Financial crime driven by money laundering has reached global and stratospheric proportions. Funds laundered are estimated to run into trillions of dollars annually, estimated to be between two to five per cent of the world’s GDP.
As the quantum of laundered money increases, so do the suspicious transaction reports (STRs), corresponding investigations and prosecutions. Each STR form provides valuable clues about the behavioral aspects of the financial crime and should not be overlooked. Using information about the transaction or the individual, we can identify key discrete factors to rank order STRs from high to low risk. Such analysis can be undertaken using the logit function, thereby, assigning score to each factor as a function of the successful prosecution. Highly predictive factors can then be presented in the form of a multivariate discriminant analysis model. Using receiver operating characteristic (ROC), we can test the predictive power of the behavioral factors shortlisted for the model.
Data-driven STR analysis can help surveillance and monitoring agencies develop a national alert system. Customized alert queries can also be designed, such as, “Create an alert when over the past X days at least Cash deposits were made with an amount between the fixed threshold value($10,000) and X% of that amount”.
AI learning can take place for uncalibrated scenarios, e.g., emerging use of P2P transactions, mules augmented by industry-based intelligence. Such alert queries can generate additional transaction alerts that usually do not get caught in the filter.
To run this analysis, we require aggregate data of the total number of STRs received and successful prosecutions per year. Where prosecution data is unavailable, we use the benchmark case-law from similar jurisdictions.
BankingBook Analytics (BBA) uses accelerators to develop a data and analytics operating model, combined with process-mapping, to assist FIUs with the implementation of the next generation of money laundering analytics and monitoring tools.
BBA’s microservice architecture uses behavioral analytics and ROC accelerators, providing an assurance that each factor included in the model is based on minimum acceptable predictive power, reducing or avoiding false-positive alerts, to avoid expensive administration and investigation efforts. Using technology enablers, such as, Optical Character Recognition (OCR), we can transform STRs into data readable files.
When seeking to upgrade or develop your system’s surveillance capabilities, you should insist on an integrated AML system that has the following capabilities:
- An alert system for regulated entities using AI-enabled STRs scoring
- Automation and integration of regulatory reporting (FATF, Egmont, etc.)
- Lateral integration with the law enforcement agencies for escalated STRs, case files, etc.
Risk-based analysis of suspicious transaction reports can help national Financial Intelligence Units transform from being reactive to proactive regulatory organizations. It is also one of the highly regarded strategic competencies particularly in anti-money laundering efforts, because meaningful risk assessment must be almost by-the-minute analysis due to the accelerated pace of changing financial, technology and social realities. Our system comes equipped with a real-time executive dashboard.
Our easily accessible archiving system, vintage risk-record keeping further empowers investigative capabilities to prevent financial crime proactively, while fighting financial crime on an ongoing basis.