KYC Perpetual Customer Behavior Monitoring at ING Bank

Overview

ING Bank has partnered with DataWalk to develop a unique and comprehensive methodology for improving the Know Your Customer (KYC) process. Leveraging cutting-edge technology including AI/ML and graph analytics, this methodology offers an automated and streamlined approach to KYC, enabling the bank to quickly meet challenging compliance requirements and reduce the risk of financial crime.

The Challenge 

ING Bank needed to fulfill compliance regulations and understand customer behavior to more accurately assess customer risk profiles. The challenge for the bank was not only to develop an accurate customer profile and identify unexpected behavior, but also to identify the nature and purpose of relationships and define what constitutes a permanent change in behavior. This requires looking at trends over time, rather than individual, isolated changes in behavior, as well as looking at the customer as an entire network of mutual relationships. Regulators gave the bank only six months to design and implement the entire solution.

The Solution

The complexity of an entire customer risk universe at the ING Bank forces the need to leverage advanced technologies such as graph analytics and machine learning. That ensured the fastest way to fuse siloed bank data and discover trends, patterns, networks, and connections. The only solution that meets these needs is the DataWalk Graph & AI Analytics platform.

Graph technology enabled the bank to capture the nature of relationships between entities, while AI (machine learning) fed by the output from graph (relationships) enabled the bank to effectively categorize customers by describing their behaviors statistically and contextually. The company selected DataWalk as a supplemental KYC system, which operates in conjunction with existing systems. The DataWalk system was selected, implemented, and shown to meet the requirement in six months.

DataWalk enabled fast and efficient data preparation and exploration to design the entire perpetual customer behavior monitoring process from scratch.

Figure 1: The Perpetual Customer Behavior Monitoring Process

A critical step was to construct the data universe, which combines transactions, KYC data, products, customer surveys, and other information. Data and analytics leaders often struggle to create a unified view of the business since much of the core business domain knowledge is embedded in traditional databases that employ inaccessible language with an unclear relationship to business reality. This was the primary reason why the DataWalk knowledge graph proved to be an excellent solution. With the knowledge graph the entire data universe was streamlined, with all data (3 billion objects) re-organized around understandable business objects such as entities, banks accounts, transactions, products and so forth, and the relationships between them. This enabled both technical and non-technical users to communicate in the same language through the no-code interface.

Figure 2: DataWalk Universe Viewer (knowledge graph) – the analytical universe for all AML/KYC users

All data was imported and linked together, then for various business segments reference profiles were designed and generated based on 12 months of transactions data, and features that reflect the nature of the relationships between entities.

DataWalk enabled ING Bank to quickly generate the reference profile of customers using machine learning fed by value from relationships. The output from the model is interpreted through descriptive characteristics such as cash transactions characteristics and their values such as "low" or "high" activity, mainly sent or received transactions, etc. This enables the creation of a self-explanatory customer profile, which facilitates communication among KYC analysts, data scientists, compliance experts and Government Financial Investigation Units (FIUs) using a shared language.

The last step was to automate monitoring of the reference profile. Instead of engaging machine learning experts every month to recalculate reference profiles, DataWalk automates this process, repeatedly checking reference profiles against new data to see if the customer's reference characteristics values have permanently changed over time. If so, then an alert is generated indicating that the Customer Due Diligence (CDD) process for that customer should be renewed. The last step in the process was to automatically send the desired report to the existing case management system to proceed with the CDD renewal process.

Figure 3. DataWalk Summary for CDD Renewal presenting the reference profile and the changes in characteristics for the following four periods.