One person, many spellings: why name matching decides screening quality

Transliteration, aliases and spelling variants make names hard to match. Good name matching cuts false alerts without missing real ones, and your team keeps the decision.

A name looks like the simplest piece of data there is. In sanctions screening it is anything but. The same person can appear under several spellings, in several scripts, with titles, nicknames, maiden names and older aliases. Whether a screening service copes with this decides whether your team spends its day on noise or on what matters.

An everyday example

Take a common first name such as Mohammed. On lists and in customer files it may appear as Muhammad, Mohamed, Mohammad or Mehmet, combined with different family-name spellings and sometimes only initials. A good screening service recognises these as the same name – and then uses further details such as year of birth or nationality to decide whether it is really the same person.

Why names are so hard

Names travel across languages and alphabets, and every crossing leaves marks. A name written in Arabic, Cyrillic or Chinese characters has no single "correct" Latin form. Different countries, authorities and banks have used different conventions over the years, so the same name may be rendered in many ways.

• Transliteration: one original name, several accepted Latin spellings depending on language and convention.
• Aliases: official lists often record other names a person or entity is known by, including older or informal ones.
• Formatting: order of given and family names, middle names, initials, prefixes, titles and company suffixes.
• Everyday noise: typos, missing accents, abbreviations and inconsistent capitalisation in your own client data.

Two ways to get it wrong

Matching that is too strict misses real cases. A listed person is found under one spelling, and the same person slips through under another. This is the risk compliance fears most, and rightly so.

Matching that is too loose produces the opposite problem. Every common name triggers an alert, queues fill up, and analysts begin to treat alerts as routine. In the worst case, a genuine hit sits among hundreds of harmless ones and receives less attention than it deserves.

Both failures have the same root: a screening service that does not understand how names actually vary. The goal is not to be strict or loose, but to be intelligent about which differences matter.

What good name matching looks like

Good matching recognises that different spellings can belong to the same person, and that similar spellings can belong to entirely different people. It takes aliases seriously, handles the conventions of different languages and scripts, and uses additional information, where available, to separate true candidates from coincidences.

The result is fewer false alerts, which means more time for the cases that deserve it. And because variants and aliases are covered, fewer real hits are missed just because a name was written differently than on the list.

Explain the match, do not just report it

Better matching only helps if your team can see why a name was flagged. A useful alert shows which entry was matched, which alias or spelling variant was involved and what the source says. A reviewer can then judge in seconds whether the similarity is meaningful or accidental.

That transparency is also what makes the process defensible. If someone later asks why an alert was cleared, the answer is not "the system said so" but a documented comparison between the client and the listed entry, with source and evidence.

Signal as a Service – never the decision. We deliver the signal, with source and evidence. Your team decides.

Questions worth asking any provider

• How does the service handle names written in other scripts and their different Latin renderings?
• Are aliases from the underlying lists part of every check?
• Can your reviewers see why a name was matched, not only that it was?
• How does the service help reduce false alerts without hiding potential hits?

Name matching will never be perfect, because names themselves are not. But with the right approach, it can be good enough to remove most of the noise, keep the real signals visible and leave the final judgement where it belongs, with the people who know the client.

If you would like to discuss how this plays out in your own screening process, we would be glad to talk.