The Consumer Financial Protection Bureau's Home Mortgage Disclosure Act dataset and the federal banking regulators' Community Reinvestment Act Public Evaluation archive — the free public-disclosure systems that Regulation C under the Home Mortgage Disclosure Act of 1975 and Regulation BB under the Community Reinvestment Act of 1977 require to be operated to make public the loan-level mortgage application and origination records of HMDA-reporting financial institutions and the CRA performance assessments of FDIC-insured depository institutions — produce one of the most comprehensive publicly accessible records of customer-side mortgage-origination and fair-lending operational activity in the United States. The HMDA Loan Application Register publishes the loan-level records of approximately five thousand HMDA-reporting depository institutions, non-depository mortgage lenders, credit unions, and savings associations that originate, purchase, or receive applications for mortgage loans secured by a one-to-four-family dwelling, with each record containing the application date, the loan amount, the loan purpose, the loan type, the property location at the census-tract level, the applicant demographics, the action taken, the rate spread, the lien status, and dozens of additional structured attributes that the 2015 final rule expanded to capture the Dodd-Frank Act Section 1094 data-point requirements. The CRA Public Evaluation archive publishes the performance evaluations that the Office of the Comptroller of the Currency, the Federal Reserve Board, and the Federal Deposit Insurance Corporation prepare for each FDIC-insured depository institution at the conclusion of each CRA examination cycle, with each evaluation containing the institution's CRA performance rating, the lending-test performance assessment, the investment-test performance assessment, the service-test performance assessment, the assessment-area delineation, and the narrative discussion of the institution's responsiveness to community credit needs. Together with the related Federal Financial Institutions Examination Council Call Report data, the National Mortgage Licensing System registration records, and the state-housing-finance-agency loan-program-participant disclosures, the HMDA and CRA archive is the largest publicly accessible source of customer-side product mentions in the mortgage-origination-technology, fair-lending-compliance, loan-origination-system, HMDA-reporting-technology, automated-underwriting, and disparate-impact-analysis sectors in the US mortgage industry — and almost none of it is being systematically extracted as social proof by the mortgage-technology, fair-lending-compliance, loan-origination-system, and HMDA-reporting-platform companies whose products are being mentioned.
The under-extraction is not because the archives are inaccessible. The CFPB publishes the HMDA Loan Application Register data through the public HMDA data browser and the bulk-download data feeds in real time after the annual March 1 reporting deadline, the federal banking regulators publish the CRA Public Evaluation documents in PDF format through their respective public-access systems within thirty days of the examination conclusion, and the FFIEC consolidated public-data platform provides cross-institutional and cross-regulator access for programmatic processing. The under-extraction is because the mortgage-industry social-proof workflow has not been constructed to handle the highly regulated and fair-lending-sensitive disclosure source format — the HMDA LAR records read as loan-level reporting records rather than as product endorsements, the adjacent CRA Public Evaluation documents read as supervisory examination reports rather than as customer outcomes, and the adjacent CRA strategic-plan filings read as Regulation BB disclosure documents rather than as customer success stories. This guide formalizes the four-stage extraction workflow that converts the archives into citable customer outcomes, the discrimination between the depository-institution Regulation C axis and the non-depository-institution Regulation Z axis, and the attribution-safe quoting framework that meets the CFPB's fair-lending marketing-restriction standards under the Equal Credit Opportunity Act and Regulation B, the federal banking regulators' Unfair Deceptive Abusive Acts and Practices framework under Dodd-Frank Section 1031, and the state-attorney-general consumer-protection-marketing frameworks that apply to mortgage-industry communications.
