# The Restricted-Party Account Clearance Standard: Clear, Escalate, or Block

*Screen any company list against the official lists, trace 50 percent ownership, grade each row Clear, Escalate, or Block, and file a dated record per row.*

- Canonical URL: https://www.refolk.ai/guides/restricted-party-account-clearance-standard
- Pillar: Process, data, and compliance
- Format: Standard
- Published: 2026-11-15
- Last reviewed: 2026-11-15
- Reading time: 22 min
- Keywords: how to screen a prospect list against sanctions lists, restricted party screening before CRM import, OFAC 50 percent rule screening customers ownership, consolidated screening list fuzzy match false positive, denied party screening account list sales, entity list affiliates 50 percent rule screening prospects

## Key takeaways

- An account row counts as cleared only when it carries Clear, Escalate, or Block plus a dated record that names the list snapshots, the reason, and a rescreen trigger.
- Under OFAC's 50 percent rule, a company owned 50 percent or more in aggregate by blocked persons, directly or indirectly, is blocked even though its own name appears on no list, so name screening alone cannot clear a foreign row.
- Published false-positive rates for name screening run from 85 to 95 percent of alerts, yet one engine in a peer-reviewed test caught only 146 of 303 true matches, so hits must be decided on score plus corroborating identifiers, never score alone.
- The Consolidated Screening List holds active entities only and is not the system of record, so every hit must be confirmed on the original agency list and every run must keep its own timestamped snapshot.
- OFAC requires sanctions records to be kept for 10 years, and a shorter EU deletion period does not excuse that, so the per-row screening record has to be designed to last a decade.
- In Refolk's index, 600 US professionals hold trade, export, sanctions, or export-control manager titles against 129 in the UK, a 4.65x gap that affects how easily a team can staff a second reviewer.

This standard is for the ops owner who has to clear a sourced target-account list before it goes into the CRM and out to reps. It covers screening every company against the government restricted-party lists, settling each fuzzy hit and ownership question the same way every time, and keeping a dated record for each row that holds up when compliance or a regulator asks how the list was cleared. You can adopt it as written, or change the thresholds and keep the structure.

## What counts as a cleared account list?

An account list is cleared when every row has one of three statuses (Clear, Escalate, or Block) and a dated record behind it. The record names the list versions screened, the reason for the status, and the event that will force a rescreen. If a list went through a search box but has no record per row, it was searched, not cleared.

Restricted-party screening means checking each party you plan to do business with against government lists of sanctioned, denied, or restricted persons and entities. For a revenue team the party is a company. The reason to screen before CRM import is simple: once a company is in the CRM, a rep will work it, so a company that should never be worked must not get in.

The three statuses:

- **Clear.** The row was screened against every snapshot list. Any hit was dismissed for a recorded reason, and the ownership trace, where one is required, found no restricted owner. Reps may work it.
- **Escalate.** The evidence in hand cannot clear or block the row. It stays out of rep queues and goes to a named compliance reviewer.
- **Block.** The row is a confirmed match, or blocked persons own 50 percent or more of it. It is never assigned, and compliance takes the decision from there.

Scope: this standard is an operational gate, not legal advice. It decides what reaches reps. It does not decide whether a transaction can be licensed. Block and Escalate rows always end with compliance or counsel, never with RevOps.

## Which lists should an account list be screened against?

Screen every row against the Consolidated Screening List and the UK Sanctions List. Add the EU and UN lists where you sell into those jurisdictions, and confirm every hit on the original agency list. The CSL is a convenient combined feed, but it is not the legal record.

The Consolidated Screening List (CSL) is the US government's combined screening feed. Its API documentation says it brings eleven export screening lists from the Departments of Commerce, State, and the Treasury together into one data feed. That covers the BIS Denied Persons List, Entity List, Unverified List, and Military End-User (MEU) List. It covers the OFAC SDN List, Foreign Sanctions Evaders List, Sectoral Sanctions Identifications List, CAPTA List, Non-SDN Menu-Based Sanctions List, and Non-SDN Chinese Military-Industrial Complex Companies List. State Department lists also feed it, but I have not verified their exact names or how many there are.

| Source | Publisher | What it covers | Caution |
|---|---|---|---|
| Consolidated Screening List | Commerce (ITA) | 11 US lists in one feed | Active entities only; not the system of record |
| OFAC SDN and non-SDN lists | Treasury | Blocked and sanctioned parties | Confirm every CSL hit here |
| BIS lists | Commerce | Denied Persons, Entity, Unverified, MEU | Entity List and MEU drive the Affiliates Rule |
| UK Sanctions List | FCDO | All UK designations | The only UK source |
| EU and UN lists | EU; UN Security Council | Consolidated designations | Feed URLs not verified here; confirm locally |

