HMDA Review — Loan Application Data Extraction

Turn a residential loan application file into the full, cited HMDA LAR field set - including the Section 8 demographics read by vision-based checkbox detection - with reportability, action-taken, and FFIEC-style edit checks.

Banking / LendingMortgage / TPR

Full HMDA extraction from each application file - including vision-based recognition of the Section 8 demographic checkboxes that otherwise have to be read by hand - with every value cited to its Regulation C data point

per application

What it extracts

12 extraction fields

Application Snapshot
The identifying spine of the file - ULI, application date, loan type and purpose, preapproval, action taken and its date, occupancy, and business-purpose flag - so a reviewer can orient before reading the record (12 CFR 1003.4(a)).
Property Information
The subject-property data points - location components, census tract, property value, construction method, unit count, and manufactured-home attributes - that HMDA requires for the dwelling securing the loan.
Loan Terms and Pricing
The loan-terms and pricing data points - amount, term, interest rate, rate spread, HOEPA status, lien status, non-amortizing features, and the loan-cost fields - drawn from the note or Loan Estimate / Closing Disclosure.
Applicant Financial Profile
The underwriting data points relied on in the credit decision - gross annual income, DTI, combined LTV, and the applicant and co-applicant credit scores and scoring models.
Applicant and Co-Applicant Age
The HMDA age data point for the applicant and first co-applicant as of the application date, reporting the whole-number age (not the date of birth) and noting where it was derived.
Ethnicity, Race, and Sex Checkbox Capture
The Section 8 government-monitoring capture - one row per applicant with every disaggregated ethnicity, race, and sex checkbox, read by vision from the actual markings only, never inferred from a name or any other cue (Appendix B to Part 1003).
Demographic Collection Method and Completeness
The compliance layer on the demographic capture - how the information was collected, whether visual observation or surname was used and was permissible, whether aggregate or disaggregated categories were used, and whether the collection is complete.
Action Taken
The application's disposition standardized to the HMDA action-taken value (originated, approved-not-accepted, denied, withdrawn, file closed, purchased, or a preapproval outcome) under 12 CFR 1003.4(a)(8).
Reportability Determination
Whether the application is HMDA-reportable, applying the covered-loan and dwelling definitions (12 CFR 1003.2) and the excluded-transaction tests (1003.3(c)), or flagging it for manual review.
HMDA Edit-Check Findings
Cross-field integrity checks mirroring the FFIEC syntactical, validity, and quality edit families - one row per check with its type, the data points involved, a pass/fail/review result, and the finding.
Consolidated HMDA LAR Record
The consolidated Loan/Application Register field set for the application, carrying the upstream values through into the data points enumerated in 12 CFR 1003.4(a), ready to load into the LAR.
Exceptions and Reviewer Notes
The exception log - one row per failed or unresolved edit check or demographic-integrity concern, with the cited rule, a severity, the finding, and a concrete corrective action for the reviewer.

Where it fits

HMDA data collection and reporting

Upstream

Loan origination / servicing - closed and in-process residential loan application files (URLA / Form 1003 and supporting documents) flowing from the LOS or a document repository

This step

HMDA data extraction and LAR field preparation

Downstream

  • HMDA scrubbing / data-integrity review
  • LAR assembly and FFIEC HMDA Platform submission
  • Fair-lending and CRA analysis

What it needs

Documents

  • Residential loan application (URLA / Fannie Mae Form 1003 or equivalent)
  • Demographic information addendum (Section 8)
  • Transmittal summary / underwriting decision
  • Loan Estimate / Closing Disclosure
  • Action or denial notice

Systems

  • Loan origination system (LOS)
  • Document repository
  • HMDA / LAR software

Prerequisites

  • The institution is a HMDA-covered filer for the collection year
  • Access to the complete application file including the Section 8 demographic page
  • Your LAR field mapping / filing layout for the collection year

What it produces

A per-application HMDA record - the consolidated LAR field set with the demographic capture, a reportability and action-taken determination, edit-check findings, and an exception log, each value cited to its Regulation C data point

Delivered to

  • HMDA / LAR software
  • HMDA scrubbing queue
  • Fair-lending analysis dataset

Review model

A HMDA analyst reviews the flagged edit-check exceptions, the demographic capture, and the cited data points before the record is added to the LAR and filed.

Who uses it

HMDA AnalystMortgage Compliance AnalystLoan Quality Control ReviewerFair Lending AnalystCompliance Officer

Volume fit

Works best for

lenders and mortgage-servicing operations preparing HMDA data across hundreds or thousands of applications per reporting cycle

Too small for

an institution below the HMDA loan-volume reporting thresholds, or a one-off single-file review

Grounded in

  • Home Mortgage Disclosure Act / Regulation C, 12 CFR part 1003verified as of 2026-07-22
  • 12 CFR 1003.4(a) (Compilation of reportable data - the LAR data points)verified as of 2026-07-22
  • 12 CFR 1003.4(a)(10)(i) and Appendix B to Part 1003 (Ethnicity, race, and sex - government monitoring information)verified as of 2026-07-22
  • 12 CFR 1003.2 and 1003.3(c) (Covered loan / dwelling definitions and excluded transactions)verified as of 2026-07-22
  • FFIEC HMDA Filing Instructions Guide (FIG), 2025 edition (edit-check specifications)verified as of 2026-07-22

Changelog

  • July 2026

based on a production deployment at a residential-mortgage operations provider

Related agents

Browse all agents →

See HMDA Review — Loan Application Data Extraction on your documents

We'll run it against a file of yours and walk through every cited field.