A screening hit is a question, not an answer. Most potential matches on OFAC, EU and UK lists are common names, and the answer usually sits in data the bank already has.

Screening teams know this. They still spend most of their day proving that a customer called Ahmed Khan, born in Ohio in 1988, is not the Ahmed Khan on a list. The work is necessary. The order in which it is done is the problem.

Why the queue fills up

Name screening is built to be generous. A fuzzy match catches transliterations, spelling variants and reordered names, which is exactly what it should do. The cost is volume: every common name on a list produces a steady stream of hits against ordinary customers.

Most of those hits reach an analyst with only the name and a score. The analyst then opens the KYC file, the account record and sometimes a registry to find the date of birth, the address and the owners. That is the slow part, and it is the same lookup every time.

At Halden Bank, one of the customer stories on this site, most matches came from the same five common names. Analysts were reading the same list entries against the same customers week after week. Nothing about the work was wrong. It was simply done by hand, in the wrong place.

Resolve before you queue

Date of birth, address, nationality and ownership resolve most matches in seconds. Doing that before a hit reaches a queue keeps analysts on the matches that remain genuinely open.

Tipline's Sanctions Screening agent runs that lookup as soon as a hit lands. It reads the attributes the list entry provides and compares them with what the bank holds on the customer:

• Full name, aliases and transliterations, with the matching method shown.

• Date or year of birth, and how far apart the two are.

• Nationality, country of residence and registered address.

• For a company: registration number, directors and beneficial owners.

• Any link between the customer and other listed persons or entities.

When the attributes clearly disagree, the hit is closed with the reason written out. When they agree, or when the bank simply does not hold the data, the hit stays open.

Keep the open ones honest

A partial match that cannot be resolved stays open, with the reason stated: which attributes agree, which are missing, which disagree. Nothing is closed on a score alone.

Take Ostrava Metals s.r.o. from the demo queue. The name returns a partial OFAC match with a score of 0.82. On its own, that score says the strings are close. It says nothing about whether this is the listed party.

The agent finds two more facts. The date of birth of one director matches the list entry, and the same director sits on the board of Halvard Trading, a customer already under review for structuring. Neither fact proves a match. Together, they are a reason to escalate, and the case arrives with both facts and their sources.

What a score is for

A match score ranks work. It does not decide it. A 0.95 on a common name with a different date of birth may be easier to close than a 0.82 with a shared director. Teams that set a single cut-off and close everything below it are trusting the string comparison with a decision it was never built to make.

Our own rule is simple: the score sets the order of the queue, and the attributes set the outcome. Every closed hit carries the attributes that closed it.

The same applies when lists change. A new designation can turn yesterday's closed hit into today's open one. Because each closed hit records the attributes that closed it, the agent can rerun only the hits that a list update could affect, instead of screening the whole book again from scratch.

What examiners will ask

How was this hit cleared, by whom, and on what data. A good screening record answers all three without opening another system. That means keeping the list version, the matching method, the attributes compared and the name of the analyst who signed, on the hit itself.

It also means keeping the closed hits. An examiner who samples twenty cleared matches should find twenty written reasons, not twenty scores.

What to do on Monday

Take one week of cleared sanctions hits and note what closed each one: date of birth, address, nationality, ownership or something else. If most were closed by a single attribute the bank already held, that lookup can happen before the hit reaches anyone. Start with the most common one.