The rise of the decision strategy in automated underwriting
Nearly two years in and COVID-19 continues to bring uncertainty to the global market. If the job demand for data scientists is any indication, businesses across all sectors are looking to quantitative analytics for clarity. But what about those financial institutions that don’t have the analytical bench strength or resources to build an analytics team that can deploy, monitor and retrain models at the speed of today’s market? Luckily, much has changed in the field of advanced analytics, and many lenders are finding a competitive edge by following a different “model” of starting with a new strategy.
Two decades ago quantitative analytics was still a relatively new concept in lending. But today we have access to robust data and attributes, and a much broader set of statistical techniques, including machine learning enabled in more mainstream applications. With all this progress, for better or worse, the analytical financial model was elevated to silver-bullet status.
However, a model is not always the end-all, be-all solution for reducing risk. That’s because it relies on historical data to work effectively. It doesn’t take a data scientist to tell you the last two years have been unlike any that came before them. With the pandemic continuing to change the economic picture for consumers by the minute, a model designed for today could in effect be outdated tomorrow.
Honestly, I don’t think it’s really about a model. The “quarter inch hole” you’re trying to buy is the ability to make automated lending decisions within risk tolerance levels. A model can provide a score, but it can’t tell you how to act on that score—it doesn’t culminate in a decision. Thus, a model alone won’t help you reach your portfolio level goals, i.e. minimizing risk or maximizing profitability.
While it’s possible to make lending decisions without a model, it’s not possible to do so without a strategy. That’s why the financial institutions succeeding in today’s market are turning to decision management strategies, modernizing them with a compute-intensive technique called mathematical optimization, and deploying them at scale.
These firms are choosing strategies because of their adaptability. While models often have rigorous governance standards, a strategy can be adapted with relatively little impact to compliance. This helps lenders better adapt to market volatility or new portfolio management goals.
These institutions are also finding improved operations due to the consistency of a decision strategy—strategies essentially allowing team members to operate from the same playbook. In smaller organizations where work is still done manually by a handful of people, a strategy allows for automated processes, like underwriting, so businesses can scale decisioning faster and more effectively.
Finally, lenders are deploying decisioning strategies because of the speed with which they can be leveraged. A custom model can take weeks or even months to build, test, deploy and optimize. Conversely, a strategy can be developed and deployed relatively rapidly, and then adapted on an ongoing basis as needed.
But beyond efficiencies, institutions are learning that a sophisticated, mathematically optimized strategy used with a generic risk score has the potential to drive even more value versus a custom model that is leveraged through a traditional credit strategy, including increasing loan funding rates by 26% one month after implementation and reducing manual reviews by 25%.
Both models and strategies have an important role to play, and each makes the other a more effective tool. But because of their adaptability, consistency and speed, strategies are often the best place to start. And if the efficiencies aren’t enough to convince you, let the numbers and new members do the talking.
Kathleen Maley is VP/analytics product management at Experian, a CUESolutions Silver provider.