by Thor Olavsrud

Toyota turns to AI to speed credit decisioning

Sep 21, 20215 mins
Artificial IntelligenceData ScienceIT Leadership

Toyota Financial Services' Intelligent Financing Decision Engine leverages machine learning and massive data sets to accelerate credit decisioning and improve customer and dealer experience.

Conceptual trend lines track + monitor data analytics [forecasting / future / what's next]
Credit: SolarSeven / Getty Images

In the world of automotive financial services, both dealers and customers are looking for decision-making speed. To gain a competitive advantage, Toyota Financial Services (TFS) has turned to AI to help it increase automated credit decisioning, improve customer and dealer experience, and efficiently manage consumer credit risk, all while adhering to fair lending practices.

“Our goal is to provide best-in-class customer and dealer experiences,” says Bharadwaj Gopal, domain information officer at Toyota Financial Services. “This required improving our credit decision-making speed and higher automation while managing our credit risk and meeting compliance standards.”

Based on dealer and customer feedback, TFS started working on its Intelligent Financing Decision Engine (IFDE) in June 2019. IFDE is a cloud-native, scalable, loan origination decision engine for credit risk analysis in customer vehicle financing. The idea was to create a state-of-the-art AI decision model that utilized targeted premier credit attributes identified by TFS to craft algorithms capable of delivering sub-second results. The project has earned TFS a FutureEdge 50 Award for applications of emerging technologies.

Driving the process with cross-functional collaboration

TFS IT partnered with the company’s consumer credit risk business unit to articulate the business case for IFDE based on real data, Gopal says.

“Over one million historical consumer loan applications were evaluated, and a swap set analysis was performed to gauge the effectiveness of the new models deployed in IFDE and to estimate the increase in straight through processing and decrease in customer charge offs,” he adds. 

Bharadwaj Gopal, domain information officer at Toyota Financial Services Toyota Financial Services

Bharadwaj Gopal, domain information officer, Toyota Financial Services

TFS built the engine on Amazon Elastic Kubernetes Service (EKS) and leveraged campaign/challenger decision management testing to understand how the new model was responding. The team also employed automated regression and load testing tools for scalability testing, which Gopal calls out as essential given that the tool had to be capable of handling some of the highest loan volumes in the industry.

Requirements were mapped out in a series of sessions involving a cross-functional team of stakeholders from consumer credit risk, sales, loan originations, dealer experience, credit analysts, compliance, data science, and agile application teams. Under the joint leadership of a business product owner and a technical product owner, the team gathered user stories and worked iteratively through a continuous development cycle using two-week sprints.

“The objective was to gather the requirements right from multiple stakeholders and catch gaps early to improve the product based on stakeholder feedback,” Gopal says.

TFS IT uses what it calls Agile Business Capability (ABC) Digital Factories, which are smaller, agile IT teams managed by a technical product owner. More than 10 such teams worked together to create IFDE, including data science, application development, cloud engineering, API services, enterprise data platform, DevOps, information security, and others.

The teams developed and deployed IFDE in 10 months. TFS launched the engine with its first dealer at the end of 2019 and began rolling it out to 2,000-plus dealers in 2020.

“There were some challenges around prioritization of these individual components being developed by horizontal teams, but continuous collaboration at all levels was crucial to manage this challenge,” Gopal says.

The art of the possible

A second challenge was the technology. Gopal says the decision engine provides credit risk analysis for more than 200,000 vehicle financing applications monthly. Building, testing, and scaling it for that kind of volume was challenging, but campaign/challenger decision management testing was key.

“IFDE has allowed us to push some boundaries within TFS and demonstrates the art of the possible,” Gopal says. “But to get here, some of the aspects that were crucial to the success of the product had a steep learning curve: to understand from experience the right level of engagement with supporting teams, to adjust our internal team’s expertise and strength based on business stakeholder feedback, and to adopt an agile mindset with a continuous improvement philosophy.”

Gopal says IFDE has improved TFS’s lending auto decisioning rate by more than 20% (from under 50% to over 60%). The engine has also helped reduce charge offs (when customers are unable to make payments and return a vehicle) to under 0.3%.

“We now have a platform to continuously iterate and improve our auto decisioning rate and manage our credit risk better,” Gopal says. “To date, IFDE has analyzed over 11 million credit applications with sub-second response times and is planned to handle credit applications from multiple tenants.”

Gopal says his key learnings from the project include:

  • Getting leadership buy-in early by presenting data insights and metrics is crucial to communicating the value that an initiative will bring to the company.
  • Identifying key stakeholders, adopting an agile mindset, and working with a cross-functional team that engages with stakeholders creates transparency and builds trust in project execution and maturity.
  • Spearheading a cloud-first strategy and restructuring your technical teams to include a mix of subject matter experts, cloud specialists, data scientists, and open source and DevOps enthusiasts helps advance a project with speed and agility.