Model Validation

TESTING and ASSESSING Anti-Money Laundering Solutions

Fincom.co - a proven Model Validation solution

Industry standard benchmarking

Fincom.co’s Model validation is being used by leading financial regulators.

Accurate Name matching

Fincom.co’s technology using patented “phonetic fingerprint” and computational linguistics, to transform the way we pronounce a name to a mathematical representation.

Bulletproof validation process

Using Fincom.co’s model validation will assure complying with all regulations, best efficiency and effectiveness benchmarking and constantly keeping ahead of industry .

Does your AML-system match Regulation Requirements and Criteria?

  • Model Validation acting as a proof of technology
  • Aiming to validate the search and matching capabilities
  • Benchmarking current screening solution
  • Model Validation should be performed:
    • Before each deployment of a new version of current solution
    • Periodically for assuring solutions effectivity prior to multiple changes on the different data sources.

Fincom.co’s Model Validation

Our model will assure:

  • Spotting Miss-hits
  • Improve Hit rate – better or equal to industry standard
  • Reduce False positive rate compared with industry standard
  • Accurate Name Comparison across different transliterations, misspellings and in origin language
  • Cross Referencing of additional data parameters.
  • Company’s risk appetite and screening policy are implemented correctly.

THE VALIDATION PROCESS

  • Sanctions lists we test against: OFAC, UK, EU, UN
  • Fincom.co’s test file is based on the above-mentioned lists, including individuals and entities, in their original writing and different name transliterations, as well as names in their original language.
  • The test includes various accuracy and performance tests – defined together with the client, as per system requirements.
  • In the evaluation session – Fincom.co’s platform is used as the Benchmark for comparison to the current system results.

MEASURMENT & KPIs

  • No false negatives
  • # of alerts will should not exceed 30% over Fincom.co
  • Average number of False positives per alert should not exceed 50%% over Fincom.co

Example: Test Case – “NAMES”

Missing Names

comparing full name (including middle names) Vs first+last only.
Eg.: Lorita Salsbury Vs. Lorita Hamiter Katrina Salsbury.
Or first + last name Vs. first\last name only. Eg.: Catharine Fang Vs. Fang

Additional Names

comparing first+last name (including middle names) Vs full name (including middle names).
Eg.: Loraine Norah Vs. Loraine Grosz Norah.

Misspelling

different spelling with similar phonetics should be a match
Eg.: Anakin Skywalker Vs. Anukin Skywoker.

Non-related characters

Existence of non-related foreign characters, signs or symbols shouldn’t prevent a match.
Eg.: Mike Mccaleb Vs. MiKe Mcדcaleb

Camelbacks

rendering “two words” as “twoWords”
Eg.: Ali Abu Ben Vs. Ali AbuBen

Names with suffix

Suffix is not considered as a word thus shouldn’t lower the score.
Eg.: Norah Swanson Jr Vs. Norah Svanson

Titles

Same as suffix, shouldn’t be considered as a word and shouldn’t lower the score.
Eg.: Mrs. Eden Manzur Vs. Eden Mantzur

Out of order names

Playing around with First and Last names’ order, shouldn’t change the score/match.
Eg.: Lorita Hamiter Katrina Salsbury Vs. Hamiter Katrina Lorita Salsbury

Names in different languages

It’s essential in cross border payments to verify original language names against transliterated.
Eg.: محمد محمود Vs. Muhammad Mahmud

Example: Test Case – “ADDITIONAL INFO”

Missing country

two relevant scenarios:

  1. Country exists on the search but missing on the data source
  2. Searching without a country but it exists on the data source.
In both cases, it should be possible to match a name understanding that “blank” data is not contradicting.

Missing dates

two relevant scenarios:

  1. Year of birth exists on the search but missing on the data source.
  2. Searching without a YOB but it exists on the data source. In both cases, it should be possible to match a name understanding that “blank” data is not contradicting.

Different date structure

Matching of full date against YOB is essential.
Eg.: 14/05/1983 Vs. 1983

Different date formats

Supporting matching of dd/mm/yy against dd/mm/yyyy against mm/dd/yy etc..

Thank you for your interest!
Please leave your details

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.