Something we said? Don’t leave just yet!

For more information about latest events, news and insights, leave us your email address below.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form
Dismiss
How can predictive analytics improve AML compliance?
Luca Primerano
September 24, 2020

We all know the fight against financial crime is by no means easy. The challenges are seemingly never-ending: trillions of transactions, millions of criminals, complex networks spanning the globe, language barriers, technology barriers, organisational silos, high levels of false positives… not to mention the unprecedented changes associated with the Covid-19 pandemic.

The need to do AML better is not just a social responsibility case – it’s a business case too. Using technology to work smarter and better placed human efforts will improve the results yielded from AML efforts while reducing costs.

Ever increasing compliance requirements are just one reflection of how regulators are rightfully relentless in ensuring regulated entities play their part in combating money laundering, terrorist financing and other financial threats.

The potential of predictive analytics in AML is still relatively untapped but it does offer an intelligent way to strengthen the fight against financial crime while making the job of compliance easier.

What is predictive analytics anyway?

Predictive analytics is all about using machine learning techniques in conjunction with data, including information on historical customer behaviour and events, to predict what will happen in the future. In the case of AML, it is about continuously and automatically comparing the customer’s expected behaviour with their actual behaviour.  This will help determine the risk of a customer committing a financial crime.

Predictive analytics can be used to detect money laundering at an early stage. It is founded on the premise that most customers stick to general behavioural trends over time.

The software uses sophisticated algorithms to track customer behaviour and, when an inconsistency arises between predictions and real events, compliance teams can decide on the most appropriate action.

Current state of play

With many organisations and financial institutions still tied to legacy technology, the adoption of predictive analytics for AML remains in its infancy. Many organisations are still heavily, if not entirely, dependent on a rules-based approach to detecting financial crime. However, while rules-based approaches have an important place, they also have limitations as they are based on known historical patterns. A rules-based approach quite simply cannot detect all fraud and financial crime in a rapidly changing digital world.

Moreover, manual processes and legacy technologies are unable to keep up with the complexity of operating environments and the vast volumes of data being produced.

There are big gaps to fill in terms of not only improving transaction monitoring but also the focus and therefore productivity of investigations.

Discover the risks and challenges of implementing AI for AML - read it here.

Predictive analytics in use

In a Thomson Reuters AI Expert article, Terry Pesce explains how while much of the focus of data analytics has been on uncovering potential business-building opportunities, the ability of this technology to detect warning signs and patterns of potential criminal activity by a financial institution’s customers should not be overlooked.

Clinton Mills’ 2017 article about predictive analytics in fraud and AML in the Journal of Financial Compliance describes how predictive analytics has been successfully used in a wide range of industries and areas, in both application fraud and transaction fraud models. This includes banking and finance, insurance, and telecommunications. Specific areas include applications for credit, credit/debit card issuing, online and mobile banking, merchant and ATM acquiring and, of course, for anti-money laundering.

Why you should use predictive analytics

Predictive analytics arms you with knowledge and, as the old saying goes, knowledge is power.

The intelligence provided by predictive analytics allows you to understand how you expect your customers to behave at any given time – to know what is normal for them. How much is your customer likely to spend this month? How many mobile phones are they expected to buy? With this intelligence you can dot the i’s and cross the t’s by comparing their expected behaviour with their actual behaviour.

This will not only improve anomaly detection. It will also assist in behavioural and trend analysis for a population to further expand understanding of real behaviours and the associated predictions.

Any significant deviation from predicted behaviour is likely to warrant further investigation. This helps ensure human efforts are allocated intelligently and resourcefully.

Other benefits of predictive analytics include better risk identification across a population of customers, and even potential cross-selling of products to customers.

What’s the catch?

While the benefits of predictive analytics are great, there are some challenges to take into consideration:

1: You need quality data – When it comes to getting ahead in AML, data is critical. The predictions made by predictive analytics are only as good as the data that goes in. The system will need a broad range of data, including transactional data, customer static data (such as KYC data), external data (such as Dow Jones or Refinitiv) and macro-economic data. Poor data quality, including old, incomplete or irrelevant information, can lead to unreliable and inaccurate predictions.

2: Choice of model – There are various types of models that can be used for predictive analytics and each has its benefits and problems.  With neural network based models it can be challenging to explain the decision.  Other models may compromise performance or accuracy, or may need more data than you have available to use.

3: Explainability – The aim of predictive analytics is to predict new behaviour and highlight differences.  For AML and financial crime, these predictions need to be clearly explainable to the human teams.  It is not sufficient just to have the prediction. Many predictive analytic systems struggle to give good reasons why the behaviour is different or predicted even in simple cases.  There are techniques that can be applied depending on the level of explanation your use case needs.

Find out how we can improve your compliance processes with predictive analytics

For information, advice or to book a demo on any Napier product, please get in touch with our expert team.

Luca is the Chief AI Officer at Napier, where he focuses on running artificial intelligence programmes partnering with tier-1 universities and clients. Luca has extensive experience in decision automation and digital transformation that he gained at Goldman Sachs, Deutsche Bank and Deloitte and has an MS in engineering from Politecnico di Milano specialising in anomaly detection, data correlation and pattern identification.