### How current is each list?

The sources disagree about update timing, so record what you actually retrieved instead of claiming the data is fresh. The CSL API documentation says each list is imported every hour, with a lag of up to one hour. Secondary pages quoting trade.gov say the CSL tools update daily at 5:00 AM Eastern. Nobody publishes how often each underlying agency list changes. That leads to two rules:

1. Log a retrieval timestamp for every list you pull, on every run. Your evidence that the data was current is that timestamp, not a claim about the feed.
2. Keep your own snapshot. The CSL holds active entities only, so a company delisted since your screen will not appear in it. If someone later asks whether a company was listed on the day you loaded it, only your stored snapshot can answer.

> **Watch out:** The old UK list is closed
>
> The OFSI Consolidated List closed at 09:00 GMT on 28 January and gets no further updates. The UK Sanctions List, published by the FCDO at sanctionslist.fcdo.gov.uk in XML, CSV, and other formats, is now the only source for UK designations, so check which URL your screen is actually pulling.

## Which fields should the screen match on?

Match on the legal name and every alternate name. Use the address, country, registration ID, and domain to confirm or rule out a hit, not to create one. Lock six fields per row before screening: legal name, alternate names, domain, street address, country, and registration ID.

| Field | Role | What it proves | How it lies |
|---|---|---|---|
| Legal name | Match key | The entity you would contract with | Trading names and suffixes hide it |
| Alternate names | Match key | Catches former and local-script names | Sourced lists often hold only the brand |
| Street address and country | Corroborator | Same entity, same place | A subsidiary can share its parent's address |
| Registration or tax ID | Corroborator | Strongest single identifier | Often blank; a blank is not a conflict |
| Domain | Disambiguator | Which company your list means | Shared or reseller domains point elsewhere |

OFAC describes a list entry as often having a full name, address, nationality, passport, tax ID or cedula number, place of birth, date of birth, former names, and aliases. For a company row, the fields you can compare are the address, country, registration or tax ID, and former names. A website is not one of the fields OFAC describes. So the domain mostly settles which company your list means, and that lets you find the legal name and registration ID you actually need for the comparison.

Check entity type before any other field. If the list entry is an individual and your row is a company, or the reverse, OFAC treats it as not a valid match.

## When does a fuzzy hit count as a true match?

A fuzzy hit is a true match only when the name score clears your threshold and the full list entry backs it up. That means the same entity type, a match on the full name rather than part of it, and at least two independent identifiers that agree. The score alone never decides a row, either way.

Fuzzy matching means flagging names that are similar but not identical, so the screen catches spelling variants, transliterations, and typos. The same tolerance that catches variants also produces most false alerts, so loosening the matcher always has a cost.

OFAC publishes the adjudication steps. Apply them in this order to every hit:

1. **Confirm the hit is on an OFAC list at all.** If it came from another list, the agency that keeps that list is the authority, so confirm there.
2. **Compare entity types.** An individual entry against a company row, or the reverse, is not a valid match.
3. **Reject partial-name matches.** If only part of the name matches, it is not a valid match.
4. **Compare the full entry.** Check the address, country, IDs, former names, and aliases against your row.
5. **Call OFAC if doubt remains.** If there are many similarities and you still cannot rule the hit out, call the OFAC hotline at 1-800-540-6322.

I found no equivalent published step list from BIS. Apply the same logic to BIS-list hits and record on the row that you did so.

### What the published false-positive figures say

| Source | Figure | Population |
|---|---|---|
| Peer-reviewed study | Over 90% of alerts false | Sanctions screening programs |
| Screening-vendor report | 85 to 95% false | Name screening systems |
| Vendor-published case | 91% cleared as false | One distributor, 612 alerts |

These figures come from financial institutions and general name-screening programs. No one has published hit rates or true-match rates for sourced B2B company lists, so measure your own: on each run, count the hits and the share of them that end as Clear. The vendor case also quotes a true-positive rate that does not agree with its own 91 percent figure, so treat it as one anecdote, not a benchmark.

Matchers also fail the other way. In the peer-reviewed test, one engine had a sensitivity of 48.18 percent, which means it found 146 of 303 true-positive names and missed the rest. The engine that produced zero false positives also missed true positives. A quiet matcher is not proof of a clean list.

