Trust is part of the product

Useful data needs visible boundaries.

MortgagesVine is being built to make public mortgage evidence easier to use without making the data say more than it can support.

Publication commitments

These rules apply before page quality, search traffic, or revenue considerations.

01

Keep historical evidence historical

HMDA fields describe reported applications and outcomes from a past activity year. We do not turn them into approval predictions or live-rate claims.

02

Compare like with like

Meaningful comparisons align period, geography, loan purpose, loan type, property context, and reporting coverage before showing a difference.

03

Protect people in the data

We publish aggregates, suppress unstable small cells, and never attempt to identify applicants or create individual application pages.

04

Separate identity from similarity

Lenders, brands, parents, brokers, servicers, and complaint-company names are distinct until a human-reviewed relationship establishes otherwise.

Evidence pipeline

From source release to public page

  1. 1Preserve

    Store source fields, release identity, and activity year.

  2. 2Normalize

    Resolve lender-year and geography identities without erasing history.

  3. 3Aggregate

    Build documented cohorts and apply minimum-count gates.

  4. 4Explain

    Publish counts, denominators, definitions, and limitations together.

Denial-reason method

Reported reasons stay tied to their denominator

County denial-reason analysis uses public HMDA loan-level CSV rows for action codes 3 and 7. A reason is counted at most once per application, even if repeated across the four published reason fields. Each share is divided by all denied decisions; totals can exceed 100% because one decision may include multiple reasons. Counts below 20 are suppressed, and missing reasons are never inferred.