What should a false-positive disposition record contain?

At minimum it should answer why the candidate was returned, which facts distinguish the screened subject, what the reviewer decided and why, who made the decision and when, and what could later change the outcome. That small record lets the next analyst reuse prior reasoning while still checking whether the facts remain current.

Should a screening program aim to eliminate every false positive?

No. Because names, aliases and incomplete records overlap across languages and jurisdictions, some false positives are unavoidable. The target is not zero false positives; it is fast, consistent and reviewable disposition.

Is a large reduction in alert volume a good measure of screening quality?

Not on its own. 'Alerts reduced by 80%' can be a good outcome, but it can also be a sign that a system stopped retrieving useful candidates. Disposition quality, such as time to disposition, structured reasons, reopening rate and repeat-alert rate, is more informative than alert volume alone.

Blog

ComplianceAugust 8, 20269 min read

A false positive is not a screening failure. An unexplained clearance is.

Q: Are false positives a sign that a sanctions screening system is broken?

No. False positives are an expected consequence of searching human names, aliases and incomplete records across languages and jurisdictions. They are not evidence that the system is broken. The actual failure is a clearance nobody can later explain: no candidate context, no identifiers compared, no reviewer, no reason and no link to the earlier case.

False positives are unavoidable in name screening. The real control is how your product explains, routes and preserves the decision to clear or escalate them.

Every sanctions-screening team knows the weary moment: a familiar name triggers another alert. The customer is legitimate. The date of birth conflicts. The country is wrong. The case has been investigated before. Someone clears it, moves on and hopes it never comes back.

That last part is where a product either reduces risk or creates it. False positives are an expected consequence of searching human names, aliases and incomplete records across languages and jurisdictions. They are not evidence that the system is broken. The actual failure is a clearance nobody can later explain: no candidate context, no identifiers compared, no reviewer, no reason, no link to the earlier case.

Short answer

Teams should treat a screening candidate as a structured review task, not as a binary alarm. A good workflow shows why the candidate was retrieved, asks for the evidence needed to distinguish the subject, records the reviewer’s disposition and preserves the outcome so the same identity does not have to be re-litigated from scratch. The target is not “zero false positives.” The target is fast, consistent and reviewable disposition.

Why false positives are unavoidable

People and companies reuse names. Records may contain aliases, transliterations, abbreviations, missing dates, different alphabets and out-of-date addresses. If a system only returned exact spelling matches, it would look clean while missing meaningful variations. If it returned every remotely similar string, analysts would drown.

This is a trade-off, not a secret algorithm problem. The product has to make the trade-off inspectable. An analyst needs to see more than similarity: 0.86; they need the attributes that aligned, the attributes that conflicted, the origin of those attributes and the information still missing.

The right unit of work: a disposition

When a candidate is returned, the next action should produce a disposition record. At minimum it should answer:

Question	Example evidence
Why was this candidate returned?	Similar legal name, alias or transliteration; matching source record
Which facts distinguish the screened subject?	Date of birth, nationality, registration ID, address, LEI, document or counterparty evidence
What did the reviewer decide?	Clear, escalate, reject, request more information or defer
Why?	Concise reason tied to the compared facts—not a generic “false positive” label
Who made the decision and when?	Reviewer identity, timestamp, policy/rule version and second approval where required
What could change the outcome?	New identifier, ownership update, source change, renewal date or monitoring event

This small record changes the economics of screening. The next analyst can reuse the prior reasoning while still checking whether the facts remain current. The compliance lead can audit the quality of clearances. Engineering can identify where product data is insufficient rather than blaming analysts for taking too long.

Design choices that reduce noise without hiding risk

Put the reasons next to the candidate

Do not require reviewers to open six tabs to understand a match. Show the original input, normalised form, matched fields, conflicting fields, source record and candidate identifiers in one review surface. A developer implementing the API should have access to the same explanation programmatically, not just in a dashboard.

Ask for the smallest useful fact

If date of birth would distinguish two people, request it. If an LEI or company-registration number resolves the entity, use it. Do not make the analyst write a narrative before the system has surfaced the most efficient discriminating attribute.

Reuse decisions carefully, never blindly

Case linking is powerful when the subject identity, evidence and risk context are stable. It is dangerous when it turns an old clearance into an indefinite exemption. Product design should retain the previous case, show the basis for reuse and reopen the review when source state, ownership, customer information or policy changes.

Measure disposal quality, not only alert volume

The meaningful operational metrics are usually:

median and 90th-percentile time to disposition by case type;
percentage of clearances with a structured reason and supporting identifiers;
reopening rate after new information;
repeat-alert rate for the same resolved identity;
analyst disagreement/rework rate;
queue age for escalations.

“Alerts reduced by 80%” can be a good outcome, but it can also be a sign that a system stopped retrieving useful candidates. It is not a sufficient metric on its own.

What automation should and should not do

Rules and models can route obvious cases, request missing data and prioritise reviewers. They can also preserve consistency by applying the same policy to similar evidence. But automation should not hide uncertainty behind a green badge.

A responsible workflow exposes the rule, record version and evidence that led to an automatic action. It gives a reviewer a path to override the result and stores that override as part of the case history. In regulated decision-making, explainability is not a decorative feature; it is the interface between automation and accountability.

Where Verifex fits

Verifex’s matching and case workflow should make candidate reasoning and disposition first-class: what matched, what conflicted, who cleared or escalated the case, and what evidence supported it. That is the practical promise behind an audit trail. Teams should validate exact matching behaviour and review capabilities against the current matching methodology and product documentation for their own policy.

Sources and further reading

This is educational material about screening operations. Verifex provides screening infrastructure and evidence records, not legal advice, transaction approval, or a replacement for your risk-based compliance program.

Run a screening and inspect the decision record.

The free plan includes OFAC and UN screening. Coverage stays explicit when a required source is unavailable.

Start screening free