Why the HMDA and CRA archive is under-extracted as social proof
The HMDA LAR records are the most counterintuitive social-proof sources in the US mortgage-industry technology sector. The records are submitted by HMDA-reporting financial institutions as required CFPB disclosures under Regulation C, and the surface content is the loan-level reporting record — the application identifier, the application date, the loan amount in thousands, the loan type (conventional, FHA, VA, USDA Rural Housing Service or Farm Service Agency), the loan purpose (home purchase, home improvement, refinancing, cash-out refinancing, other), the property type (one-to-four-family dwelling, manufactured housing, multifamily), the occupancy type (principal residence, second residence, investment property), the action taken (loan originated, application approved but not accepted, application denied, application withdrawn by applicant, file closed for incompleteness, loan purchased, preapproval request denied, preapproval request approved but not accepted), the action-taken date, the property location (state, county, census tract), the applicant ethnicity, race, sex, and age, the co-applicant demographics where applicable, the income relied upon in the credit decision, the purchaser type for sold loans, the rate spread for higher-priced loans, the HOEPA status, the lien status, the credit score (where the 2015 final rule data points apply), the automated-underwriting-system result, the debt-to-income ratio, the combined loan-to-value ratio, the loan term, the introductory rate period, the non-amortizing features, the property value, the manufactured-home secured-property type, the manufactured-home land-property interest, the total units, the multifamily affordable units, the submission-of-application channel, the initially-payable-to-institution flag, the mortgage loan originator NMLS identifier, the automated-underwriting-system name and result for each system used, the reverse-mortgage flag, the open-end-line-of-credit flag, the business-or-commercial-purpose flag, and the denial reasons for denied applications. The technology platform names — the specific loan-origination system, the specific automated-underwriting platform, the specific HMDA-LAR reporting platform, the specific fair-lending-monitoring platform, the specific disparate-impact-analysis platform, the specific pricing-engine platform, the specific point-of-sale platform, the specific document-preparation platform — are not directly named in the HMDA LAR structured record itself; the platform mentions appear in the adjacent CRA Public Evaluation narrative discussion of the institution's lending-test performance, the institution's annual Call Report Schedule RC-P loan-servicing-asset disclosure, the institution's Section 1071 small-business-lending-data submission narrative, the institution's National Mortgage Licensing System branch-and-loan-officer registration records, and the institution's annual audited financial statements that disclose the technology-platform capital-expenditure and lease commitments.
The surface read of the HMDA archive is therefore reporting-neutral — the records document the institution's compliance with the Regulation C reporting requirements without editorializing about any vendor. The under-extraction is the failure to recognize that the joint HMDA and CRA disclosure is the regulatory-anchor evidence that supports the technology-platform mentions in the adjacent CRA Public Evaluation, the Call Report Schedule RC-P, and the annual audited financial statements. The HMDA archive documents the institution's loan-origination volume, the institution's fair-lending-performance trajectory at the census-tract level, and the institution's automated-underwriting-system adoption pattern; the CRA Public Evaluation documents the technology platforms the institution has deployed to support the lending-test, investment-test, and service-test performance; the annual audited financial statements document the technology platforms the institution has capitalized as long-term assets. The extraction workflow that joins the HMDA record with the CRA Public Evaluation and the annual audited financial statements produces a social-proof asset that documents both the technology-platform deployment and the institution's regulatory-anchor status — the CRA Public Evaluation establishes the supervisory-recognition relationship, the HMDA record establishes the institution's loan-origination volume scale, the financial statements establish the institution's capitalized commitment, and the joined record establishes the citable customer outcome that the mortgage-origination-technology, fair-lending-compliance, loan-origination-system, or HMDA-reporting-platform company can use as social proof.
The CRA Public Evaluation lending-test narrative is the second source. The CRA Public Evaluation is the supervisory examination report that the federal banking regulator publishes at the conclusion of each CRA examination cycle, and the lending-test narrative discusses the institution's responsiveness to the credit needs of the assessment areas, the geographic distribution of the institution's lending across low-and-moderate-income census tracts, the borrower-profile distribution of the institution's lending across low-and-moderate-income borrowers and small businesses, the community-development-lending activity, the innovative-or-flexible-lending practices, and the institution's use of technology to expand access to credit. The lending-test narrative includes the disclosure of the technology platforms that the institution has deployed in the loan-origination workflow — the loan-origination-system platform, the automated-underwriting-platform, the fair-lending-monitoring-platform, the pricing-engine-platform, the point-of-sale-platform — and the supervisory observations about the institution's use of the platforms to expand credit access. The CRA Public Evaluation lending-test narrative is extractable as social proof of the technology platform's role in the institution's CRA-rated lending workflow; the surface-read approach misses the proof because the narrative is framed as a supervisory examination observation rather than as a customer success story.
