Leveraging Analytics in KYC - CREOBIS · KYC analytics can reduce the complex and expensive burden...

Copyright © SAS Inst itute Inc. A l l r ights reserved.

Leveraging Analytics in KYC

Luxembourg, June 29 2017

Eric Malherbe

Compliance solutions manager EMEA


Agenda

1. The context of KYC

2. What are analytics in AML?

3. Which analytics use to improve KYC process?

4. What are the benefits of analytics?

5. Conclusion


Context of KYC


Context of KYC

KYC Issues

• Often paper based

• Multiplication of data

sources to achieve certainty

and accuracy

• Time-consuming

investigations

• Speed has negative effect

on customer experience

• Ongoing regulatory changes

• Spiraling costs

KYC & identity resolution

Offshore leaks

DBs & External registrars

Sanctions & PEPs

Networks & Siloes


Global Anti-Money Laundering Software Market 2017-2021

"One trend in market is increasing demand for know your customer analytics. AML and KYC analyticsare becoming important focus areas for all financial services firms worldwide. Advances in technologyenable banks and financial institutions to collect an enormous amount of data, but these organizationsface multiple challenges in compiling the data to build strong relationships, improve return, andreduce risk.

KYC analytics can reduce the complex and expensive burden of compliance on both AML and KYCdepartments.”

Further, the report states that one challenge in market is high cost of implementation. (...) It alsorequires expensive additional infrastructure, complex programming, and extra time and money toensure data integration and data quality.

For more information please click on: http://www.researchandmarkets.com/publication/m4ucawy/4143977

http://www.researchandmarkets.com/publication/m4ucawy/4143977



For more information please click on: http://www.researchandmarkets.com/publication/m4ucawy/4143977


What are analytics in KYC area ?


Analytics in a nutshell


Data managementAccess and gather all type of data…• any data source (oracle, db2, Teradata, Hadoop…)• any format (flat, structured, unstructured…)

… on the fly…• direct connection with the source systems• real time data ingestion

…while ensuring a consistent data quality, and preparing the way for an accurate detection• Entity resolution, standardization, validation & correction, deduplication, integration…• Data enrichment (matching, data correction, surviving record…)• Continue Data quality control & monitoring• Include Quality Knowledge Database (grammatical, sonority, lexical rules)

J. Smith 18 Queen Street13/01/1980 0208 123 45676

Smith 0208 123 4567613/01/1980

John Smith 13/01/1980 18 Queen Street JN 12 34 56 A

John Smith 13/01/1980

John Smythe JN 12 34 56 A

Entity resolution and network generation

Precision

Structure

Unicity

Validity

Completeness

Consistency


It’s all about dataPrecision

Complete

ness

Consist

encyStructure

Unicity

Validity

• End to end data management capabilities (cleansing,enrichment and quality improvement) added to a flexibleand extensible data models

• Data quality is monitored over time

• Input data are profiled to easily detect, report, alert anyanomaly and take immediate action (alert, reject,correction, enrichment)

Expected benefits of Data Management and Data Quality

Limit manual intervention and guaranty the relevance ofthe alerts generated

Connect to internal and external data sources (OpenCorps, National Registrars, D&B, BVD, Bloomberg, Reuters,other providers)

Accelerate the loading of the Data Model

Manage the entire data quality life cycle : a continuousprocess for alerts accuracy

Ease the addition of new data Provide an end to end consistent solution, from the

source systems to the alerts dashboard

Data management


Social network analysisEntity resolution





Smith 0208 123 4567613/01/1980

Dis

par

ate

dat

a so

urc

es





Smith 0208 123 4567613/01/1980

FROM

TO


Entity resolution





Smith 0208 123 4567613/01/1980

Social network analysis


Social network analysisNetwork building

More entities are resolved and link together to form a network


UNSUPERVISED

When no targets exist

Examine current behavior to identify outliers and abnormal

transactions

Use when a known target is available

Use historical information to predict suspicious behaviors similar to previous patterns

Rule and analytic based network scoring

Automatically risk score while building relevant networks

SUPERVISED NETWORKS

Clustering, K-means, anomaly detection, (neural networks)… Linear & logistic regressions, decision tree, neural networks

Analytical methods


Unsupervised machine learning is the machine learning task of inferring a function to describe hidden structure from

"unlabeled" data (a classification or categorization is not included in the observations). Since the examples given to the

learner are unlabeled, there is no objective evaluation of the accuracy of the structure that is output by the relevant algorithm—which is one way of distinguishing unsupervised learning from supervised learning and reinforcement learning.

