Name-Screening Calibration
Name screening finds possible matches against sanctions, PEP, adverse media or other risk datasets. It is not a yes/no engine: false positives and false negatives are part of the control design problem.
1. What screening actually produces #
A screening system usually produces candidate matches. A candidate match is an alert that needs review. It is not automatically a confirmed match, and it is not automatically a false positive.
| Result | Meaning | Risk if misunderstood |
|---|---|---|
| No hit | No relevant candidate was found in the checked source and matching scope. | Can be wrongly treated as full AML clearance. |
| Possible match | The system found a similarity requiring review. | Can be wrongly treated as confirmed sanctions/PEP exposure. |
| False positive | The candidate match was reviewed and reasonably discounted. | Can consume resources if matching is poorly tuned. |
| False negative | A relevant match was missed. | Can create serious control failure. |
2. Why false positives happen #
- common names;
- partial name overlap;
- different transliteration systems;
- missing date of birth, address or nationality;
- poor customer data quality;
- alias-heavy list entries;
- conservative thresholds designed to avoid missed matches.
3. Why false negatives happen #
- overly strict exact matching;
- failure to screen aliases or transliterations;
- outdated lists or stale customer data;
- screening only the direct customer while ignoring relevant related parties;
- weak data ingestion or mapping from onboarding systems;
- poor handling of non-Latin scripts, abbreviations or name order.
4. Calibration is a governance activity #
Calibration means finding a defensible balance between too many irrelevant alerts and the risk of missing relevant matches. It should be governed, tested and documented. It is not a one-time technical setting.
| Calibration area | Control question |
|---|---|
| List quality | Which lists are used, how current are they, and what do they contain? |
| Customer data quality | Are names, dates, addresses, IDs and related parties captured consistently? |
| Matching logic | How are exact, fuzzy, phonetic, alias and transliteration matches handled? |
| Thresholds | Are alert thresholds reviewed against risk, workload and quality outcomes? |
| Alert review | Are decisions documented with identifiers and source evidence? |
| Testing | Are sample cases, tuning results and changes recorded? |
5. Analyst review discipline #
A reviewer should look for identity evidence, not only name similarity. Useful match-resolution fields can include date of birth, place of birth, nationality, address, identifiers, entity registration number, ownership/control links and source list context.
6. Source basis #
- OFAC Sanctions List Search for approximate/fuzzy search concept.
- Wolfsberg Sanctions Screening Guidance for sanctions-screening control framing.
- Evidence Weight for match-resolution and source-limit recording.
See Also#
References#
- FATF risk-based approach and AML/CFT programme concepts listed in the Source Register.
- Wolfsberg/Basel-style governance and control-quality framing where applicable.
- Editorial & Source Policy for source hierarchy and reviewer-accountability boundaries.