The Call Report Schedule RC-P loan-servicing-asset disclosure is the third source. The FFIEC Call Report Schedule RC-P is the schedule that depository institutions with $1 billion or more in total assets and selected smaller institutions file quarterly with the federal banking regulators to disclose the institution's residential-mortgage-loan-servicing assets, the loans serviced for others by loan type, the loans sold with recourse, and the technology platforms the institution uses for the servicing workflow. The Call Report Schedule RC-P disclosure includes the technology platforms the institution has deployed for the servicing workflow — the servicing-platform, the loss-mitigation-platform, the escrow-administration-platform, the investor-reporting-platform, the default-management-platform — and the operational metrics for the platforms. The Call Report Schedule RC-P is extractable as social proof of the technology platform's role in the institution's mortgage-servicing workflow; the surface-read approach misses the proof because the schedule is framed as a regulatory financial disclosure rather than as a customer success story.
The three sources are complementary because they cover different stages of the institution's operational relationship with the technology-platform ecosystem. The HMDA LAR record covers the institution's loan-origination volume and demographic-distribution profile that anchors the regulatory legitimacy of the institution. The CRA Public Evaluation lending-test narrative covers the technology-platform deployment that supports the CRA-rated lending workflow and the supervisory recognition. The Call Report Schedule RC-P covers the technology-platform deployment that supports the mortgage-servicing workflow for the institution's servicing portfolio. The extraction workflow that handles all three sources produces a social-proof asset library that covers the regulatory-legitimacy axis, the CRA-lending-platform axis, and the mortgage-servicing-platform axis — and the library reads as more credible than a marketing-constructed social-proof library because the source materials are public CFPB, federal-banking-regulator, and FFIEC filings that the prospective customer can independently verify against the public HMDA data browser, the regulator CRA Public Evaluation archive, and the FFIEC consolidated public-data platform.
The four-stage extraction workflow
The extraction workflow consists of four sequential stages that convert the source archives into citable customer outcomes. The workflow is designed to maintain the legal and reputational safety of the extracted content; the staged construction prevents the premature publication of content that has not been verified for the attribution-safe quoting requirements that the CFPB fair-lending, the federal-banking-regulator UDAAP, and state-attorney-general consumer-protection-marketing rules require.
Stage 1 — Source-archive identification and corpus construction
The first stage identifies the source archives relevant to the technology-platform company and constructs a corpus of source documents for extraction. The HMDA Loan Application Register bulk-download data feed and the federal-banking-regulator CRA Public Evaluation archive are queried by the institution-type filter (FDIC-insured depository institution only, federally-chartered credit union only, non-depository mortgage lender only, state-chartered depository institution only, dually-supervised institution), the institution-size filter (small institutions with assets below $336 million, intermediate-small institutions with assets between $336 million and $1.34 billion under the 2025 thresholds, large institutions with assets above $1.34 billion, wholesale-and-limited-purpose institutions), the loan-volume filter, and the platform-mention filter (the company's product names, the company's parent organization name, the company's product-category terms); the query returns the institutions that mention the company's products in the CRA Public Evaluation lending-test narrative, the Call Report Schedule RC-P disclosures, the Section 1071 small-business-lending-data submission narrative where applicable, the National Mortgage Licensing System registration records, and the annual audited financial statements filed with the appropriate primary federal banking regulator.
The corpus construction is the foundation of the extraction workflow because the corpus determines the upper bound of the extractable customer mentions. The corpus should be constructed broadly in the first iteration — all currently HMDA-reporting depository institutions, all currently CRA-rated FDIC-insured depository institutions, all platform-mention queries the company has documented, all CRA Public Evaluation cycles and their associated lending-test narratives — to maximize the recall of the extractable mentions, and then narrowed in subsequent iterations to focus on the highest-yield sub-corpora that produce the most citable mentions.
Stage 2 — Product-mention extraction with structured-attribute capture
The second stage extracts the product mentions from the corpus and captures the structured attributes that support the attribution-safe quoting framework. The extraction is performed on the CRA Public Evaluation lending-test narrative, the Call Report Schedule RC-P loan-servicing-asset disclosure, the Section 1071 submission narrative where applicable, and the annual audited financial statements — the documents that contain the narrative descriptions of the technology platforms the institutions have deployed. The extraction identifies the product name, the institution name (including the holding-company parent affiliation where applicable), the institution RSSD ID that anchors the federal-banking-regulator supervisory record and the institution LEI that anchors the HMDA LAR record, the deployment context (the institution-type classification, the asset-size range, the loan-origination-volume range, the assessment-area composition, the CRA performance rating), the outcome (the CRA-lending benefit the platform enables, the fair-lending-compliance benefit the platform supports, the operational-efficiency benefit the platform delivers), and the source-document citation (the institution RSSD ID, the CRA Public Evaluation date, the lending-test page reference, the Call Report filing date, the Schedule RC-P line reference, the annual financial statement filing date where applicable).
