Methodology
If a number cannot be computed honestly, Fat Boy Data does not publish it. Every definition, limit and rounding rule is here so you can see exactly what the numbers mean.
Where the numbers come from
Every number starts withHome Mortgage Disclosure Act (HMDA) loan-level filingspublished by the Consumer Financial Protection Bureau, covering 2023 to 2025. HMDA-covered lenders file one row per application: loan amount, purpose, lien position, owner occupancy, action taken and census tract. Branch locations come from the FDIC institution directory for banks and the NCUA credit union branch file. Analysis is ours. Seedata sources.
Definitions
| Decided | An application that ended in approval, origination or denial. Withdrawn and incomplete applications are excluded because they do not show whether the lender would have said yes. |
|---|---|
| Approval rate | Applications approved, including originations and approvals not accepted, divided by applications decided. |
| Borrower take-up | Loans originated divided by applications approved. This shows whether borrowers actually took the offer. A lender can approve nearly everyone and still fund very few loans, which is why this column matters. |
| Typical loan | The median loan amount. For one lender, this is the median of the loans it originated.For a market it is the median across lenders, weighted by how many loans each one made, not a loan-level median, because that is not derivable from what is published to us in aggregate. |
| Active lenders | Institutions that decided at least one application of that type in that place. |
| Rankable lenders | Lenders with at least 20 decided applications qualify for rankings. |
Loan amounts are $10,000 brackets
Federal law requires HMDA loan amounts to be published as the midpoint of a $10,000 range. A typical loan shown as $85,000 means the bracket around $85,000. A minimum of $5,000 means somewhere between $0 and $10,000.No amount on this site is exact to the dollar, and none should be read as a quote or offer.
How Fat Boy handles small samples
A lender that approved its only application has a 100% approval rate, but it should not outrank a lender that approved 82 of 102. So a lender needs20 decided applicationsbefore it can be cited for a “highest approval rate” claim. Below that, it still appears in the table, markedLimited sample.
Two more exclusions apply to editorial rankings only.Wholesale lenders never win a ranking, because a borrower cannot apply to a broker-only channel at all; they are shown marked Broker only. A lender whose borrower take-up is under 70% is also not cited as a top approver. Approving nearly everyone while funding few borrowers is not the same as being easy to borrow from.
Geography
HMDA is published at census-tract resolution by law, never by address. County figures sum the tracts in that county. ZIP figures assign each tract to the ZIP code holding most of its land area using the Census ZCTA relationship file. Tracts and ZIP codes do not nest cleanly, so ZIP figures are close approximations, not boundary-exact counts. Nothing on this site is ever presented for an individual property.
Connecticut replaced its counties with planning regions in 2022 and HMDA changed over mid-series, so Connecticut markets are built from 2025 and the year before it only.
What this data cannot tell you
An approval rate is not a lender policy.Credit decisions turn on the borrower: credit score, debt-to-income, loan-to-value and income. Underwriting by geography would be redlining, which is illegal. An approval rate measures the joint outcome of who applied in an area and how they were treated. Read it as “who has said yes to people around here”, never as “who will say yes to me”.
Coverage is 616 counties, about 79.6% of the US population. A lender's figures here describe its lending inside that footprint, not its entire national book.
Data period 2023–2025. Last updated August 2026.