Financial institutions depend on a combination of proprietary models and credit scores to assess risks.
Depending on the assessed risks, banks design terms to loans and define interest rates. Banks are incentivized to have the most data available when making decisions to provide the best possible risk-reward tradeoff for the institution and their customers.
That’s why a major American bank turned to Unsupervised to help better assess lending risk. Before Unsupervised, they had depended on a number of proprietary models and assumptions around how changes to credit can impact lending risk.
Turning Insights to Action
Unsupervised was able to set up within the bank’s heavily secured data ecosystem and begin finding insights in weeks. With so much data available, the firm had previously struggled to analyze information across silos. But Unsupervised AI is able to easily ingest and blend data across sources, breaking down data silos.
With Unsupervised, they realized their existing models and assumptions didn’t hold true in some common scenarios. By reviewing patterns from our platform, they were able to find a number of strategic business opportunities to better assess risk and build out their competitive advantage in the lending space.
Key insights were found around:
Return on Average Assets – It was previously assumed that decreases in ROAA should directly increase risk. However, through the lens of Unsupervised, they were able to realize while on the whole this assumption is true, it actually did not hold when reviewing medium risk banks. These mid-tier risk institutions showed almost no change to their risk scores with decreased ROAA month-over-month.
Tier 1 Capital Ratios – The bank had assumed that increases in capital ratio would decrease risk, which at a macro level held true; however, there were a number of large segments within that where it did not. As a whole, they realized that Tier 1 Capital Ratio had complex, unexpected interactions when used in their model that caused incorrect risk classifications in some cases.
Loan Types – The AI found a number of trends around lending risk for firms that had decreased exposure to agriculture loans. Previously, loan type was not considered when assessing risk and the bank was able to use this to improve their risk modeling going forward.
By finding these insights, the bank was able to improve their risk management and modeling to better reflect the market. Using Unsupervised, they are able to see what’s happening in their business with each incremental data refresh and continuously find new opportunities for improvements across their risk management group.