The structured-attribute capture is the workflow's protection against the attribution-safe quoting requirements under the CFPB fair-lending marketing-restriction standards, the Equal Credit Opportunity Act and Regulation B, the federal banking regulators' UDAAP framework under Dodd-Frank Section 1031, and the state-attorney-general consumer-protection-marketing frameworks. The extracted mentions are stored as structured records with the institution RSSD ID and the CRA Public Evaluation date as the anchors; the citations allow the workflow to reconstruct the original context of the mention when the marketing team uses the mention in a social-proof asset. The citations also allow the workflow to detect mention-context decay — when an institution has been acquired or has consolidated, when the institution has changed its CRA performance rating, when the institution has filed an amended HMDA LAR that materially changes the loan-origination-volume profile, when the institution has experienced a material supervisory event that affects the regulatory-anchor reliability, when the institution's primary federal regulator has changed, when the institution has ceased originating mortgages — and to retire or update the mention before it becomes inaccurate.
Stage 3 — Attribution-safe quoting framework application
The third stage applies the attribution-safe quoting framework that converts the extracted structured mentions into publishable social-proof assets without violating the CFPB, federal-banking-regulator, and state-attorney-general marketing rules. The framework is built on four principles. The first principle is the public-record-only attribution principle — the social-proof asset cites the institution RSSD ID, the CRA Public Evaluation date and page reference, and the public-filing URL as the source, never representing that the institution has authorized the use of the mention in marketing, and never representing that the institution's borrowers have endorsed the product. The second principle is the platform-scope-bounded outcome-attribution principle — the social-proof asset attributes only the outcomes that fall within the platform-deployment scope the institution has disclosed in the CRA Public Evaluation lending-test or the Call Report Schedule RC-P, never extending the attribution to outcomes that are outside the disclosed deployment scope (a loan-origination-system platform disclosed in the lending-test should not be attributed for outcomes that fall under the Schedule RC-P servicing workflow). The third principle is the no-fair-lending-implication principle — the social-proof asset does not extract from or reference the institution's enforcement-action history that appears in the CFPB consent-order archive, the federal-banking-regulator enforcement-action archive, or the state-attorney-general consumer-protection-action records, because the enforcement history is a sensitive attribute that the institution has not authorized for marketing use even though the history is publicly available, and the social-proof asset does not present demographic-distribution data from the HMDA LAR in a way that could be interpreted as making fair-lending-performance claims. The fourth principle is the no-cherry-picking principle — the social-proof asset selects platform mentions that represent the institution's typical platform deployment, never selecting outlier mentions that misrepresent the institution's normal operational profile and that could trigger the federal-banking-regulator UDAAP-misleading-communication provision.
The framework application produces a social-proof asset library that the marketing team can use across the company's website, sales-enablement collateral, and prospective-client outreach. The library is organized by institution type, asset-size range, platform-deployment category, and outcome category, and the marketing team selects the assets that match the prospective client's institution-profile to construct the institution-specific social-proof presentation.
Stage 4 — Mention-context decay monitoring and library maintenance
The fourth stage monitors the source archives for mention-context decay and maintains the library currency. The monitoring is performed on a quarterly cycle that ingests the CFPB HMDA data browser, the federal-banking-regulator CRA Public Evaluation archive, and the FFIEC Call Report archive's new and amended filings, re-runs the platform-mention queries against the new filings, identifies the new mentions that should be added to the library and the existing mentions whose context has decayed (the institution has been acquired or has consolidated, the institution has changed its CRA performance rating, the institution has filed an amended HMDA LAR that materially changes the profile, the institution has experienced a material supervisory event, the institution has ceased originating mortgages), and updates the library. The decay monitoring is the workflow's protection against the use of mentions that have become inaccurate; the marketing team that uses a mention from an institution that has subsequently ceased originating mortgages is presenting the prospective client with a mention that no longer represents an active customer relationship.
