Financial crimes programs continue to evolve, with banks and other financial institutions intensifying their focus on the use of data analysis and advanced analytics to gain better insights into the risks they face.
Recognizing that technology is only part of the solution, a growing number of institutions are deploying advanced analytics to refine their organizational approaches to identifying, monitoring, and mitigating financial crime risk. Through the application of analytical context principles and supporting tools, those responsible for risk management, compliance, and data management can prepare for the next steps in bolstering their financial crimes programs.
Analytical context: What and why
Analytical context is best described as a collection of tested methodologies and analytical techniques designed to augment conventional anti-money laundering (AML) and sanctions controls as part of a comprehensive financial crimes program. Applying analytical context to financial crime detection increases effectiveness and efficiency, reduces risk, and provides better reporting to law enforcement.
Analytical context places customer behavior within a broader context, looking beyond the customer reference and immediate transaction data that forms the basis of conventional know-your-customer (KYC) and transaction monitoring (TM) reviews. By also analyzing broader, real-world connections, relationships, and events, financial institutions can gain a more complete perspective on their exposure to financial crimes risk.
Traditionally, financial crimes programs consider each event – such as a TM alert or a suspicious activity report (SAR) – at a point in time. By introducing advanced analytics and big data tools, analytical context enables continuous re-evaluation and consideration of events within the context of changing information and the financial institution’s risk tolerance.
One important advantage of such an approach is the ability to predict the probability that the activities that triggered an alert might, in fact, represent good behavior. This capability reduces noise and false positives from the event population, enabling improved program efficiency and more effective deployment of resources.
Analytical context: Tools and techniques
In the case of transaction monitoring, analytical context transforms traditional rules-based monitoring into an event-driven dynamic process based on the financial institution’s risk appetite. It can dynamically create a networked, relationship view of all entities including direct customers and pseudo-customers (that is, entities that do not have a direct relationship with the financial organization). The view also encompasses originators, beneficiaries, and other related entities.
Several advanced technological techniques figure prominently in the application of analytical context initiatives. These include:
- Data profiling and analysis. A prerequisite to the application of advanced analytics, data profiling and analysis involve the process of understanding available data, how it is connected, how it should be employed, and what impacts it could have that need to be taken into account as models are developed.
The results of this effort should be documented and, prior to commencing the analytical project, any data quality issues should be remediated within the context of a broader financial institution data governance standard.
- Natural language processing. Natural language processing looks at the interactions between computers and human languages, particularly the processing and analyzing of unstructured text data.
In the context of financial crimes, this technique can be used to gain additional investigation insight into an event to understand the reasons why an alert should be either closed or escalated. Natural language processing is useful in recognizing and interpreting regulations that need to be incorporated into a financial institution’s policy or risk assessments.
- Entity consolidation and resolution. Entity consolidation is the process of preparing data prior to monitoring in order to resolve multiple instances of the same entity name. Such instances can happen when transaction data (such as a wire transfer) originates from multiple sources that might identify a given financial institution’s customer or customer’s customer differently. Entity consolidation makes it possible for entities to be monitored holistically. Without it, the same entity is monitored in isolation and could be either alerted multiple times or not alerted when it should be.
Entity resolution is a similar process that occurs post-monitoring. In this step more aggressive connections are made – for example, through telephone numbers and other identifiers – in order to aid investigators looking at the real-world connections of an entity.
- Exploratory machine learning. Machine learning is a commonly used term in the field of data analytics. In the context of financial crimes, there are two different uses. The first, exploratory machine learning, describes the process in which unsupervised techniques are used to understand data and to seek out typologies and insights that can be used to build rules or controls.
An example of exploratory machine learning would be when an unsupervised approach is taken to gain insight on risk categories and then target selected variables in order to create targeted rules. Exploratory machine learning is often used to detect common activities over time to identify patterns related to human trafficking or terrorist financing investigations.
- Predictive machine learning. Unlike exploratory machine learning, which seeks to understand data and gain insights, predictive machine learning seeks to forecast future insights.
For example, predictive machine learning could draw on historic activity to predict the probability that a given event is in fact good behavior and not anomalous. This feature enables the automation of event triage and, through the application of broader techniques, the hibernation of those events that analytical scoring indicates are not likely to warrant further investigation. Such hibernated events then could be continuously re-evaluated when information changes.
- Network analysis. Network analysis is a technique used to identify connections between entities based on all available data in order to develop financial crime typology detection scenarios. The models that are developed from these scenarios can then be deployed to monitor for unusual network activity using precise rules for individual institutions.
Analytical context: Getting started
The application of analytical context principles is still in its early stages in many financial institutions. In fact, it is not even on the radar of risk, compliance, and data executives in many organizations. For example, in a recent Crowe webinar, only one-third (33.6%) of the participating executives reported their institutions had a dedicated financial crimes analytics function. More than half said their institutions were not using any of the specific analytical techniques associated with the application of analytical context.