Choice and ease of payment is great news for customers, but while the rise of digital transactions brings new value for banks and processors, it also poses challenges in how they offer excellent services at scale.
Cleber Martins, Head of Payments Intelligence, and Damon Madden, Principal Fraud Consultant MEASA, recently discussed with me the impact of real-time and open on the Indian payments ecosystem, and specifically what it means for fraud prevention.
RT: What fraud prevention challenges has India faced as a result of the incredible adoption of real-time payments, driven in large part by UPI?
Cleber Martins: The challenge for financial institutions in any fast-growing digital payments market is scalability. India’s banks and processors were already used to providing #sleepatnightability for very high volumes of transactions, but the drive towards cashlessness, in combination with the advent of digital overlay services for RTP, has caused explosive growth. Scalability is a critical non-functional requirement (NFR) for any payments player. When it comes to fraud prevention, those same strategies and controls have to scale with the volume of transactions, particularly in a real-time ecosystem.
Damon Madden: It’s important to remember that the scalability challenge is not just for UPI transactions. They account for a huge portion of the volume, but the bigger challenge for India’s financial institutions is managing the fraud ecosystem. India’s experience really demonstrates the criticality of an enterprise fraud solution that can scale across all payment types, and react at the speed of the market. The proliferation of digital overlay services means that there are constantly new data streams to account for in fraud prevention. The velocity, volume, value, variety, and veracity of that data demands rapid speed of correlation and decision-making in fraud systems: It’s an obvious use case for machine learning.
RT: How can India’s financial institutions leverage machine learning to manage fraud across this complex ecosystem with its huge transaction volumes?
DM: Machine learning is optimal for exactly this kind of environment. It has real-time capabilities that can enable an organization to rapidly react to marketplace trends. The speed of training, and absorption of that information, accelerates the deployment of new strategies to market. With machine learning, banks and processors can react in real-time to protect every transaction, as well as train models on macro trend analysis to ensure that new patterns and trends are taken into account.
Without machine learning, it’s unlikely that any organization could keep pace with the changes in fraud; as India experiences rapid innovation in real-time payments fraud threats will accelerate at the same pace, so strategies must quickly evolve to combat them.
CM: It’s not just the ability to stop the fraud, it’s also the operational efficiency in how you deploy and manage machine learning against your overall fraud strategy. The easiest way for fraud prevention to scale against transactions is to decline more, but increased false positive rates are bad for customer experience and operational effectiveness. Organizations cannot afford to let their fraud losses and overheads scale at the same pace as digital transactions; they cannot afford to increase false positives, or it will negatively impact their brand reputation and customer experience.
Banks and processors need to be able to correlate big data without relying on human beings. For this to truly deliver operational efficiency, it has to drive down false positive rates. It’s better to reserve human intelligence for training the machine learning models based on their experience and knowledge in the market. Providing feedback from the analysts is critical to the reinforcement learning techniques that are applied in machine learning.”
RT: In the Indian market, where real-time payments and digital overlay services are now table stakes, how can the country’s financial institutions use machine learning to differentiate themselves and win market share?
CM: India’s banks and processors are rising to the challenge of scalability. The customer experience is paramount, and relies on quality of service. The government mandates around cashlessness and real-time payments mean that every single consumer, merchant, corporate and payments player is adopting these services. Customers have more choice than ever and can move to any other provider. Your customer experience becomes your differentiator because happy customers are loyal. Differentiating on customer experience at scale relies on machine learning.
Many of these new digital payments are no longer made via a single plastic card but instead via electronic instruments with more capabilities than cards. In this new payments ecosystem, each customer has a range of devices and overlay services, and generates rich data for their payments provider to leverage for the benefit of the customer. This data can be used to provide more robust security and further differentiate on customer experience.
DM: Fraud strategies can leverage device information, behavioral biometrics data and location to assess the transaction risk. But machine learning is necessary to process and correlate the data points in real-time. As account providers integrate machine learning into their fraud prevention strategies and solutions, they can begin to leverage those learnings against related services for customers. For example, in India the issuing banks can look to support customers to make better choices about overlay services based on fraud trends data from across their account holder base. This kind of advice will become more and more central to retaining the primary relationship with customers in an open ecosystem.
CM: The payments ecosystem needs to be aware that if they do nothing to modernize their fraud prevention, the fraud and operational losses they currently experience will only rise as transaction volumes continue to grow. They need solutions that allow the organization to focus on capturing market share and bringing new services to market so they can capitalize on that transaction growth. Machine learning is the solution to that challenge.