Introduction
iComply’s admin tools enable compliance, operations, and IT managers to set and enforce AML risk screening procedures for their compliance and operations teams. Search parameters can be configured for applications such as name screening, payment screening, and daily ongoing monitoring.
To effectively screen for AML risk, you must identify if your customer is a match with or an affiliate of sanctioned or high-risk entities. Both individuals and legal entities can change their name, jurisdiction of domicile, attempt to alter their name or spelling, or simple typos can occur by both the client and the agent.
While the information contained in sanctions, adverse media content, and PEP lists can also contain typos, false information, and errors - most regulators expect compliance teams to be able to account for this in their risk screening program and procedures. This document provides an overview of how the iComply system can be configured to support unique screening profiles, search algorithms, and automation thresholds.
Search Overview
While there are many reasons why it can be difficult to match your KYC subjects with the names in Sanction, Adverse Media, or PEP databases, the most common pitfalls are listed below with tips for how to configure your search profiles to overcome these challenges.
Fuzziness Settings
Fuzzy logic, Edit Distance, or Levenshtein Distance is an algorithm used in information theory, linguistics, and computer science to reduce the impact of spelling errors, typos, deviations, phonetics, translations, and other variations in name spellings.
Fuzziness settings can be controlled systemwide for your entire iComply instance using default or customized Search Profiles by Admin Users. If enabled by an Admin, any User can fine-tune the fuzziness setting for any customer from within the AML Search Dashboard or KYC Profile Pages.
The Fuzziness setting will enforce the minimum word length to which Edit Distance can be applied. Edit Distance is a way to measure how different two words are by counting the number of changes needed to turn one word into another. For example, to turn "cat" into "bat," you only need to change one letter.
Typos and Spelling Errors
Fuzziness is used to measure the difference between two sequences of text or numbers. In layman’s terms, the Levenshtein distance between two words is the minimum number of single-character edits (insertions, deletions, or substitutions) required to change one word into the other.
When combined with name derivatives, phonetically similar names, variation in the name forms, and non-Latin character searches - fuzzy logic is a powerful application of machine learning that can improve efficiency, effectiveness, and automation, and separate the signals from the noise.
To capture potential typos and spelling errors without returning a significant number of false positives, we have capped the maximum edit distance to one character.
Minimize False Positives
Filtering
To enhance the accuracy of your searches and reduce false positives, the iComply system allows for additional filtering criteria such as entity type, country, year of birth/incorporation, and source. By default, the iComply system is configured without filtering. Contact the iComply Customer Support team to change your default filtering settings.
Search Profiles
Contact your iComply Customer Support team to adjust your default search profile, or create multiple search profiles to execute based on triggers and thresholds from user activity and data. Search Profiles can be configured across broad categories (i.e. sanctions, watchlists, political exposure, adverse media), sub-categories (i.e. violent crime, fraud, sex crime, etc), or by specific data sources (IOSCO warnings, OFAC, FBI Most Wanted, etc).
Each iComply instance will have a default search profile configuration. Admin users can create multiple unique search profiles to support unique risk screening procedures and automatically support multiple jurisdictions, risk profiles, transaction thresholds, etc.
Consolidated Risk Profiles
In most cases, risk information on one entity can come from many different publicly available data sources. By consolidating all of these Listings into a single Risk Profile for any entity we can uniquely identify we can reduce the false positives and more effectively identify risk. This allows users to quickly determine if the potential match is a true or false positive, and make a prompt decision to accept or reject the entire risk profile.
For example: a search through all data sources for sanctions, watchlists, PEP, and adverse media may produce 200 potential matches consisting of multiple references to only 5 unique entities. To simplify the assessment process, all potential matches associated with a particular entity identified in the risk profile will be combined into a single risk profile. Consequently, only 5 distinct risk profiles will require evaluation.