Skip to main content

Benchmark Methodology

Every market comparison TermVerify shows is computed from public federal data using the rules on this page. Last updated July 2026.

Data source

Our benchmarks are derived from the 2025 Modified Loan/Application Register published under the Home Mortgage Disclosure Act (HMDA) by the Consumer Financial Protection Bureau and FFIEC on March 31, 2026 — loan-level records of U.S. mortgage originations that lenders are required by federal law to report. Our analysis population contains 2,706,510 purchase loans (3.78 million including refinances) across all 50 states, DC, and Puerto Rico. The raw data is public and free: ffiec.cfpb.gov. Anyone can reproduce our numbers from it.

Filters we apply

  • Originated loans only (not applications, denials, or purchased loans)
  • First-lien loans on site-built, 1–4 unit homes
  • Principal residences, consumer-purpose (no investment or business loans)
  • No reverse mortgages
  • Loans with reported total loan costs greater than $0
  • A benchmark cell is shown only when it contains at least 30 loans; below that we fall back to the state, then national, cohort — and we always print which cohort and how many loans a comparison uses

Purchases and refinances are never mixed

Refinances typically close with lower costs than purchases. Blending them makes refi-heavy lenders look cheaper than they are for a homebuyer: one major national lender's California median is $5,425 when purchases and refinances are mixed, but $8,205 for purchase loans alone. Every TermVerify benchmark is computed separately for purchases and refinances, and your Loan Estimate is only ever compared to loans of the same purpose.

What the comparison covers

HMDA's total_loan_costs field corresponds to sections A (origination charges), B (services you cannot shop for), and C (services you can shop for) of your Loan Estimate — the "Total Loan Costs" line. It excludes prepaids, escrow deposits, transfer taxes, and recording fees. TermVerify compares only your sections A+B+C total to it. When a report shows a "Market Check," that is the comparison being made — nothing else.

Why we don't publish per-fee benchmarks

No public dataset discloses individual fees (appraisal, title insurance, recording, credit report). HMDA reports five cost fields, none of them fee-level; the fee-level dataset that does exist (the Uniform Closing Dataset collected by Fannie Mae and Freddie Mac) is not public. Sites that display per-fee "lender averages" sourced from "verified closing disclosures" are presenting estimates or relabeled aggregate data as if it were fee-level records. We show per-fee dollar figures only where a verifiable source exists — government fee schedules, state-filed rates, or statutory tax formulas — and label everything else as a typical industry range, not data.

Vintage and updates

Benchmarks currently reflect 2025 originations (published 2026). HMDA data is annual; we refresh when each new release is published. Comparisons in existing reports are snapshots — they state their vintage and are not retroactively recalculated. Costs shown are medians and quartiles of what similar borrowers actually paid, not quotes, and your pricing depends on your credit profile, property, rate lock, and loan terms.

Demographic data firewall

TermVerify does not use, display, or make any decision based on borrower demographic information (race, ethnicity, sex, or age) contained in HMDA data. Our build pipeline structurally excludes those fields.

Corrections

If you believe a figure is wrong, contact us via the dashboard and we will investigate against the source data. This page is educational information, not financial, legal, or lending advice; TermVerify is not a lender or mortgage broker.