The business was using ‘risk-based pricing’ (based on a credit risk score) prior to Zinia. This doesn’t solve for the real business KPI, which is ‘net returns on the loan’ (whilst still managing for risk of bad lending). Thereby, the business was not capturing all of the business value available.
First, Zinia automatically cleans and standardizes the data fields and also selects the data variables for the AI model. In this case, Zinia selected the following types of data variables (about 10 final model variables).
|Product features||Customer credit history||Employment details||Loan performance|
|Loan term in months||Bureau score||Employment length||Default/ Paid in full/Early settlement|
|Interest rate offered||Debt to income ratios||Annual income|
|Loan amount||Bureau enquiries|
Then, Zinia builds deep neural networks for customer behaviors such as price sensitivity to take up the loan, customer default and how much they would pay back. Finally, Zinia combines these models and simulates for the optimal price (interest rate) that maximizes the net returns on the loan. Zinia also accommodates for any constraints such as interest rates restricted within a certain range.
Zinia projects an increase of over 25% in total net returns on loans. It starts with 6.9% for the base model and reaches upto 25.6% as the models self-learn with more data fed through Zinia across a period of 24 months.
The underlying AI Models produced by Zinia also improve in predictive power as they ingest more and more data in an automated way. (Base data + 6/12/18/24 months of additional underwriting data)