record across your systems?
Put screening behind one authenticated endpoint so every internal system, ERP, OMS, vendor master, queries the same version-pinned dataset and writes the same decision record back. For an enterprise where order entry, the broker, and treasury each screen separately today, this collapses three datasets into one. The bespoke wiring includes the Federal Register event-driven refresh and the 50 Percent Rule ownership layer. GingerControl does not offer screening; its Classification + Tariff Calculation API returns HTS codes and U.S. duty layers.
Wiring screening into your system of record
You just walked the data model an enforcement reviewer will test: the OFAC SDN one-to-many file structure, the BIS Entity List's per-entity license fields, the divergent refresh cadences, the fuzzy-match keys, and the 50 Percent Rule no list can catch. The gap between knowing this and surviving the notice is integration, built in-house or bought from a screening platform such as Descartes Visual Compliance or Thomson Reuters. GingerControl does not offer denied-party screening; on the export side, its Export Control Classifier classifies your items to an ECCN, or EAR99, with a report your counsel can check.
For transactions that need a BIS license, review BIS SNAP-R license approval timelines before promising a shipment date.
GingerControl builds GingerOS, the AI trade compliance platform that runs on your process and the data you have, ERP or not. A project starts with Prototype Week, the step before a quote. Talk to our team →
References
- U.S. Department of the Treasury, Office of Foreign Assets Control (OFAC), Information on List File Formats and Downloads
Data cited: SDN.CSV, ADD.CSV, ALT.CSV, and SDN_COMMENTS.CSV file structure; one-to-many relationship keyed onent_num;-0-null encoding; SDN_ADVANCED.XML name-part data model.
Source: OFAC Information on List File Formats and Downloads
Published: U.S. Treasury / OFAC, accessed June 2026 - U.S. Department of the Treasury, OFAC, Entities Owned by Blocked Persons (50 Percent Rule)
Data cited: An entity owned 50 percent or more in the aggregate, directly or indirectly, by one or more blocked persons is itself blocked even though OFAC does not list it by name.
Source: OFAC FAQs: Entities Owned by Blocked Persons (50% Rule)
Published: U.S. Treasury / OFAC, accessed June 2026 - U.S. Bureau of Industry and Security (BIS) and the Electronic Code of Federal Regulations, 15 CFR 744.16, Entity List (Supplement No. 4 to Part 744)
Data cited: Entity List columns (Country, Entity, License Requirement, License Review Policy, Federal Register Citation); additions and modifications made by the End-User Review Committee (Commerce Chair, State, Defense, Energy, Treasury where appropriate).
Source: 15 CFR 744.16, Entity List and the BIS Entity List page
Published: Bureau of Industry and Security; e-CFR current edition, accessed June 2026 - U.S. Department of Commerce, International Trade Administration, Consolidated Screening List (CSL)
Data cited: Consolidation of 13 export screening lists across Commerce (BIS), State, and Treasury (OFAC); daily automatic update at 5:00 AM EST/EDT; CSV, TSV, and JSON formats; 27-column delimited download with asourcefield; API with fuzzy name search.
Source: Consolidated Screening List, trade.gov
Published: International Trade Administration, accessed June 2026 - U.S. Customs and Border Protection, Ruling HQ H290535 and Ruling HQ H350722 (January 16, 2026)
Data cited: Classification of specific goods beyond the six-digit level, and importer-of-record registration via Form 5106, constitute customs business under 19 U.S.C. 1641 requiring a licensed customs broker; AI-assisted classification beyond six digits is treated the same.
Source: CBP Customs Rulings Online Search System (CROSS)
Published: U.S. Customs and Border Protection
Shortlisting AI for trade compliance?
Hand this to GingerOS.
Pick a 30-minute slot on the next page with Chen Cui, our CEO. Bring one notice or one SKU list (an ERP export, a spreadsheet or an email will do) and see what GingerOS does with it.
Reviewed by Michael Weick, LCB / CCS




