Rosette Enables Fintech to Reduce False Positives by 75% for KYC/AML
In Europe, businesses from the largest financial institutions to individual bookkeepers, attorneys, and real estate brokers must identify their customers to comply with European anti-money laundering (AML) screening regulations. These requirements are often burdensome, involving paper forms, collecting passport information, and ultimate beneficial owner (UBO) data. Finland-based financial technology provider DOKS® introduces cost and labor savings through cloud-based Know Your Customer (KYC) and AML tools and processes.
DOKS electronically collects and stores data for the entire KYC/AML process, digitizing the paperwork journey. It references data from local trade/business registers to determine UBOs and potential politically exposed persons (PEPs) in the organization, verifies identities through local banking IDs of Nordic customers and other electronic ID services, and assesses risk by screening against local and country-specific watchlists.
The Challenge
DOKS has to match names and identities when:
- Screening against sanctions lists
- Verifying electronic IDs
DOKS’ in-house name matching tool wasn’t able to handle the multitude of variations in personal and company names and the increasing volumes of sanctions list screening. Their existing tool was also producing too many false positives in electronic ID verification. The cost of a DOKS customer missing a match to a sanctions list could result in significant penalties and damage their reputation.
Although nearly all DOKS transactions are in languages that use the Latin alphabet, the data on corporations, entities, and owners might be in any language or alphabet.
The Solution
DOKS evaluated commercial and open source name matching options. Besides accuracy — including the capability to match names in different languages and scripts — the ability to process large volumes of data quickly was a core requirement. RosetteⓇ by BasisTech met the need for accuracy and speed because it plugs into Elasticsearch, which DOKS engineers were already familiar with.
Rosette has two other features that attracted DOKS: match confidence score and explainability
Match confidence score: Rosette produces an intuitive score from 0 to 100%, enabling the user to establish a threshold score above which two names are considered a match. Some businesses may use a higher threshold to reduce false positives. Others may set a lower threshold to minimize missing true matches.
“This relevancy score in Rosette is essential,” said Petteri Tenhunen, Co-founder and Partner of DOKS. “It makes it possible to configure the name matching according to the customer’s specific requirement and reduce false positives.”
Explainability: With each name comparison, the AI and technologies powering Rosette provide the reasons behind each match score calculation. Because DOKS is under continuous audit by third parties, it must be able to show the rationale for assigned scores.
“When reviewing our due diligence reports, customers are often very interested to learn why our system ended up with a certain conclusion and why the match percentage was 80% instead of 65%,” said Tenhunen.
“Name matching is central to reducing false positives, and we found few commercial providers with that focus,” said Tenhunen. “Our goal is to reduce the time our customers spend managing regulatory obligations. Rosette is state-of-the-art and complements our KYC product, which aims to deliver the best solution to our customers.”
Impact: Fewer errors in KYC/AML screening
Since DOKS switched all of its customers to the system using Rosette, the feedback has been positive. For every match to a sanctions list entry, DOKS customers are shown the queried name and the initial information from the list, such as date of birth, title, and place of residence, so they can quickly eliminate false positives and decide on the next course of action.
“Using Rosette, the number of false positives has decreased by up to 75% in the best cases, yet hasn’t compromised the reliability of the results,” said Tenhunen. “That means significant savings for our customers running their AML processes in DOKS.”