Data Quality Remediation for Loan Purpose Codes in Mortgage Portfolio

Leveraging ML and AI to identify and remediate misclassified loans.

Project Statement

Investment property loans were misclassified as owner-occupier loans and vice-versa, among many other examples of incorrect classification. This issue led to the misrepresentation of the bank’s mortgage portfolio and, in some cases, resulted in incorrect interest rates being charged to customers. The inaccurate data also affected the regulator’s ability to make informed decisions based on the bank’s reporting.

Problem Solution

We adopted a statistical method and leveraged third-party market data to identify loans that were likely misclassified. This predictive approach was validated by an experienced testing team of seasoned bankers, ensuring its accuracy. The method was scalable over a population of well over 1 million home loan accounts. To prevent future issues, a front-book monitoring control was embedded into the process to detect misclassification as new loan accounts were being opened. This proactive solution ensured that all new loans were classified accurately from the outset.

Value Delivered

The solution provided significant benefits:

- Improved Data Quality: The machine learning and statistical approach, enhanced by third-party data, successfully identified misclassified loans, allowing for quick correction of errors.
- Regulatory Compliance: By addressing the misclassification issues and working closely with the regulator, the bank strengthened its relationship with regulators and ensured future data submissions were accurate and compliant.
- Operational and Financial Benefits: The solution resulted in lower capital costs, better growth potential, and increased confidence in the bank’s data accuracy and reporting practices.
- Enhanced Relationship with Regulators: The transparent and proactive approach improved trust with the regulator, positioning the bank as a responsible and compliant institution.

Liked it? Share this case study to inspire other designers