The CSL API offers a fuzzy-name option. I have not verified how it works, including which name fields it searches and how much variation it tolerates. Read the current documentation at developer.trade.gov before you set a threshold, and record the matcher setting on every row.

#### Score versus corroboration

Horizontal axis runs from Low match score to High match score. Vertical axis runs from Identifiers conflict or are missing to Identifiers agree.

| Quadrant | What it means |
| --- | --- |
| Low score, identifiers agree | Escalate; the matcher may have underrated a variant |
| High score, identifiers agree | Block once confirmed on the agency list |
| Low score, identifiers conflict | Clear, recording the conflicting identifier |
| High score, identifiers conflict or missing | Clear if one positively conflicts; Escalate if they are missing |

*A fuzzy hit is graded on both axes, and only the top-right corner is ever a Block on name evidence.*

## How do you trace ownership for the OFAC 50 percent rule?

Add up every stake that blocked persons hold in the company, both directly and through intermediaries they own 50 percent or more of. If the total reaches 50 percent, the company is treated as blocked, even though its own name appears on no list. That is why a name screen alone can never clear a foreign row.

OFAC FAQ 401 states the rule: entities directly or indirectly owned 50 percent or more in the aggregate by one or more blocked persons are considered blocked. "Indirectly" means ownership of shares through another entity or entities that are themselves 50 percent or more owned by the blocked person(s). A stake of exactly 50 percent counts. The FAQ's first three examples reach exactly 50 percent, and each entity is blocked.

### Aggregation, worked

These illustrations apply the FAQ 401 definition. They are not quoted from it.

- Blocked person A holds 30 percent directly and blocked person B holds 20 percent directly. Together that is 50 percent, so the company is **Block**.
- Blocked person A owns 60 percent of a holding company, and the holding company owns 50 percent of the target. The holding company is itself blocked, so its stake counts toward the target's total. **Block.**
- Blocked person A owns 40 percent of a holding company, and the holding company owns 60 percent of the target. The holding company is not 50 percent or more owned by blocked persons, so its stake does not count under the definition. The rule does not block the target, but there is a listed party in the chain, so it is **Escalate**.

The rule covers ownership only. Control is outside it, as FAQ 398 says. OFAC still urges caution where blocked persons hold significant minority stakes or control, so this standard sends any known stake or control role held by a listed party below 50 percent to Escalate.

### The BIS Affiliates Rule

The BIS Affiliates Rule applies the same 50 percent test to export controls. It counts parties on the Entity List and MEU List, and also certain SDNs. BIS published the rule (90 FR 47201) and then suspended it (90 FR 50857). It has since extended the suspension by Federal Register notice. FR Doc. 2026-22793 set the end date at 10 January 2027 and left the 50 percent threshold and the underlying license requirements unchanged. The end date has moved before and could move again, so treat it as a value to look up each time, not a fixed fact.

| Item | OFAC 50% rule | BIS Affiliates Rule |
|---|---|---|
| Threshold | 50%+, direct or indirect, aggregate | Same |
| Lists counted | Blocked persons (SDN) | Entity List, MEU, certain SDNs |
| Status | In force | Suspended; end date set by latest Federal Register notice |
| Control alone | Not covered (FAQ 398) | Not established in sources reviewed |

#### Four layers of restricted-party exposure

1. **Listed directly** - The company itself appears on a snapshot list
2. **Owned 50%+ by blocked persons** - OFAC treats it as blocked; in force
3. **Owned 50%+ by Entity List or MEU parties** - BIS Affiliates Rule; suspended, not withdrawn
4. **Listed-party stake or control below 50%** - Not blocked by rule; OFAC urges caution

*Name screening sees only the top layer, and every layer beneath it needs an ownership trace.*

> **Rule:** A clean list search is not a clearance
>
> Every non-domestic row needs a recorded ownership trace before it can be Clear. If the chain cannot be resolved from public corporate records, the row is Escalate, never Clear by default.

For each link in the chain, record the owner, the percentage, the source document, and the date you retrieved it. Public company registries and public corporate filings are the usual sources. If a link has no source, it has not been traced.

## Who should grade the Escalate lane?

A named compliance reviewer grades Escalate rows, and a second qualified reviewer grades a sample so you can check the two agree. RevOps runs the screen and writes the record, but does not decide Escalate rows. The workload is real. One screening-vendor report puts a manual investigation at 20 to 60 minutes, and a vendor write-up puts each false positive at 5 to 20 minutes. Multiply by your expected hits before you commit to a load date.

