Risk-Based Transaction Monitoring Through Analytical Segmentation

Compliance officers are required to adapt quickly to new regulatory requirements and changes in how bad actors conduct financial crime. Unfortunately, most risk-based transaction monitoring solutions are static, outdated, and not easily adaptable to new developments in the anti-money laundering (AML) space. Many banks need to consider new solutions that are cost effective and not only meet but exceed these evolving requirements.

Analytical segmentation modeling (ASM) is one way to design an effective AML monitoring strategy through the development of a quality model to achieve segmentation. ASM involves combining customer or bank accounts with similar properties and transaction behavior to make it easy for banks to formulate risk signals based on their various classes of customers. This model sets threshold levels for the segments monitored by identifying patterns based on the groups of similar customers and/or accounts. ASM also allows enhanced monitoring of high-risk segments, which in turn provides authentic alerts that can then be investigated.

How to develop an Analytical Segmentation Model

Developing an ASM requires using the data to inform which framework fits best. Steps to develop the model include:

• Identify the attributes that define the bank’s customers. Some important attributes to consider include:
  • Customer risk rating
  • Client type
  • Account type
  • Product type
  • Transaction type
  • Transaction volume
  • Transaction amount
  • Income
  • Revenue
• Decide if the focus of the model should be on the account and/or customer level
• Add the risk classification for each account and/or customer that will be used
• Perform data quality checks on the customer attributes to ensure the data set does not contain missing or invalid values
• Run a statistical clustering algorithm, such as a k-means algorithm, to group the accounts/customers and build the model
• Analyze and test the integrity of the model
• Perform an exploratory analysis of the data to make sure outliers are removed
• Rerun the model to confirm that the model is showing segments with similar attributes

The segments created by this process are more accurate for grouping than traditional rule-based methods. Monitoring rules specific to these segments can be designed to improve the quality of transaction monitoring and reducing false positives.

Once this process is set-up, the model should be evaluated monthly to establish the range of transactions, statistical distribution of segments, and anti-money laundering risk included in the model. Like every other model, regular upgrades help a bank stay up-to-date with changing customer behavior and reduce false money laundering alerts. This creates efficiency as well as reduce the cost of developing these systems.