Mid-size lender achieves 7% increase in approval rate using TruIQ Custom Models*

Scenario

A mid-size lender wanted to understand what scores might be best suited for its personal loan portfolio, and if a new custom model would drive a lift in approvals. With a reputation for data science expertise and impact in model development, the lender turned to TransUnion's Advanced Analytics Consulting team for a full assessment of their loan origination models.

The solution

To get started, TransUnion’s Advanced Analytics Consulting team conducted an initial assessment of the lender’s business objectives. They learned about the lender’s target segments, which helped assess the suitability of their historical application data samples and existing credit policies. Then, to get an estimate of the unobserved risk of declined and not booked applications, TransUnion architected a multi layered reject inference (RI) process that combined policy exclusions, bureau-based peer proxies and machine learning models. From this assessment, the team discovered the existing model was not materially outperforming generic models (Table 1). 

Table 1 – Retrospective Benchmark analysis results*

Based on the results, TransUnion’s Advanced Analytics Consulting team was asked to develop a new custom model for the underwriting process. The team followed a Model Risk Management (MRM) compliant model development process involving segmentation, model specification, and in-time and out-of-time validation.

The results

The custom model developed by TransUnion materially outperformed the existing model and competing generic benchmarks (Table 2).

Table 2 – Custom model benchmark analysis results*

It created significant opportunities for both portfolio expansion and risk mitigation, including a:

• 7% increase in new approvals with no impact to the default rate
  
• 27% decrease in default rate at the current approval rate (Table 3)
• 6.3% lift over their existing, internally developed model

Swap set summary

Table 3 – Swap Set analysis summary results*

With the help of TransUnion’s Advanced Analytics Consulting team, the lender was able to implement a new model to help approve more applications, all without pulling full-time resources away from current projects or compromising portfolio quality.

*Note: The figures and tables shown are not derived from any particular customer but are generated for illustrative purposes.