Banks and other creditors have used formulas to model creditworthiness for decades, a successful strategy centered in effective risk assessment. External trends aren’t going to disrupt this. Lending will always be a risk game.
But data and technology are refining the art and science of risk assessment in ways that are critically important for banks to understand as they work to maintain and grow market share in an increasingly competitive industry.
The amount of data banks own and use is exploding. This is driving a realization that credit risk modeling needs to adjust in response because it has direct impact on compliance, revenue, reputation, and competitiveness.
Compliance: Post-financial crisis regulations such as Basel II, III, and Dodd-Frank have added new forms of credit risk management, such as stress testing, that require more frequent and more detailed reporting.
Revenue: Customer expectations are changing. They want banks to evaluate their credit worthiness on a continuous basis and be proactive in communications, including pre-approved loan amount information.
Reputation: Errors in execution or reporting are more harmful, as customers and society have a diminishing tolerance for mistakes. Adding more contextual data to credit risk modeling improves the efficacy of decisions and the ability to justify decisions.
Competitiveness: Banking has been historically slow to change, an unhealthy trait in an era dominated by fast-moving innovation. Not modernizing credit risk modeling is a huge risk by itself considering the value it adds to competitors that do, like the Fintechs.
A data management platform is the first requirement of modern credit risk modeling.
We see many banks that have invested in data lakes, which are valuable to the risk function in general because they are able to possess high volumes of structured, unstructured, and semi-structured data. This allows for near-real-time data ingestion and processing, which is needed more and more across the banking enterprise.
Data lakes also add significant processing power for soft analytics and allow for new data processing techniques. They can also save money because they make it possible to store data at low cost while enabling banks to store more and more diverse data. All that comes together in one single repository.
On top of the data are analytical capabilities needed to build the quantitative models, and a growing number of qualitive models for risk scoring and default detection for credit risk.
Qualitative credit risk modeling is becoming more common to fill the need to serve personalized offers to customers on an ongoing basis and to expand to new markets and customers.
For example, if you look at traditional credit scoring for retail loan approval, credit risk is assessed most of the time on a customer's financial history. Adding more data and analytics capabilities provides banks the ability to build richer models by factoring in new types of data, i.e., demographic data, financial data, employment data, and behavioral data. This is where data lakes come into play: They feed in contextual data that expands possibilities, as well as confidence in the possibilities.
The same is true for commercial lending. Current models assess things like sales margins, liquidity ratios, and total debts; but there's an opportunity to factor in new data types as well. Things like capacity utilization, social capital, social media, family records, and other archives can be meaningful.
While the analytics engines of modern credit risk modeling are automated and could include machine learning (ML), human input using soft analytics are still important in credit risk modeling to make sure credit decisions are transparent and legal. Second, not all data-rich credit risk models are the same, and one organization might emphasize one input more than another depending upon strategy and other factors.
One of the things that stands in the way of adoption of modern credit risk modeling is outdated technology. Cloud is essential to provide the storage and computing power needed for resources such as data lakes and real-time analytics. Cloud resources also can be rapidly scaled up or down. Most of the time, current infrastructures can't do.
Using hardware and software that is engineered specifically for the bank’s requirements provide additional advantages. Typically, engineered system stand out in performance, scalability, fault tolerance, and manageability — all important aspects in modern credit risk modeling. To add to that, using an autonomous system such as Oracle’s autonomous database, which automatically tunes and manages itself with no downtime, could mean faster response times and quicker decisions.
From a performance point of view, there's more data and more models, but also more expectations from regulators and customers. From all angles, there's the requirement to deliver results faster, so performance is certainly a very important factor.
Scalability. With new types of data constantly being added, the bank’s underlying infrastructure needs to rapidly scale up. Risk modelers will need data lakes to experiment with their models and test out new data sources. The scalability of those environments is getting more and more important for data scientist and data analyst productivity. Both in the development part, as well as the production part, the ability to scale very fast with the least cost is becoming more important.
Fault tolerance. Credit risk management is a business-critical function. At any point in time, banks should be able to deliver results or credit positions to customers. Engineered systems help to effectively do that, as well as regulatory reporting and stress testing.
Manageability. Modern credit risk modeling systems become more and more complex as time goes by, so manageability of the whole platform is an important factor; and the autonomous capabilities of Oracle Cloud Services are an important part of ongoing manageability.
Engineered systems can be a good place to start cloud modernization. These systems can be part of a hybrid cloud architecture and/or provide cloud-ready assets when a bank is ready to move more workloads into the cloud.
And while it seems that some workloads will remain on-premises for the foreseeable future, banks are taking steps to modernizing their on-premises infrastructure to achieve cloud benefits while taking into account the compliance and risk management profiles they need to satisfy their customers. HDFC Bank in India chose Oracle Exadata to help increase daily financial reporting speeds by 4x and liquidity risk reporting speeds 7x—moving from 52 hours to just under eight hours. Replacing its legacy infrastructure has helped it meet a new, more demanding service-level agreement and improved overall credit risk management.
Learn more about how Oracle can provide ready resources that are engineered for the cloud.
Wiljo van Beek is the Director for Analytics and Big Data Banking & Insurance, Oracle EMEA.