Expertise

Blog

Hiding in Plain Sight: How Duplicate Records Let Sanctioned Customers Slip Through

Many bank customers have received the same letter twice: one copy addressed to J. Smith, the other to Mr John A. Smith. 

For the customer, it is a minor annoyance. For the bank, it is a sign that two of its systems hold two versions of the same person, while it’s likely that neither knows the other one exists.

If a bank can’t tell that both letters are going to one customer, it raises an important question: does it have a complete picture of the people it screens? The answer is often no.

Mergers, product silos, legacy systems and migrations can leave one customer spread across several records. Each record is screened on its own, so alerts multiply, analysts clear the same person more than once, and a true match on one record can stay invisible on the others. 

Institutions that connect fragmented records into a unified customer profile are better positioned to close these gaps.

How does one customer become five records?

Nobody sets out to build records this way. A customer opens a current account, then takes out a credit card run on a different platform. Years later they get a mortgage through a business the bank acquired, whose records were migrated across and never fully merged.

Each system captured the customer slightly differently. One holds a full middle name, another an initial. One stores the date of birth day-first, another month-first. A third has an old address from before they moved.

Complicating matters further, some records are simply incomplete. When the Financial Conduct Authority (FCA) reviewed sanctions controls at more than 150 firms in 2026, it found customer records holding placeholder dates of birth [1]. 

These stand-in values, entered when the real date was never captured, strip out one of the main details screening relies on to confirm or rule out a match. A record like that is hard to match to a sanctions list, and just as hard to match to the other records of the same person.

The result is one real customer and several digital versions of them, each looked after by a different system and, often, a different team.

Why is a fragmented customer a sanctions risk?

Duplicate records cause three problems, and they get worse together.

The first is volume. Every record is screened separately, so one customer can generate several alerts for the same possible match. Different analysts then clear the same person, each time from scratch.

Duplicate records can add unnecessary workload during large-scale rescreening, when banks are already managing significant alert volumes. The consequences of an unmanaged backlog can be severe.

When Bank of America moved to a new sanctions screening platform, rescreening all existing customer records produced so many alerts that, in OFAC’s words, the bank was “insufficiently prepared”. As a result, a business account whose beneficial owner had been placed under US Russia-related sanctions in May 2023 went unreviewed until around July 2024, roughly 14 months after the designation [2].

The second is invisibility. An analyst reviewing one record can’t see what sits on the others. A strong match on one record, or a link to a sanctioned person, may never reach the analyst looking at another version of the same customer.

The third is inconsistency. Two analysts looking at two records of the same person can reach two different decisions, with no way of noticing they disagree.

None of these failures requires anyone to do a bad job. Each analyst can be careful, each system can work as designed, and the bank can still miss the person in front of it.

What happens when a regulator asks who you deal with?

Sooner or later, someone asks the question that a fragmented customer base can’t answer cleanly: show us every relationship you have with this party.

In the US, it’s routine. FinCEN regularly sends banks lists of subjects under investigation under section 314(a) of the USA PATRIOT Act, and banks must search their records for matching accounts and transactions [3]. A bank that only searches one system and not any others has only partially responded to that request.

The stakes rise for banks outside the US. Under the Bank Secrecy Act (BSA), US authorities can subpoena records from any foreign bank with a US correspondent account. If a foreign bank fails to comply, US authorities can require US banks to terminate its correspondent relationships within 10 business days of receiving the relevant notification [4]. 

In October 2026, FinCEN issued its first administrative subpoena under this authority as part of an investigation into a Greek bank's potential ties to Iran. It also issued separate requests for information about banks' relationships with sanctioned Iranian financial institutions [5].

An incomplete response also works against you. In its 2026 settlement with Bank of America, OFAC criticised the bank for piecemeal and incomplete disclosures, and for not promptly answering all its requests for information. The bank only identified the full scope of one problem after OFAC issued an administrative subpoena, and it received no reduction to its penalty [2].

OFAC’s message to every institution was to manage risk across the whole enterprise, connecting the dots to find systemic problems [2]. That is increasingly hard to do when the dots are stored as separate customers.

What do leading institutions do differently?

The institutions that can respond to those regulatory requests quickly treat customer data as a control, not an admin chore.

They resolve before they screen. Records belonging to the same person or company are linked into a unified profile, giving screening systems a more complete view of the customer and reducing unnecessary duplication.

Linking two records is a decision with consequences: merge two different John Smiths and you’ve created a new risk. A governed approach uses automation to identify likely duplicates, with clear confidence thresholds, audit trails, and human review where necessary.

They make decisions consistently and record why. A unified customer profile helps analysts identify related alerts, apply consistent judgements, and maintain a rationale that an examiner can follow.

They treat migrations and acquisitions as risk events. OFAC has said changes such as moving to a new screening tool should be carried out in a controlled way, to avoid gaps and backlogs that can’t be managed [2]. 

Resolving duplicates before a rescreen means fewer alerts to clear when it matters most.

One customer, one complete view

A duplicate letter is harmless. The same gap in a sanctions programme isn’t. 

Fragmented records can generate unnecessary alerts, obscure genuine sanctions matches, and make it harder for banks to give regulators a complete picture of their customer relationships.

Fixing it starts before screening, by knowing who your customers really are. 

That's what Silent Eight's Entity Resolution Agent is built for: identifying records that belong to the same customer and linking them into a unified profile, with governed decisions, audit trails, and human oversight where required.

FAQs

What is entity resolution in financial crime compliance?

Entity resolution is the process of working out which records belong to the same real person or company, and linking them into one profile. In compliance, it means screening and investigating the whole customer rather than each account in isolation.

Why do banks end up with duplicate customer records?

Mostly through growth. Separate product systems, mergers and acquisitions, data migrations and inconsistent data entry all create extra versions of the same customer, each with slightly different details.

How do duplicate records affect sanctions screening?

They multiply alerts for the same person and split the customer’s history across records. An analyst clearing one alert may never see a match or a link sitting on another version of the same customer.

Can merging records create new risks?

Yes. Wrongly merging two different people is as dangerous as missing a duplicate. That’s why the safest approach lets technology propose matches and keeps an analyst responsible for approving each merge.

How does data quality affect AI in compliance?

AI decisions are only as reliable as the data behind them.

Read more in Feeding the Machine: Why Financial Crime Data Must Evolve for the Age of Agentic AI.

Share article

Related Posts

Related Posts

Related Posts

Discover how AI is Revolutionising Compliance and Risk Adjudication

Download our latest collateral to stay ahead.


Discover how AI is Revolutionising Compliance and Risk Adjudication

Download our latest collateral to stay ahead.


Discover how AI is Revolutionising Compliance and Risk Adjudication

Download our latest collateral to stay ahead.