This tool is built and works, but a single year of data spread across five profile dimensions leaves too few files in each lender cell for the figures to be stable. Rather than publish rates built on twenty or thirty applications, we are holding it until the 2023–2024 records are processed, which will roughly triple the sample. The method is sound; the sample is not yet. — peer-adjusted rates and the lender table are published and stable.
Not a prediction · a record
People like you. Doors like these.
Everyone knows their credit score. Almost nobody knows what happened to applications shaped like theirs — at each lender, in the same year. That is in the federal record, and it has never been published this way. Pick five facts and look.
Look at the count column as much as the percentage: a rate built on 40 files says less than one built on 400. Rows with fewer than 50 are dimmed for that reason. No signup, no credit pull, nothing stored. Nothing here predicts your outcome; it reports what happened to other files with the same shape.
Source: CFPB HMDA 2025, processed by FinanceRateCalc. Universe: FHA applications reaching a credit decision (action_taken 1, 2, 3); denial = action 3. A profile cell is published only where at least 40 applications were decided, and a lender appears within a cell only where it decided at least 30 (40 to 60 at aggregated levels) — small cells are withheld rather than shown as noise. HMDA contains no credit scores, so profiles match on geography, loan amount, income, debt-to-income and loan-to-value only; two files with identical values here can still differ in ways an underwriter sees. These are historical observations about lenders, not predictions about any individual application, not advice, and not evidence of wrongdoing by any institution. Free · no ads · no lender pays us · CC BY 4.0 · DOI 10.5281/zenodo.21575106