BSA/AML and Fraud Detection
Optimizing models to protect customers & the bottom line.
The costs of reputation risk, regulatory issues, and monetary penalties are very real. Where BSA/AML and fraud detection are concerned, a dependable, robust validation is critical to ensure models are functioning as intended to provide real-time information to protect against the threat of money laundering, terrorist financing, and fraud.
“Criminals don't take time off. It is more and more important that institutions are adaptable and flexible to deal with new and increasing challenges.”
Given the importance of BSA/AML and fraud models, the customer, account, and transaction data that underpin them are critical. With poor data, the models would be calibrated improperly and could produce results that fail to catch criminal activity. DCG's data validations are guided by data science and ensure that BSA/AML and fraud models run on trustworthy data.
Rules Analysis and Transaction Testing
DCG's BSA/AML validations use a proven transaction testing process that frequently detects modeling issues for clients that previous validations have missed. The DCG team recreates and sensitivity tests AML rules to identify potential gaps in the ruleset. DCG also performs optimizations on the false positive and false negative rates of AML rules, as well as on composite AML risk scoring models.
Sanctions Screening and Watch List Filtering
Some of the largest regulatory fines issued to banks in recent years are due to deficiencies in their ability to detect transactions from sanctioned, criminal, or suspicious entities. Combining a risk-based approach for customer/transaction screening with an intimate understanding of the technical methodologies used in the models, DCG's validators help ensure that sanctions screening and watch list filtering models perform at their best.
DCG has a real-world understanding of the different types of fraud commonly perpetrated by criminals, such as identity theft, email phishing, and credit card fraud. Our experts develop, train, and implement machine learning models for fraud detection. Not only do we provide rigorous and practical fraud model validations, we also provide in-depth training on topics related to fraud modeling.
Peer Into Black Box Models
Confirming that vendor models are working as intended is critical to preventing money laundering and fraud. However, with ever-increasing complexity in methodologies used by vendors, institutions need more than a high-level program audit. DCG's team includes programmers who can recreate rules for transaction testing as well as experts who provide independent risk assessments and guidance for program improvement.
How We Help Clients
Examples of Client Results
Strengthened anti-money laundering (AML) programs at $10B+ institutions.
Recreated and tested rules-based strategies to identify rules that have blind spots or loopholes, create an excess number of false positives, or produce too many false negatives among high-risk customers.
Helped management revise rule thresholds to ensure they provide more useful information.