In this guide
False positives in AML and sanctions screening are reduced mainly by improving data, handling names properly, and configuring matching logic under governance.
Screening is a probabilistic comparison against incomplete information. A measure of over-inclusion is intentional so that genuine matches are not missed.
Why false positives occur
- Sparse customer records: a name with no secondary attributes.
- Common names, which generate high volumes of matches.
- Transliteration variation, where a name has many valid renderings.
- Recurrence: the same cleared match re-alerting at every cycle.
Data quality first
Data quality is the highest-yield intervention. Validate at capture, keep names in structured fields, and capture secondary identifiers like date of birth and nationality.
Configurable matching and thresholds
Matching configuration should be owned by the firm. That means control over the algorithm, per-attribute thresholds, and separate treatment of individuals and entities.
Governance and testing
Treat the screening configuration as a regulated control. Maintain documented rationale for each setting, pre-deployment testing, and version history.
Metrics that show real improvement
How Agora supports screening quality
The Agora Due Diligence Platform provides screening with configurable matching and structured customer data capture that improves attribute completeness. Screening sits inside the wider customer due diligence software platform, runs continuously against existing customers under periodic and perpetual KYC software, and can be re-run across a whole population using KYC remediation software.
Frequently asked questions
What causes false positives in sanctions and AML screening?
Most false positives come from a small set of causes: poor or incomplete customer data, names that are common or transliterated in several ways, missing secondary attributes such as date of birth, and matching thresholds set conservatively.
Can false positives be eliminated entirely?
No. Screening is a probabilistic match against incomplete data, so some level of false positives is unavoidable. The objective is to reduce avoidable noise while maintaining detection of genuine matches.
Is it acceptable to raise thresholds to cut alert volume?
Threshold changes are legitimate only as part of a governed, tested tuning exercise with documented rationale and evidence of the effect on detection.
What metrics show screening is improving?
Useful measures include alerts per screened record, the proportion of alerts discounted at first review, and detection performance against a maintained set of known-match test cases.
How should tuning changes be evidenced?
Record the rationale, analysis and test results, the approval, and the post-implementation monitoring. A reviewer should see why the configuration is set as it is.
Next step
Optimise your screening
Review screening configuration and alert handling against your own population.