The depository-institution-Regulation-C-axis discrimination from the non-depository-institution-Regulation-Z-axis attribution
The depository-institution Regulation C axis and the non-depository-institution Regulation Z axis are the two analytical axes that calibrate the social-proof attribution scope. The depository-institution Regulation C axis applies to the FDIC-insured depository institutions and federally-insured credit unions that report HMDA data under Regulation C and that are subject to the CRA Public Evaluation cycle under Regulation BB — the platform mentions for these institutions are bounded to the CRA-rated lending workflow and the Call Report-disclosed servicing workflow, with the supervisory-recognition attribution from the CRA Public Evaluation strengthening the credibility of the platform mention. The non-depository-institution Regulation Z axis applies to the independent mortgage banks, broker-dealers acting as mortgage lenders, and other non-depository mortgage originators that report HMDA data under Regulation C but that are not subject to the CRA Public Evaluation cycle and are instead supervised under the Truth in Lending Act Regulation Z framework and the CFPB UDAAP framework — the platform mentions for these institutions are bounded to the HMDA LAR loan-origination workflow and the CFPB supervisory-examination findings where the institution is a CFPB-supervised non-depository, with the supervisory-recognition attribution weaker because no CRA Public Evaluation cycle applies. The axis discrimination determines the social-proof construction strategy: depository-institution-axis mentions are constructed with the CRA Public Evaluation lending-test narrative as the primary anchor and the platform deployment bounded to the CRA-rated workflow, while non-depository-institution-axis mentions are constructed with the HMDA LAR loan-origination-volume profile and the CFPB supervisory-examination findings as the primary anchors and the platform deployment bounded to the Regulation Z workflow.
The eight-week implementation routine
The extraction workflow is implemented across an eight-week routine that builds the social-proof asset library at scale.
Week 1 — Corpus construction and platform-mention query design
The implementation team constructs the source-archive corpus by downloading the HMDA Loan Application Register bulk-download data feed and the federal-banking-regulator CRA Public Evaluation archive and designs the platform-mention queries against the CRA Public Evaluation lending-test narrative, the Call Report Schedule RC-P, the Section 1071 submission narrative where applicable, and the annual audited financial statement attachments.
Week 2 — Initial extraction and structured-attribute capture
The implementation team runs the initial extraction across the constructed corpus and captures the structured attributes for each mention. The week's output is the initial mentions database with the institution RSSD ID, the institution LEI, the CRA Public Evaluation date and rating, the institution-profile attributes, the platform-deployment scope, and the outcome-attribution data fully captured.
Week 3 — Attribution-safe quoting framework drafting
The implementation team drafts the attribution-safe quoting framework templates that the marketing team will use to convert the extracted mentions into publishable social-proof assets. The templates incorporate the four attribution principles and produce a consistent voice across the library.
Week 4 — Compliance review and legal sign-off
The implementation team submits the drafted social-proof templates to the company's compliance and legal teams for review. The review confirms that the templates satisfy the CFPB fair-lending marketing-restriction standards, the federal-banking-regulator UDAAP framework, and the state-attorney-general consumer-protection-marketing rules that apply to the company's product offering.
Week 5 — Initial library publication
The implementation team publishes the initial library of attribution-safe social-proof assets across the company's website, sales-enablement collateral, and prospective-client outreach materials.
Week 6 — Mention-context decay monitoring setup
The implementation team sets up the quarterly mention-context decay monitoring that will maintain the library currency as new and amended HMDA LAR submissions, CRA Public Evaluation reports, and Call Report filings are submitted.
Week 7 — Sales-team enablement and training
The implementation team trains the sales team on the use of the library, the selection of assets that match the prospective client's institution-profile, and the attribution-safe presentation of the assets in the sales conversation.
Week 8 — Library performance measurement and refinement
The implementation team measures the library's performance against the company's sales-pipeline metrics and refines the corpus construction, query design, and attribution-safe quoting templates based on the measured performance.
The HMDA Loan Application Register and CRA Public Evaluation archive is the highest-volume mortgage-industry source-document corpus in the US, and the ProofShow extraction workflow converts the corpus into a defensible, attribution-safe social-proof asset library that the mortgage-origination-technology, fair-lending-compliance, loan-origination-system, HMDA-reporting-platform, and disparate-impact-analysis company can use to construct credible customer-outcome narratives for the prospective-client pipeline. For broader mortgage and consumer-financial-services social-proof workflow guidance, see the ProofShow customer CFPB consumer complaint narrative and consent order remediation disclosure product mentions extraction workflow and the ProofShow customer FINRA BrokerCheck and Form U4/U5 broker disclosure product mentions extraction workflow.