Skip to content

Why Banks Must Democratize Machine Learning for Fraud Prevention and Payments Intelligence

banks democratize

Banks are already actively on the path to digital transformation, considering new technologies, new customer experiences and new business models. A critical piece of this digital transformation centers on better understanding the wealth of data within the banks’ systems and mining it for improved customer insight. In the New Payments Ecosystem, data is as valuable to the bank and its customers as the deposits held in their accounts, and it should be protected, and leveraged for the benefit of the customer.

Machine learning is well acknowledged as one of the new technologies needed to interrogate the vast volume and variety of data within the banks’ operations. Humans cannot possibly process the data at the same speed and scale. However, a bank-wide machine learning project is a huge undertaking, requiring major investment in technology and human resources. Some banks feel disillusioned by these very projects that, in the past, have failed to deliver a return on investment. But there is an area of the banks’ business where smaller, more tactical machine learning projects can be rapidly deployed in a way that delivers value to the business from day one: fraud prevention.

Machine learning in fraud prevention

In the fraud shop, there is a wealth of human expertise that can be used to guide new machine learning models. The current barrier to delivering this often lies in the solutions themselves. They require data scientists to create the initial models. A fraud analyst knows that a specific correlation between transaction types in a sequence is a strong fraud indicator, but the data scientist will need many more interactions to draw the same conclusion. The machine learning model is only as good as the signals it is given. Solutions that enable business users (the fraud experts) to input the initial correlations will deliver results faster in terms of identifying new correlations across different data sets. This ‘democratization of machine learning’ empowers business users to ‘download’ their knowledge and experience into the models. The machine learning models depend on fraud occurrences in order to learn, but fraud is actually a relatively small percentage of successful transactions. Fraud analysts can direct machine learning to investigate and improve in areas where fraud has not yet reached critical mass, but the expert knows that this is a potential ‘greenfield’ for fraud. It’s about bringing machine learning closer to the user and their instincts, rather than bringing the user to the machine learning. In this way, the organization can shorten the time to Return on Investment (ROI) of their machine learning projects.

Democratizing machine learning

The democratization of machine learning can be used to support business users to express their instinctive decisions around desirable traits for transactions or customer behaviours, even if these are not yet fraudulent or highly indicative of fraud. Feedback on these kinds of instincts will aid the machine learning model to fine tune itself, and improve accuracy and consistency in identifying more complex fraud indicators. Business users are closer to the models; they have transparent views of the model and can apply strategies and controls to best leverage the outcome of the intelligence from the models. Continuous involvement of business users is key to developing the machine learning model over time. They will still input to the model and investigate the output as the model generates intelligence and use their human expertise to confirm fraud instances. But they can also combine their omnichannel customer view, with insights correctly generated by the model. If the human intelligence confirms the insights, these can be fed back into the model: if the model consistently flags a correlation between data sets as potential fraud, and the analysts consistently confirm this as fraud, then a strategy and control based on this information can be added to underpin the model. These can be used to automate the decisions that business users have consistently made, and reduce the need for auto actions that impact the customer experience. It’s about managing that customer experience with the arsenal available to them, and turning machine learning for fraud prevention toward a customer-centric financial services relationship.

The market recognizes that machine learning as a technology has passed its peak, and is now establishing itself across organizations. This means that the business must bring that technology closer to its business case to drive value from what is fast becoming a commodity. In the payments fraud context, I expect this to materialize in the democratization of machine learning. I see a future where the business user, in this case the fraud analyst, will have increasing influence on the evolution of models, in order to deliver the best possible customer experience.

Discover more about leveraging machine learning for payments intelligence. Listen to this webinar from Marc Trepanier, Principal Fraud Consultant at ACI Worldwide, and Julie Conroy, Research Director at Aite Group.