Voice of the Industry

Transaction monitoring - next generation – reality or myth?

Monday 12 July 2021 08:39 CET | Editor: Alin Popa | Voice of the industry

What does the next generation of transaction monitoring has to offer? Will the false positives still be an issue in the future? Colin Whitmore, senior analyst for Aite Group answers these and many other questions

To meet their anti-money laundering (AML) commitments, financial institutions and other regulated firms need to monitor their customers on an on-going basis and report activity that is considered suspicious. Transaction monitoring (TM) systems have been used by firms as the means of identifying unusual activity amongst their customers, which on further – human – consideration may lead to reporting of a suspicious activity report (SAR). TM systems deploy rule- and mathematical-based detection models to look for unusual activity and raise alerts should it occur. These alerts are then worked by analysts and investigators and reported where appropriate.

TM systems have been in use for a number of years, however, they suffer from the issue of false positives – alerts raised, which on inspection by a human are not considered unusual. This is very costly for firms as alerts require human consideration before they can be closed.  

Fundamentally, if the net is cast wide then it draws into many false positives, if it is too narrow it risks missing true cases of money laundering. FIs often err on the side of caution, setting the net wide, with the result of many false positives, all of which need to be subsequently closed. Other reasons for false positives include outdated models and rules, or customers who were classified at the time of onboarding into the wrong segment or peer group.

The next generation of technologies for TM promises a lot, including a solution for the false positives, the ability to identify ‘unknown unknowns’ i.e. the true cases of money laundering, and to automate the end-to-end process. The next generation brings together a number of techniques, it is bedded on the application of machine learning techniques that have been around for many years in other competencies. They are now starting to appear in the AML TM domain and include:

  • Robotic process automation (RPA) and natural language procession (NLP) for information gathering and case creation;

  • Dynamic segmentation, with a detailed focus on analysing customers’ transactional activity and using that to assign appropriate risk categorisation;

  • Contextual analysis through advanced link analysis;

  • Improved accuracy through the separation of noise from signal;

  • Monitoring and updating detection models in real time;

  • Using clustering and other deep learning for better prediction and identification of unknown unknowns;

  • Applying supervised learning to mimic human decision-makers and automate decisions.

Many software vendors offering these new techniques, often pointing to a 40-50% reduction in false positives. With these outcomes you would think that every FI would be busily implemented next-gen solutions, enthusiastically replacing their existing legacy solutions. The reality on the ground is different, and there are several reasons for that. One of the most common is the need for a FI to understand the techniques and to be able to explain them to their regulator. Why an alert or SAR was produced, or equally as important why it was not. Then there are the sunken investments in existing systems and the costs of putting in new systems, then the selection of vendors especially given the fact that many fintechs are small and not always able to weather larger institutions’ vendor onboarding scrutiny.   

So where are we now? Many institutions are now considering, experimenting with, and implementing solutions based on machine learning. Smaller FIs, payment providers, and MSBs are more likely to adopt cloud-based services from smaller vendors who have machine learning at the heart of their products. Larger institutions are working in partnership with vendors, seeing how and where the technology fits, often augmenting exiting TM systems. In several cases, they are financially investing in smaller fintech firms to ensure they can deliver and support solutions.

Next-generation approaches are not only for fintech firms. Larger legacy vendors are not standing still, they are augmenting their current functionality, applying machine learning techniques including link analysis, supervised, and unsupervised learning to enhance their solutions.

Finally, let’s not forget the human in the loop. Currently, humans are making decisions on whether an alert is true or false, often inefficiently as they spend time collecting base information, bringing facts together, and documenting outcomes. Techniques such as RPA and NLP can streamline this process, saving significant time and effort. Advanced analytics can triage alerts ensuring that cases of highest concern are escalated first, allowing the human to focus on priority alerts. Even where next-generation approaches promise fully automated decision-making, we are still seeing the ‘human in the loop’ with the human being responsible for the ultimate decision. 

Conclusion

Transaction monitoring, like so many other areas of banking, is on a journey. The next generation promises much and could revolutionise monitoring, delivering the Holy Grail, of fewer, better quality alerts, the ability to detect new unknowns, and amend models in real time to adapt to emerging threats.

It is an evolving space, and there are already benefits from applying machine learning, for many in the industry we are nearer to the next generation but not quite there. Whatever the outcome, the good news is that AML and innovation now go hand in hand, with augmented intelligence supporting the human decision- maker but for the foreseeable future not replacing them.

About the author

Colin Whitmore is a senior analyst in Aite Group’s Fraud & AML practice, specialising in AML and sanctions with over 20 years of experience within banking and financial services, including Thompson Reuters, Aviva, Barclays, Royal Bank of Scotland, and HSBC.

 

 

About Aite Group

Aite Group is a global research and advisory firm delivering comprehensive, actionable advice on business, technology, and regulatory issues and their impact on financial services. With expertise in banking, insurance, wealth management, and capital markets, we partner with our clients, delivering insights to make their businesses smarter and stronger.


Free Headlines in your E-mail

Every day we send out a free e-mail with the most important headlines of the last 24 hours.

Subscribe now

Keywords: transaction monitoring, transaction fraud, AML, artificial intelligence, data analytics
Categories: Fraud & Financial Crime
Companies:
Countries: World
This article is part of category

Fraud & Financial Crime