The new science of credit risk

Financial institutions have long relied on rigorous criteria, rigid rules, and hard cuts to filter customers before using analytics to manage or avoid credit losses. This traditional approach is typically rooted in a generalized yet narrow view of customers and their behaviours, which often leads banks to decline the perfectly legitimate, low-risk requests of good customers. Taking a more holistic view of each customer and his or her needs or developing new products for marginal customers would entail completely rebuilding the decision framework.

But this black-and-white approach to credit approvals and straightforward decision-making can lead to missed revenue and frustrated customers—who, until recently, have had few lending alternatives. Fintechs and other new players are adeptly using machine learning with more expansive data sets to make quicker lending decisions. They’re connecting people to capital more quickly, more effectively, and more cheaply than traditional banks and other lenders.

To stay in business, traditional organizations are going to have to transform how they make lending decisions.

Personalized lending decisions

For the most part, and for various reasons, lending decisions fail to capitalize on one of the greatest resources available to any financial institution: the wealth of customer data they have at their disposal across the enterprise. The time is ripe to change that. The low cost of cloud-based storage and computing power enables financial institutions to more easily deploy machine learning and other technologies to develop much more detailed and complex insights into each customer. Over time, these insights can be refined further to create a highly nuanced understanding of each customer, which can then be used to make real-time lending decisions tailored to the individual that delivers a superior customer experience and opens up new revenue opportunities.

That’s how credit risk management begins to transform into credit risk science—it’s lending decision-making for insight-driven organizations.

Credit risk science isn’t simply about applying traditional analytics methods to customer data, since these methods focus on correlations or narrow insights drawn from a subset of the customer base. More data produces increasingly accurate predictions, but only for standard, habitual behaviours, such as a customer’s weekly gas and grocery purchases. This approach isn’t as effective at dealing with individual customers’ unique and rare spending needs, like an unexpected home repair or finding a treasure at the antique market.

Relevance analytics: the next level of insights

On the other hand, relevance analytics use cognitive memory and model-based algorithms to capture and integrate the rare behaviours that are so important to understanding specific customers, and ensures these insights are incorporated into a fuller, deeper picture of each individual customer. This enables organizations to more effectively differentiate between customers who may be virtually indistinguishable day by day yet have unique needs at different points in time.

By constantly scanning customer activity at a more detailed level, identifying spending needs and spending sequences of non-standard activities as well as habitual ones, relevance analytics allows financial institutions to better understand and respond to the credit authorization needs of individual customers.

Credit risk science has already shown its value, from the top line to the bottom line, in the lending industry. And right now it’s more feasible than ever before to achieve resilient growth by applying data-driven insights to lending decisions.