Behavioural analytics and AML
February 2021 | TALKINGPOINT | RISK MANAGEMENT
Financier Worldwide Magazine
February 2021 Issue
FW discusses behavioural analytics and anti-money laundering (AML) with Araliya Sammé at Featurespace.
FW: Broadly, could you outline how behavioural analytics can help companies to detect financial crime, such as money laundering?
Sammé: Almost all customers and transactions are genuine. This means that looking for illicit behaviour is like looking for a needle in a haystack. By flipping the problem and using behavioural analytics to truly understand what good or normal behaviour looks like, we are able to better spot anomalies that could indicate criminal activity. By removing the noise and highlighting anomalous behaviour, the haystack is removed, and the most suspicious activity can be investigated first. Criminal organisations are often technologically savvy, using innovative methods of conducting their criminal activity. So, if your approach to prevention and detection is to design rules based on what is deemed to be the latest typology, you will always be one step behind the criminals, who are extremely motivated to act swiftly. Using behavioural analytics enables financial institutions to reduce the number of alerts to sift through. Prioritisation of those alerts means they can focus on the highest risk first, enabling more effective financial crime detection, as well as more efficient and cost-effective controls.
FW: Drilling down, how does behavioural analytics technology work in practice? What kinds of tools can be used to collect and analyse data, to flag aberrant and potentially criminal behaviour?
Sammé: Rather than specific tools, there are various analytical methods that can drive meaningful prediction from data. With the ability to process enormous volumes of data from a multitude of sources, machine learning (ML) provides plentiful opportunities for financial crime teams. ML models are taught what to look out for by using past alerts that were deemed worthwhile, or suspicious. The machines look at historic data to predict the future, spotting patterns in huge swathes of data. In order to spot entirely new suspicious activity, such as human trafficking for example, anomaly detection uses exploration to identify potentially suspicious activity that can be checked by human investigators, continually feeding back to the models. This approach can be combined with rules to form a hybrid system, which feeds into the model to automatically retrain them.
FW: To what extent has the emergence of coronavirus (COVID-19) escalated the threat of criminals looking to exploit weaknesses in the finance system to fulfil their money laundering objectives?
Sammé: The coronavirus (COVID-19) pandemic has greatly increased the threat posed by criminals, ranging from the nature of their activities to the means and methods in which they aim to disguise and process their illicit gains. Inherently with any type of socioeconomic challenge, you see an increase in criminal activity with the primary goal of exploiting people’s fears, pressures and lack of financial stability. As lockdowns were implemented worldwide, spending habits changed overnight. Thresholds and rules that many financial institutions relied on to flag suspicious behaviour were no longer relevant. Even ML models that relied upon manual retuning were ineffective, as these retunes can take months. Automatically adaptive ML models enable financial institutions to maintain consistent decisioning that catches as much financial crime as pre-pandemic.
“Innovation must be matched with innovation if the industry is to successfully chip away at the enormous levels of laundered money.”
FW: With false positives the leading cause of delayed financial crime detection, how can companies go about setting thresholds to separate subtle deviations in normal activities from genuine red flags?
Sammé: Thresholds alone are not enough to detect the constantly shifting and changing approaches of criminals. They are rigid, with siloed data points, and are unable to adapt to keep up with savvy criminals. The key to achieving false positive reduction is to be able to understand better the data you have and be able, with greater accuracy, to attribute specific behaviours to certain actions. This is a core concept within adaptive behavioural analytics. ML models create profiles for multiple entities, such as customers, accounts, merchants and policies, not only at an individual level but also between peers, each specific to the use case – from retail banking to correspondent banking, merchant acquirers or even life insurance, resulting in a far more accurate risk score than more conventional solutions. Christmas Eve perfectly illustrates the impact ML can have on false positives. Rules-focused systems reliant on thresholds would trigger an enormous number of alerts, as spending patterns changed with last-minute gift purchases. However, an ML model would profile individual behaviour against that of the customer’s peers, realise that the behaviour was unexceptional, and not trigger an unnecessary alert.
FW: To what extent can behavioural analytics assist companies with anti-money laundering (AML) regulatory compliance?
Sammé: Behavioural analytics enables financial institutions to better understand their risk exposure and address it directly, which is the main goal of any compliance programme. It does this by allowing wider risk coverage and erasing blind spots with anomaly detection, surfacing unknown threats. In addition, financial crime teams are more efficient, working with fewer false positives and therefore a clearer picture of what activity is suspicious. Add to this the speed at which suspicious behaviour is spotted, and you have a really strong compliance programme. ML can only be a magic wand with which to eliminate financial crime when combined with input from financial crime experts, whether or not they are in-house at financial institutions.
FW: What advice would you offer to companies looking to implement behavioural analytics as a core component of their financial security framework?
Sammé: Knowledge is power. Companies should do research and utilise the plethora of articles, webinars, white papers, videos and events that explain how ML facilitates better detection and prevention of financial crime. Speaking to peers is absolutely key for knowledge sharing and feeling part of a community. If still unsure, beginning with a proof of concept (POC) is a fantastic way to test ML with your institution’s specific challenges and understand the benefits that can be achieved. With POCs it is important to be clear on your success criteria and to collaborate with both data scientists and financial crime experts.
FW: Looking ahead, do you expect to see increasing demand for behavioural analytics tools? To what extent could this be a game changer for companies’ AML efforts?
Sammé: We have seen a strong acceleration in demand across the industry for more data-driven and innovative solutions, particularly over the last five years. This really is a crucial concept. There is often a misconception about criminal and terror organisations and the nature of how they conduct their business. Criminal organisations, at their core, are like any other: innovative, well-organised, well-funded and mission-driven. Innovation must be matched with innovation if the industry is to successfully chip away at the enormous levels of laundered money.
Araliya Sammé joined Featurespace in 2016, bringing deep subject matter expertise in all areas of financial crime to customers and finance industry conferences around the world. Prior to joining Featurespace, Ms Sammé worked as a management consultant at both Deloitte and EY and advised at multiple global financial institutions on financial crime and anti-money laundering. She can be contacted by email: araliya.samme@featurespace.co.uk.
© Financier Worldwide