Introduction

First Franklin is one of the oldest and most established lenders in the South East US. Founded in 1941 in Toccoa, Georgia, this 4th generation family owned and operated lender spans over 400 branches and 10 states and is expanding.

They have dispersed over 6 million loans and over $13 billion dollars of loan volume since inception. With revenue topping $300 million annually, they have an aggressive digital and online expansion strategy leveraging the latest technology offered by LendAPI for faster and smoother transactions.

The Legacy Wall

As First Franklin Corporation enters into an expansion phase, their credit risk underwriting has become ever more sophisticated. From revitalizing their enterprise data warehouse to sophisticated machine learning credit risk management models, their existing lending stack has been pushed to the limit.

Older Loan Originations Systems are sometimes hard coded, impossible to make changes and unable to handle multi-variate and non-linear underwriting models. That’s where LendAPI’s decision engine and model run time environment comes into play. 

Deployment of rules, models and new data providers has proven to be impossible with legacy loan origination systems. Sometimes small changes could take months and are very costly for lenders such as First Franklin to continue to offer best in class financial services products. 

Scalability has also become an issue with legacy LOS platforms where they are unable to process more than 1 or 2 transactions at the same time. For 400 branches to simultaneously access multiple data sources and perform complex calculations, a modern system designed to take on parallel processing is not just a want, it’s becoming a must have to launch new products, meet new customers and launch new branches

The Shift

We engaged 1st Franklin a few months ago and was able to launch their newly formulated credit risk underwriting model within a few days. After several iterations of refinement, we’ve launched the new underwriting decisioning with the most sophisticated machine learning model to control for losses and to grow responsibly.

We introduced new data sources, new derived variables and new scores to exercise new credit risk models created by First Franklin’s data science team. 

First Franklin is able to implement all of their sophisticated non-linear models without dumbing it down for their legacy loan origination system. This way, they can stay at the forefront of credit risk management without sacrificing the robustness of their underwriting model.

In addition, implementing a new version of the model has become a day worth of work as opposed to months if not years. The credit risk team can self-serve and implement new versions of the model themselves with all of the visual user interface made available through LendAPI.

They can easily create derived variables, map additional data sources, attributes and scores as inputs to their model without having to wait for technologies or vendors to intervene. This greatly increased First Franklin’s ability to experiment, perform A/B testing and stay ahead of the curve with respect to their customer demand and competition.

Our API first architecture allows First Franklin’s engineering team to intercede their models at any stage of the underwriting process. We can implement a variety of different models and underwriting criteria for a whole host of situations with an API endpoint to provide maximum flexibility for their engineers to integrate and deploy with any enterprise system.

The two teams worked together hand in hand to map all of the latest variables needed for their machine learning model. Leveraging our no-code decision engine to run these models without any IT involvement and tested high volume transactions within days. It was one of the fastest implementation projects we’ve done to date.

The Results

First Franklin is now able to run thousands of transactions per minute with the latest machine learning model. They now have access to sophisticated tools to dial in both approval rates and default rates with their latest probability based models at the same time, reducing their reliance on external engineering vendors and internal staff.

Legacy systems aren't just old; they are restrictive. For a company like First Franklin, the software was no longer a tool, it was a boundary. LendAPI turned the software back into a conduit for their intellectual property (the risk model).