If you do not have that reviewer in-house, the market for one is uneven.

**600 - US professionals with trade, export, sanctions, or export-control manager titles in Refolk's index**

The same titles return 129 people in the UK, so the US pool is 4.65 times larger.

| Country | Total in Refolk's index | "Trade Compliance Manager" (n=25) | "Export Control Manager" (n=25) |
|---|---|---|---|
| US | 600 | 14 (56%) | 1 (4%) |
| UK | 129 | 11 (44%) | 4 (16%) |

The title shares come from a sample of 25 profiles per country, so read them as directional. In the UK sample, titles lean more towards export control (16 percent against 4 percent in the US). That fits the UK employers in the sample, which include defence names such as Raytheon UK and Parker Meggitt, along with HP, Honeywell Technologies, and GE Aerospace. The US sample includes Apple, DHL Express, Tesla, GE Aerospace, Parker Hannifin, and Emerson. In practice, a UK team that needs a reviewer with ownership-tracing depth has fewer people to choose from and should start looking before the list is ready, not after the hits arrive.

When you need someone who has actually done restricted-party screening, rather than anyone with "compliance" in their title, [Refolk](/) lets you ask for that in plain English and returns named people you can approach as a second reviewer or an advisor.

I ran this search: `Trade compliance managers at US industrial manufacturers who mention restricted party screening` - [see the full result list](https://www.refolk.ai/s/jkeqfe5ts1).

*Returns named trade compliance managers with their current employer and title, ready to shortlist as reviewers.*

## The clearance procedure, step by step

Run these seven steps in order on every list, every time. Each step has a definition of done, and no row moves forward until its current step is done. No published source gives a time per step, so time your first run and use that as your planning figure.

#### Clearing an account list before CRM import

1. **Freeze and normalise the list** - Fix six fields per row: legal name, alternate names, domain, street address, country, and registration ID. Done means a version-stamped file nobody edits during the run.
2. **Snapshot every source list** - Pull the CSL, the OFAC SDN and non-SDN lists, the UK Sanctions List, and any EU or UN lists that apply, with a retrieval timestamp for each. Done means every list version is logged and stored.
3. **Run the automated screen** - Match each row's legal name and every alternate name against every snapshot list. Done means each row is marked hit or no-hit, with score, list, and matched entry ID.
4. **Adjudicate each hit with OFAC's steps** - Confirm on the agency list, compare entity type, reject partial names, compare the full entry, and call the hotline if doubt remains. Done means every hit has a recorded outcome and reason.
5. **Trace ownership for the 50 percent rules** - On every non-domestic row and every unresolved hit, add up direct and indirect stakes held by blocked persons, and separately by Entity List and MEU parties. Done means the chain is recorded with a source per link.
6. **Grade each row Clear, Escalate, or Block** - Apply the grading rubric below. Done means one status and one reason code per row, with a second reviewer's grade on every Block and a sample of the rest.
7. **Write the record, set the trigger, load the CRM** - Write the per-row record, set its rescreen triggers, and load a status field that keeps Escalate and Block rows out of rep queues. Done means no row reaches a rep without a record.

### The grading rubric

Paste this into your policy document. Each criterion is either met or not met, which is what lets two reviewers reach the same grade.

**Clear, Escalate, Block grading rubric**

```
BLOCK if any of:
  B1. Hit confirmed on the original agency list, same entity type, full-name match, and 2+ independent identifiers agree.
  B2. Blocked persons own 50% or more in aggregate, directly or through entities they own 50%+ (OFAC FAQ 401).

ESCALATE if any of (and no Block criterion is met):
  E1. Confirmed hit, same entity type, full-name match, but fewer than 2 identifiers agree because data is missing.
  E2. Many similarities remain after the full-entry comparison (OFAC hotline step).
  E3. Ownership chain unresolved on a non-domestic row.
  E4. Any known listed-party stake or control role below 50%.
  E5. Entity List or MEU parties own 50%+ (BIS Affiliates Rule, suspended).
  E6. Low-score hit where 2+ identifiers agree.

CLEAR only if all of:
  C1. Screened against every snapshot list, with timestamps recorded.
  C2. Every hit dismissed with a reason code: NOT-ON-LIST, ENTITY-TYPE, PARTIAL-NAME, or IDENTIFIER-CONFLICT (name the field).
  C3. Ownership trace recorded for non-domestic rows, with no Block or Escalate finding.
```

*Change the corroborator count only by written policy amendment, and record the version on each row.*

What each criterion proves, and how it lies. B1 proves identity, but it fails if your corroborators are not independent. For example, an address and a country taken from the same registry line count as one identifier, not two. B2 proves blocked status without any listing, but it is only as good as the ownership sources behind it. C2's IDENTIFIER-CONFLICT reason needs a field that positively conflicts, such as a different registration ID. A blank field is not a conflict, so it sends the row to E1, not to Clear.

