Business Intelligence for Lending: Origination, Portfolio, and Regulatory Reporting

A mortgage lender's BI team builds origination and portfolio dashboards. The compliance team reviews the HMDA-related views and finds they don't match what actually gets submitted to the CFPB. The reasons are familiar at every lender that builds BI without HMDA-specific discipline: the dashboard counts loans by origination date; HMDA counts by action date. The dashboard uses application-level race and ethnicity; HMDA requires both applicant and co-applicant with specific code mapping. The dashboard's rate spread calculation uses APR minus the average prime offer rate differently than the CFPB methodology. Each difference is small. Together they make the dashboard useless for compliance and fair lending analysis, which is the highest-value use case for lending BI.

Lending BI done right encodes the regulatory reporting specifications correctly from day one. HMDA data structured to match the Loan/Application Register (LAR) — 110+ fields with the specific code values the CFPB expects, calculated using the CFPB methodology for rate spread, loan-to-value, and the derived fields. Fair lending regression analysis as a first-class analytical capability rather than an ad hoc project. Origination funnel dashboards using consistent definitions of pull-through, fallout, and cycle time. Portfolio dashboards with delinquency curves that roll up to vintage, ALLL, and CECL reporting. Done with this discipline, BI supports origination operations, portfolio management, and regulatory reporting from one source. Done generically, every regulatory question requires a parallel calculation.