Supervised learning is the machine learning task of inferring a function from labeled training data.[1] The training

data consist of a set of training examples. In supervised learning, each example is a pair consisting of an input object

(typically a vector) and a desired output value (also called the supervisory signal). A supervised learning algorithm analyzes

the training data and produces an inferred function, which can be used for mapping new examples. An optimal scenario will

allow for the algorithm to correctly determine the class labels for unseen instances. This requires the learning algorithm to

generalize from the training data to unseen situations in a "reasonable" way (see inductive bias).

Social network analysis (SNA) is the process of investigating social structures through the use

of network and graph theories.[1] It characterizes networked structures in terms of nodes (individual actors, people, or things

within the network) and the ties, edges, or links (relationships or interactions) that connect them. Examples of social

structures commonly visualized through social network analysis include social media

networks,[2] memes spread,[3] friendship and acquaintance networks, collaboration graphs, kinship, disease transmission,

and sexual relationships.[4][5] These networks are often visualized through sociograms in which nodes are represented as points and ties are represented as lines.

NotesAnalytical methods

https://en.wikipedia.org/wiki/Machine_learning

https://en.wikipedia.org/wiki/Supervised_learning

https://en.wikipedia.org/wiki/Reinforcement_learning

https://en.wikipedia.org/wiki/Machine_learning

https://en.wikipedia.org/wiki/Supervised_learning#cite_note-1

https://en.wikipedia.org/wiki/Training_set

https://en.wikipedia.org/wiki/Inductive_bias

https://en.wikipedia.org/wiki/Network_theory

https://en.wikipedia.org/wiki/Graph_theory

https://en.wikipedia.org/wiki/Social_network_analysis#cite_note-1

https://en.wikipedia.org/wiki/Social_media


https://en.wikipedia.org/wiki/Internet_meme


https://en.wikipedia.org/wiki/Weak_ties

https://en.wikipedia.org/wiki/Collaboration_graph

https://en.wikipedia.org/wiki/Kinship

https://en.wikipedia.org/wiki/Disease_transmission

https://en.wikipedia.org/wiki/Sexual_network



https://en.wikipedia.org/wiki/Sociogram


Advanced analytics in DataLab solutionAnalytics made easy to use

Predictive and descriptive modeling techniques

– Linear Regression

– Logistic Regression

– Generalized Linear Model

– Decision Trees

– Clustering

Compare Models

Generates Scoring code for deployment


Which analytics use to improve KYC process?Two use cases:

• Shifting the paradigm of UBO identification

• Customer on boarding process improvement


ICIJ – Panama Papers – Bahama Papers – Wiki Leaks

• Bribery• Corruption• Illegal trade• Money laundering• Tax evasion• Higher costs to society• Economically not a level playing field

If you don’t find it, one will probably do it for you…

UBO identification


Basic business pain: Various sources of information

Who’s who

Internal DB’s

- Customers

- Alerts

- Cases

Sanction & watch-lists

PEP Registrar

International Chambers of Commerce

International Data Providers

(D&B, BVD, OpenCorps, Kyckr, etc.)

ICIJ

•Wikileaks

•Panama Papers

•Bahamas Papers

3.000.000

500.000

1.000.000

300.000

200.000

500.000 • Many checks to perform • Disparate DBs contain the information• Data quality issues prevent easy identification• What about the GDPR which will make this even

more difficult to manage?