### Calibrating two reviewers

Before you rely on the rubric, have two reviewers grade the same sample of rows independently. Agreement is the target. Every disagreement either exposes a criterion that can be read two ways, which you then rewrite, or a reviewer applying it wrong, which you then train. If you have no second reviewer in-house, the same Refolk search above gives you a shortlist of outside ones. Repeat the calibration whenever the rubric version changes.

## How does account screening go wrong?

Most failures are not bad matches. They are rows cleared on evidence that could never have shown the problem. Each failure mode below has a symptom and a check you can add to the run.

### 1. Screening the CSL only

What it looks like: every row comes back no-hit, and the list loads. But a subsidiary that is not listed, owned by a blocked parent, sails through, because no name screen can see ownership. Check: require a recorded ownership trace on every non-domestic row.

### 2. Treating a stale list as current

What it looks like: a hit you cannot reproduce later, or a delisted company that is missing from a historical review. The CSL can lag by up to an hour and drops removed entities. Check: store the snapshot and its timestamp, and confirm every hit on the original agency list.

### 3. Pointing UK screening at the closed OFSI list

What it looks like: UK screening runs cleanly against a list that no longer changes. Check: confirm the configured source is sanctionslist.fcdo.gov.uk. The FCDO's own search tool now uses fuzzy logic and ranks results by relevance, so use it to confirm UK hits by hand.

### 4. Clearing on name mismatch alone

What it looks like: a low score treated as a Clear. In the peer-reviewed test, one engine missed 157 of 303 true positives. Check: test your matcher against known true listings and their spelling variants before each threshold change, and never Clear a row without a recorded reason code.

### 5. Ignoring control below 50 percent

What it looks like: a company with a listed minority owner or a listed party on the board is cleared because the rule is about ownership only. OFAC urges caution in exactly that case. Check: criterion E4 in the rubric.

### 6. Assuming the BIS rule is dead

What it looks like: Entity List affiliate exposure dropped from the trace because the rule is suspended. It is suspended, not withdrawn, and its threshold is unchanged. Check: keep criterion E5 in the rubric, and set a calendar trigger on the end date in the latest Federal Register notice.

### 7. Keeping records for five years

What it looks like: a retention policy written before OFAC extended its requirement. Check: set retention to at least 10 years under 31 CFR 501.601.

> A company with no listing can still be blocked, so a clean name search is never by itself a clearance.

## What goes in the per-row record, and when do you rescreen?

The per-row record has to show, on its own, what was screened, against which list versions, by whom, with what result, and what will trigger the next screen. Design it to last 10 years. OFAC extended its sanctions recordkeeping requirement from five years to 10, and it has said that a shorter period required elsewhere, such as under EU rules, does not excuse compliance.

**Per-row screening record**

```
record_id:
input_file_version:
legal_name:
alternate_names:
domain:
street_address:
country:
registration_id:
list_snapshots: CSL @ <timestamp>; OFAC SDN @ <timestamp>; OFAC non-SDN @ <timestamp>; UK Sanctions List @ <timestamp>; EU @ <timestamp>; UN @ <timestamp>
matcher_and_threshold:
hits: <list> :: <entry ID> :: <score>
adjudication: <step reached> :: <outcome> :: <reason code>
ownership_chain: <owner> :: <percent> :: <source> :: <retrieved date>
status: CLEAR | ESCALATE | BLOCK
rubric_criteria_met:
rubric_version:
reviewer_1: <name> :: <date>
reviewer_2: <name> :: <date>
rescreen_triggers:
retain_until: <screen date + 10 years, or longer if counsel requires>
```

*One record per row per run; append on rescreen, never overwrite.*

> **Note:** BIS retention is not covered here
>
> This guide's sources did not cover the BIS recordkeeping period under EAR part 762. Confirm it with counsel and set retain_until to whichever period is longer.

