AML Name Screening for Singapore Fintechs: Watchlists, Fuzzy Matching and False Positives

What AML Screening Covers in Singapore

AML name screening in Singapore operates under MAS Notice 626 (for banks) and MAS Notice 641 (for capital markets firms), with cross-references to MAS Notice PSN01 for payment service providers. The obligation is to screen customers, beneficial owners, and counterparties against designated lists before onboarding and on an ongoing basis when lists are updated.

The Three Screening Lists You Must Cover

Fuzzy Matching: What Threshold Is Compliant?

There is no MAS-prescribed fuzzy match percentage. However, MAS examinations regularly challenge firms whose thresholds are so high that they miss obvious name variations. Industry practice for Singapore-regulated firms:

• 80–85% match score: Catches most transliteration variations (Arabic, Chinese, Russian names) with manageable false positive volume
• Below 80%: Risk of missing designated persons with common name variants — be prepared to justify this to MAS
• Above 90%: May miss names written in different orders or with minor spelling differences

MAS expectation: Your threshold must be defensible with evidence. Document why you chose it, what testing you ran, and what false negative rate you accept. MAS examiners will ask.

Reducing False Positives Without Increasing Risk

False positive rates of 95–99% are common at 80% match thresholds. Managing this volume is where most compliance teams struggle. Proven techniques:

1. Date of Birth Scoring

Weight matches where date of birth confirms or excludes the sanctioned person. A "Mohammed Al-Rashid" match with a DOB 30 years different from the sanctioned individual can be algorithmically downscored.

2. Country of Birth / Nationality Weighting

If a sanctioned individual is a national of Syria and your customer is a Singapore PR born in China, flag this as requiring analyst review rather than automatic block.

3. Name Component Analysis

Parse full names into components before matching. "John Smith" vs "Smith, John" are the same person. Many screening systems fail this basic tokenisation step, generating duplicate alerts.

4. Allowlist Management

For cleared false positives, maintain an allowlist with reviewer name, date cleared, and reason. This allows automatic suppression on future scans while maintaining an audit trail for MAS inspection.

Ongoing Screening: List Update Response Times

Onboarding screening is not enough. MAS expects you to re-screen your entire customer book when a new designation is added to the UN or MAS list. Your operational SLA should be:

• Detect new list update: within 4 hours of publication
• Complete re-screen of full customer book: within 24 hours
• Freeze matched accounts / file STR: immediately upon confirmed match
• Report to MAS and CAD: within 24 hours of confirmed sanctions match

What MAS Looks For in AML Inspections

Based on public MAS enforcement actions and industry feedback, MAS focuses on:

• Whether screening covered all required lists (not just UN, but also MAS TFS)
• Whether threshold was documented and tested
• Whether allowlist reviews were re-performed periodically (not just at onboarding)
• Whether the firm took the right action within required timelines on confirmed matches
• Whether staff were trained to make correct alert disposition decisions

Key Takeaways

• Screen against UN, MAS TFS, and OFAC as a minimum — and document why your list selection is complete
• Fuzzy match threshold of 80–85% is industry standard; document your choice
• DOB, nationality, and name parsing significantly reduce false positive volume
• Ongoing screening must respond to new list updates within 24 hours — build this into your ops workflow
• MAS inspections probe threshold documentation and allowlist review quality, not just whether screening exists