The Financial Brand Insights - Spring 2023

• Relying upon a limited set of data sources • Leveraging an outdated set of predictive variables • Using simple analytical engines that cannot accommodate complex business rules • Depending heavily upon subjective assessments from underwriters • Operating with inflexible, legacy models that have been patched over time • Addressing concerns over the length of time and cost to implement • Being unprepared for regulatory reviews of the predictive variables in the model These challenges can be significant obstacles to risk management professionals who try to rebuild and deploy new credit scoring models inside their organizations. They are real hindrances and should not be minimized. However, the benefits of surmounting challenges should not be discounted either. This article explores four best practices lenders should follow when building new credit scoring models or upgrading existing ones. BEST PRACTICE 1 Centralize connectivity to data sources to more easily develop and deploy new credit scoring models Auto, mortgage and credit card lenders rely on one or more of the three major credit bureaus to complete the loan origination process. The credit bureaus, as well as alternative data sources, provide the raw data needed to generate the predictive attributes required by best-in-class custom credit scoring models. These custom credit scoring models typically live within a larger digital ecosystem that might include: • A legacy loan origination system (LOS) to capture loan application data and process the loan request • A data access layer that will retrieve data files from all the major credit bureaus and other data sources with a single request • A product approval and pricing decision engine that will leverage the scores • Additional downstream services needed to book the loan

By Michael Sogomonian at Digital Matrix Systems

As advances continue in big data analytics, opportu- nities exist for lenders to improve how they design and deploy the credit decisioning models that underpin their processes. New, high-performance models built with state-of-the-art predictive attri- butes allow lenders to define lending criteria more precisely and thus enhance their ability to approve more creditworthy customers. At the same time, they can decline requests from consumers who are either not creditworthy or cannot afford more debt. Lenders that have deployed such new models into their production stream have already increased revenue, reduced loss rates and made considerable efficiency gains. Using a well-designed infrastructure supports the rapid design and deployment of these high-performance models. But some lenders struggle with transitioning to a more advanced credit model that can leverage data across multiple bureaus or alternate data sources. They face significant technology, capability and cultural hurdles. These hurdles include: Big data analytics offer lenders new ways to evaluate borrowers’ creditworthiness and enhance lending, but implementing new, high-performance models is complicated and challenging. Ensuring they have a well-designed infrastructure is a critical step for lenders that want to adopt new credit scoring models or upgrade the ones they have.

Rapid Design and Deployment of New Credit Scoring Models

Lenders face significant technology, capability and cultural hurdles when transitioning to an advanced credit model that can leverage data across multiple sources.




Powered by