### When to rescreen

No agency publishes a required rescreen schedule. Vendor guidance usually stops at "screen when a record enters the CRM". One export-compliance advisory goes further: it says one-time screening at onboarding is not sufficient and recommends screening at onboarding, at order acceptance, before shipment, and before payment. Two clocks matter. Lists change at the source as often as hourly, and BIS has already moved the Affiliates Rule end date once. So the trigger should fire on change, not on a fixed date.

Set three triggers on every row:

- **List change.** A new snapshot differs from the one on the row's record.
- **Rule change.** A Federal Register notice changes an ownership rule, or the Affiliates Rule end date arrives.
- **Deal event.** Order acceptance, before shipment, and before payment.

#### What happens when a rescreen trigger fires

1. **Trigger fires** - New list snapshot, Federal Register notice, or deal event
2. **Diff** - Compare the new snapshot with the one on each row's record
3. **Rescreen** - Rerun the match on affected rows, or all rows after a rule change
4. **Re-grade** - Apply the same Clear, Escalate, Block rubric
5. **Append** - Write a new dated record and update the CRM status

*A rescreen applies the same rubric to new evidence and appends a new dated record, leaving the old one intact.*

#### Before the list loads into the CRM

- [ ] Every row has exactly one status: Clear, Escalate, or Block.
- [ ] Every source list has a stored snapshot with a retrieval timestamp.
- [ ] The UK source is the UK Sanctions List, not the closed OFSI list.
- [ ] Every hit was confirmed or dismissed on the original agency list, with a reason code.
- [ ] Every non-domestic row has a recorded ownership trace with a source per link.
- [ ] Every Block row was graded by two reviewers.
- [ ] Escalate and Block rows are excluded from rep assignment by the CRM status field.
- [ ] Every record has rescreen triggers and a retain_until date at least 10 years out.
- [ ] A calendar trigger exists for the Affiliates Rule end date in the latest Federal Register notice.

To keep the standard current, check the Federal Register for new BIS notices and the CSL documentation for changes to the matcher whenever a rescreen fires. Recalibrate your two reviewers each time the rubric version changes.

## Frequently asked questions

### Is screening against the Consolidated Screening List enough to clear a prospect list?

No. The CSL combines eleven US lists from Commerce, State, and the Treasury, but it is not the system of record, holds only active entities, and carries no UK designations. It also cannot see companies that are blocked through ownership. Use it for the automated pass, confirm every hit on the original agency list, add the UK Sanctions List and any EU or UN lists that apply, and trace owners on every non-domestic row.

### Does the OFAC 50 percent rule count control or board seats?

No. The rule covers ownership only, and OFAC FAQ 398 says so. A company where blocked persons hold board seats or a minority stake below 50 percent is not automatically blocked. OFAC still urges caution in those cases, so this standard sends any known listed-party stake or control role below 50 percent to Escalate, not Clear. The compliance reviewer then records the decision.

### Do I need to screen for the BIS Affiliates Rule while it is suspended?

Trace for it, but while the suspension holds, do not Block on it alone. The rule is suspended, not withdrawn, and the extension notice left the 50 percent threshold unchanged. Flag rows owned 50 percent or more by Entity List or MEU parties as Escalate. Set a calendar trigger on the end date in the latest Federal Register notice, and rescreen those rows when that date arrives or a new notice appears.

### How often should a CRM account list be rescreened?

No agency publishes a required rescreen schedule, so use triggers instead of a calendar. Rescreen when a source list snapshot changes and when a Federal Register notice changes an ownership rule. Also rescreen at deal events. One export-compliance advisory says one-time screening at onboarding is not sufficient and recommends screening at onboarding, at order acceptance, before shipment, and before payment.

### How long do I have to keep sanctions screening records?

Keep them for at least 10 years. OFAC extended its sanctions recordkeeping requirement from five to 10 years under 31 CFR 501.601, and it has said that a shorter period required elsewhere, such as under EU rules, does not excuse compliance. This guide's sources did not cover the BIS retention period under EAR part 762, so confirm it with counsel and apply whichever period is longer.

### What fuzzy match score should count as a sanctions hit?

No published score works across all matchers, so set your own by testing. Run your matcher against known true listings and their spelling variants, and lower the threshold until it catches them. Expect most of the resulting alerts to be false, because published figures put false positives at 85 to 95 percent of alerts. Then decide each hit on entity type, a full-name match, and corroborating identifiers rather than on the score.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/guides/restricted-party-account-clearance-standard*