• Is this prospect really new?• Does he have ties with other customers?• Are there existing AML alerts or cases?• What about the Credit loans DB?• Sanctions lists / PEP lists• What does the National Registrar say?• Do I need to check information from the international

chambers of commerce • Do I need data providers to identify UBOs?

Data management & exploitation are key

UBO identification


Who’s who

Internal DB’s

- Customers

- Alerts

- Cases

Sanction & watch-lists

PEP Registrar

International Chambers of Commerce

International Data Providers

(D&B, BVD, OpenCorps, Kyckr, etc.)

ICIJ

•Wikileaks

•Panama Papers

•Bahamas Papers

3.000.000

500.000

1.000.000

300.000

200.000

500.000

Data management combined with analyticstechniques such as

Fuzzy Matching & Entity resolution Social Network Analytics Visualization

Analytics proposal for shifting the paradigmUBO identification

Officers and shareholders are automatically included

in the social network

Watch list & sanction list matches in red

Known ownership information indicated by

the link thickness


- Search external databases

- Search for the possible matches and try to find the right ones

- Search databases to find UBO’s and Officers

- Investigate them

- Ownership percentage: 25%, 10% or less ?

- How many levels down ?

- How much time do you have ?

- What level of certainty ?

- When will you be exhausted searching entities where you find nothing…

- Calculate the CDD risk score

Client On-boardingWithout Analytics


- Automated search on external databases

- Automated retrieval of potential matches

- Automated retrieval of UBO’s and Officers

- Automated analytics prioritizing your search- Fuzzy match all UBOs and officers with sanctions lists

- Fuzzy match them with the database of existing customers

- Check whether alerts and/or cases do exist for any of them

- Check whether any of the linked entities are compromised with Panama Papers and the likes

- Go on with the CDD risk scoring process

Client On-boardingWith Analytics


All known Shareholders automatically retrieved

All known Officers automatically retrieved

and matched againstSanctions lists

All known filings


Officers and shareholders are automatically included in

the social network

Watch list & sanction list matches in red

Known ownership information indicated by the

link thickness


Officers and shareholders are

automatically includedin the social network


What are the benefits of analytics ?


Analytics life curveAdvanced analytics push the boundaries

SUSP

ICIO

US

CA

SES

DET

ECTE

D

POPULATION

█ Advanced analytics with Risk-Scored Networks

█ Advanced analytics

█ Random samples

█ Business rules


AML AnalyticsOngoing AML Program Improvement

Reports / Dashboards

Data Exploration

Predictive Analytics

Data ScientistData Stewards / CDOsBusiness Analysts

PredictiveDiagnosticDescriptive


Using AML Analytics and pre-canned reports also improve compliance oversight and prepare for decision

AML Analytics


Alert

Generation

Process

Database

Searches

Text

Mining

Predictive

Modeling

Anomaly

Detection

Automated

Business Rules

LEVELS OF DETECTION

EVENT

ENTITY

NETWORK

SAS HYBRID ANALYTICAL METHODS

Risk

Networks

Ability to improve transactions monitoringHybrid Approach


Conclusion



Datalab

Detection accuracy

Efficiency Phased approach Cloud

Visualisation

Time savings


- Data lab approach - White box - Non-Levensthein screening- Optimization- Risk scored networks- Advanced Analytics

Standard AML package- AML & CDD- Screening Watch lists

- Analyzed data (ETL/real time)

Co

mp

lian

ce le

vel

Covered by competition

Our unique value

Detection efficiency& data coverage

R.O.I.

• Better detection accuracy

• Increased regulatory compliance

• Reduced risk of fines• Manpower efficiency

Question about analytics is not “Why” but “When”


Contact Details

ERIC MALHERBE

Compliance solutions manager EMEA

[email protected]

+32 475 846.951

Leveraging Analytics in KYC - CREOBIS · KYC analytics can reduce the complex and expensive burden...

Documents

Transcript of Leveraging Analytics in KYC - CREOBIS · KYC analytics can reduce the complex and expensive burden...