SCREENING ASSURANCE

MICRO-SERVICES
AML SYSTEM ENHANCERS &
OPTIMIZATION TOOLS

SCREENING ASSURANCE

Ensuring no Missed Hits is the very core of any AML system. Yet Missed Hits (False Negatives) are inevitable when using almost any AML system currently existing on the market.

Fincom’s Accurate Name Matching ensures NO Missed Hits thanks to its proprietary Phonetic Fingerprint technology. This is achieved by the system’s unique ability to capture different spelling mistakes and variations, typos, and out-of-order names (unstructured names) in both the language of origin (different alphabets) and transliterated forms in 40+ languages.

Identifying foreign names in 40+ languages, including Russian, Chinese, Arabic, Korean, and many others.

Unique understanding of unstructured names, guaranteeing NO misses, while ensuring no AML Discrimination.

Unprecedent accuracy, capturing approximately 13% of Missed Hits that occur in both legacy and current AML solutions.

ACCURATE NAME MATCHING

Fincom’s ‘Accurate Name Matching’ mechanism provides an exceptional ability to establish a true level of Phonetic Proximity and can validate match against all typical instances of name variations, such as: out-of-order names, misspellings, non-related characters, missing names/additional names, initials, names with suffix, corporate identifiers, and so on. The system is designed to identify the sound of a word or a name, rather than its spelling. This enables the system to match names written in different languages (alphabets) and spelled differently based on accent variations.

KEEPING A WIDER FUNNEL

WHILE REDUCING FALSE ALERTS RATE AND MITIGATING OPERATIONAL BURDEN

The high accuracy of the phonetic-linguistic engine allows open a wider funnel, applying a far more conservative approach, to ensure NO MISSES. Due to the combination of Phonetic Fingerprint technology and 4 advanced distance algorithms (within and in-between languages), the rate of False Positives is extremely low and does not overload suppression stages.

INSIGHTS: CHALLENGES

THE LANGUAGE CHALLENGE - CAUSING HIGH RATE OF MISSED HITS

The global financial system today must be able to handle client’s payments in either domestic or international transactions, with names derived from various origins, be it from Spanish, Russian, German, Chinese, Arabic, Korean, or any other language.
Current AML systems do not “recognize” languages, since most of them do not use Phonetic-Linguistic technology. Thus, they can only compare one string to the other. Further, most systems do not have the ability to screen sanctioned names in the language of origin, although this might be required in some cases (example: SWISS sanction list that includes names in their language of origin). All foreign names are being transliterated either by the source database, or in the process of AML screening by the banks and FI’s AML systems, or in the sanction list itself as opposed to the ”sanctioned name”, as it appears in various official documentation. While the transliterated names may sound similar, their spelling varies considerably. As a result, if such a name appears in any of the Sanctions or PEP lists, it is not recognized by the AML system due to spelling variations resulting from transliterations. This leads to Missed-Hits. As a result, many sanctioned individuals and/or companies are not flagged and continue carrying out illegal transactions.

ADDING RULES TO STRING-MATCHING ALGORITHMS WON'T HELP

The true measure of any AML system is the level of Effectiveness that needs be 100% – meaning NO MISSED HITS. However, current AML systems often fail to achieve a reasonable level of name-matching accuracy since they are based on string distance algorithms, such as N-grams, Levenshtein, and others, measuring the difference between two strings, i.e. how many characters (letters) in a name are identical. These technologies, once applied as a basis for any AML screening, fall short when names are derived from different languages or are written in different structure. The alterations may happen due to typos, transliterations, spelling variations, not to mention the common problem of unstructured names. The risk of Missed Hits and consequent potential fines is very high when a name screening system is set to find an exact or a very high probability match. In order to overcome this problem, banks and financial institutions apply Fuzzy Logic thresholds to a minimum score that may ensure a high probability of a true name match. The most common threshold is roughly 72%. As a result, the Alert Rates (and consequent False Positives) are enormously high, which often leads to adjusting the threshold of Fuzzy Logic back to a high level of 85% or even higher, extremely narrowing the funnel and, thus, missing a lot of names. That, again, results in Missed Hits, since the system can no longer recognize spelling or structural variations of the searched name.

Example: Monic – Mounica – Mounique – NO MATCH → MISSED HIT

Unstructured names:
A major challenge that causes Missed-Hits is the challenge of unstructured names.
Unstructured names may be represented in various forms and orders in different databases, which makes it virtually impossible for a system based on string distance algorithms to recognize and match them. As an example: First, Middle, Last → Middle, First, Last, etc. Names that have 3-6 parts are very common for Spanish, Arabic, Russian, and many other names.

Fincom’s Screening Assurance Micro-Service based on Accurate Name Matching, handles 44+ languages ensuring NO MISSED HITS while reducing alert levels below 5%.

Copyright © 2016-2022 FinCom.co All rights reserved.
|
|

Thank you for your interest!
Please leave your details