Updating credit risk models doesn’t have to be a lengthy, delay-ridden process
By
Credit risk vectors are constantly in flux, with everything from regulatory updates to new data sources impacting business decisioning needs. To
However, tech resource limitations and legacy systems mean that critical changes can take weeks or months to deploy and in the time it takes financial institutions to modify credit models, more nimble emerging players and fintech providers have already updated their risk models—making them better able to compete against traditional financial institutions.
The challenge is that the typical financial institution loan origination process is monolithic and changing any step in the process—including the credit risk model—is difficult, costly, and time-consuming.
Instead, what in theory should be the tweaking and fine-tuning of the models turns into a full-blown IT project, along with the associated costs and inherent delays. If the credit risk model was a patient, it would be undergoing open heart surgery for an ingrown toenail.
The Open-Heart Surgery Approach
There’s a multitude of reasons why updating credit models has traditionally required that banks and credit unions embark on a long, expensive development process. For one, data comes from a myriad of external sources and must be aggregated, rationalized, and transformed into the same format before the financial institution can feed that data into a credit risk model. Often the financial institution is at the mercy of their vendor or professional services provider to prepare the data. In essence, the data is held hostage until the vendor makes the data both accessible and usable.
Even if the data exists within the financial institution, there’s an excellent chance that it’s dispersed in multiple, siloed systems and may still not be in the correct format. And even if it is in the right format, it’s probably not easily accessible without following stringent data governance rules.
And once the financial institution finally gets the data, in a format they can use, they find that integrating sophisticated, real-time data analytics with legacy systems is almost impossible.
Now add in the challenge of the lengthy, delay-ridden software development lifecycle. Forget just tweaking a few variables or calculations. Changing the risk model means requirements gathering and analysis, design, coding, testing, and then deployment.
Another challenge is that credit models themselves are typically written in programming languages that are not readable by bank and credit union systems. Data scientists prefer to use open source programming languages like
The Less Invasive, Fine-tuning Approach
If given a choice between open heart surgery or a less invasive procedure that could achieve the same outcome yet is done quickly in a physician’s office, it’s pretty clear which approach most people would choose. While surgical procedures haven’t advanced far enough yet that a simple technique can replace full-blown surgery for most patients with heart disease, the same isn’t true of updating credit risk models.
Even though surgeons work on grafting an artery or repairing damaged tissue, the entire heart is exposed during open heart surgery. The surgery can take hours, and the patient recovery is lengthy. No wonder medicine has pursued far less invasive options.
If we continue the analogy of open heart surgery to updating credit risk models, wouldn’t it be better to simply change the model in a less invasive procedure rather than exposing the entire loan origination process? That’s where microservices come in. Microservices break a monolithic process like loan origination into discrete services that do one thing really well.
For example, a credit scorecard could be a service used by the loan origination system (LOS), as the LOS only cares about capturing the credit score to underwrite the loan—not how the scorecard calculates the score. With microservices you can update the scoring model without impacting the credit underwriting process and quickly adapt to a changing risk landscape, regulatory updates, and your competition.
Credit risk models are just one use case for microservices. Any function in the loan origination process, such as sanctions screening or Know Your Customer (KYC) verification or machine learning models for predictive analytics, can be exposed as a microservice and easily updated as data sources, regulations, or analytical techniques change.
Less Time, More Opportunity
Microservices allow you to be much more agile and react quickly. Because you can reuse microservices throughout the organization, you’ll need fewer resources and decrease costs for IT and professional services.
But perhaps the most interesting promise of microservices is how this architecture can transform a financial institution from a product-centric organization to a customer-centric one. Financial institutions have traditionally focused on selling products to customers by starting with the product design. But the products they develop may not actually be useful to a large enough segment of customers to make the product profitable. Or, the long product development lifecycle means that by the time the product comes to market, customer demand has shifted. The result is lost revenue and less satisfied customers.
Microservices support a customer-focused strategy. With the agility and flexibility provided by microservices, you can analyze customer behaviors and attributes, such as income and payment history, in real time and design products that can best suit your customers by delivering a more personalized product.
Microservices means you don’t need to redesign and recode entire processes. Instead, you can fine-tune credit risk models rather than embarking on a lengthy, costly, and risky open-heart surgery.
About Provenir
Provenir makes risk analytics faster and simpler for financial institutions. To learn more, visit
