How the data is built
Methodology
Every number here is computed directly from public records, not estimated, modeled, or supplied by lenders. This is exactly where the data comes from and how each figure is calculated.
- Funded loans
- 1,036,074
- Coverage
- FY2010–2026
- Distinct lenders
- 3,304
- Current as of
- 2026-03-31
What we never do
- Estimate, model, or project any figure
- Use numbers supplied by lenders
- Guess that two lender names are the same institution
- Impute, sample, or invent missing values
- Show approval or denial rates: that data is not in the source
Source and coverage
Source: the SBA's FOIA 7(a) and 504 funded-loan datasets from data.sba.gov, published by the U.S. Small Business Administration. The current build covers 1,036,074 funded loans across approval fiscal years 2010–2026, current as of 2026-03-31, and was last generated 2026-07-18. We refresh after the SBA posts a new quarterly extract.
What each row is
One row is one funded loan, exactly as the SBA published it. We do not de-duplicate, sample, or impute missing values; a loan with no usable lender name is dropped, and a blank numeric field is read as zero rather than guessed. Because these are funded loans and not applications, there is no approval-rate or denial-rate figure anywhere on the site — that data is not in the source.
How we group lenders
Loans are grouped by the institution name as it appears in the SBA file: the BankName
field on 7(a) loans and the CDC_Name field on 504 loans. This carries two
limitations we want to be explicit about:
- Names are not normalized across mergers, rebrands, or spelling variants. If a bank reported under two different names, or was acquired and its loans appear under both the old and new name, it shows up as more than one lender. We group by the literal name string rather than guess that two names are the same institution, because guessing wrong would misstate a lender's track record. The current build contains 3,304 distinct lender names on that basis.
- A “lender” means the bank on 7(a) loans and the Certified Development Company (CDC) on 504 loans. On a 504 deal the CDC is the named institution in the SBA data, not the third-party first-mortgage lender, so 504 rankings reflect CDC activity.
The figures we report
For each lender, state, and industry we report total funded volume (the sum of SBA gross approval amounts), loan count, and average loan size (volume divided by count). Volume uses the SBA's gross approval amount as reported; for 504 loans that is the SBA/CDC debenture portion, not the full project cost.
Charge-off rate
Charge-off is reported only on 7(a) loans — the SBA's 504 dataset never carries
CHGOFF status. So the charge-off rate is the count of a lender's 7(a) loans
marked CHGOFF divided by its 7(a) loans, and a lender or category with no 7(a)
loans (for example, a 504-only CDC) shows “not reported,” never 0 percent. Because the
denominator includes recently disbursed loans whose outcome is not yet known, this understates the
eventual rate for fast-growing recent lenders and is not a lifetime default rate. Read it as a
relative track-record signal, not a prediction.
Industries
Industry groups are the NAICS sector — the first two digits of the loan's NAICS code. We collapse the SBA's split sectors so each appears once: Manufacturing (31–33), Retail Trade (44–45), and Transportation and Warehousing (48–49). Loans with no usable NAICS code are excluded from industry rollups only.
Geography and metro areas
Metro pages aggregate loans by OMB Core Based Statistical Area (July 2023 delineations), joining the SBA's recorded project county to its CBSA. A CBSA is a union of whole counties, so the join is exact, never estimated. The project county can differ from a borrower's mailing address, and the "top cities" listed on state pages use the mailing-address city, so the two views are keyed on slightly different geography. Areas below the reporting threshold get no page. Connecticut has no metro pages: Census replaced its counties with planning regions in 2022 and the SBA's records straddle the change, so no complete Connecticut metro total can be computed; Connecticut loans are reported on the state page.
Spot something off? We publish corrections.
Tell us and we'll fix it. Our editorial policy covers how we handle corrections and keep advertising separate from the data; our data sources page documents full provenance; and the processed datasets behind these figures are on our open datasets page.