How AI Empowers Credit Access For Thin Credit Individuals

Introduction 

Lease-to-own financing options open up access and purchasing power for those with bad credit or no credit. In the US, Shield Leasing offers simple, straight-forward options to help automobile owners get the tires, wheels, and minor auto repairs needed to keep their vehicles on the road.  Shield Leasing’s brand promise is an easy application process with instant approval for applicants with low to no credit.

‍

Challenges Faced

‍

1. The applicant pool was vast and diverse with low to no credit, so the data used to evaluate credit risk had to be orchestrated from many different legacy data systems - traditional and non-traditional - in different formats - APIs, databases and spreadsheets. Integration with legacy is challenging and time-consuming.
2. Easy application process with instant approvals meant Shield Leasing had to do more with less data and make decisions in real-time. ​​The accuracy and completeness of the data used in the models are crucial for their reliability.
3. Shield Leasing was scaling their business so the applicant profile would change over time, so migration from a static rule based decision process to a dynamic AI based decision engine was an important goal.
4. Shield Leasing wanted to scale and modernize their 360 degree view of customers and business using modern BI applications. 

‍

Solution Implemented

‍

‍Data orchestration 

Internal and external data from multiple sources - third-party and internal and user provided - were orchestrated and stitched together and analyzed to better understand the applicants , by studying the distributions, patterns and anomalies in the data. 

Creating features specific to risk evaluation 

The better the data that is provided to models to make decisions, the better the decisions. To gather the most relevant information to train the model, additional features like ratios, velocities, and frequency counters, were created from the available input data. For example, standard features like debt/income ratio or non-traditional features like email trust. This was done seamlessly using the AutoAI capabilities of the RapidCanvas platform. 

Automated modeling and explainable AI 

The AutoAI platform automated the creation of the best possible model to predict, at the time of credit application, which applications are risky. With this white box approach, the internal working of the model and the importance of each factor used for prediction can be easily explained. In situations involving credit risk, it's important to understand not only if someone is risky but also why they are risky. Explainability is important for ensuring accountability, fairness, and transparency in automated decision-making systems.

What-if Analysis: Credit evaluation depends on individual applicant profiles as well as the macro economic environment. It is important to be able to simulate ‘What if?’ situations. Play with different features and find how they impact predictions.

Comprehensive business intelligence application  

Interactive data apps were generated for business users to review credit predictions and make data-driven decisions. With increased visibility into the risk profile of each applicant, the Shield Leasing team was able to better understand the factors that influenced credit and trends arising from the data. 

Continuous model updates 

With an ever-increasing pool of applicants and changing trends, the model is continually updated to ensure effective predictions are always available for the team at Shield Leasing.

‍

Results and Benefits:

‍

Ability to scale while reinforcing brand promise

Shield Leasing’s brand promise is an easy application process with instant approval for applicants with low to no credit. AI and machine learning allowed Shield Leasing to scale its customer base while ensuring the brand promise could be reinforced.

‍

Increased revenue 

Shield was able to detect risky credit applications and positively impact their revenue, to the tune of 10%. 

‍

Improved  credit risk management  

With the insights provided using dynamic real-time machine learning models to predict future outcomes, Shield Leasing could better assess and manage risk both during the credit application and the ongoing payback period.

‍

Deeper customer insights 

The interactive data apps gave the Shield Leasing team a deeper understanding of customer insights. The data apps showcase a 360-degree view of each customer, segment and cluster of users to better understand groups of customers with similar patterns and behaviors, and to analyze and explore alternative outcomes.

‍