Red Flag RadarForensic public-record signals

US small-business credit outcomes — SBA 7(a)/504

A free reference of historical loan-level credit outcomes on US small businesses, from public SBA 7(a) and 504 FOIA data (approvals since FY1991), with point-in-time vintage capture from 2026 onward (the FOIA feed carries a current outcome per loan; we snapshot each vintage as it publishes). The failed loans stay in the set — survivorship-bias-free, which is exactly what makes a credit-outcome benchmark usable. Descriptive only; a charge-off is a late, loan-level outcome, not a company default.

Source: U.S. Small Business Administration 7(a) & 504 FOIA data, vintage as of 2026-03-31 · data.sba.gov · 2,168,733 loans.

Gross charge-off rate by program

programloansresolved (observability)charge-off rate
504227,202125,503 (55%)9.7%
7(a)1,941,5311,386,694 (71%)15.8%

Charge-off rate = charge-offs / resolved, where resolved = paid-in-full + charged-off (loans still exempt / undisbursed are excluded from the denominator). Observability = resolved / all loans — the share of the cohort that has actually reached an outcome.

7(a) charge-off rate by approval cohort

The by-cohort curve traces the credit cycle: pre-2008 vintages charged off far above the benign post-2011 cohorts. Recent cohorts are too young to read — most of their loans have not yet reached an outcome, so a rate would be doubly selected (early charge-offs and early payoffs both). Those cohorts show their observability but no rate.

approval FYloansresolved (observability)charge-off rate (of resolved)
199118,43316,185 (88%)12.8%
199223,64620,952 (89%)9.1%
199326,28223,184 (88%)8.2%
199436,03431,879 (88%)10.2%
199555,53948,344 (87%)12.7%
199645,83639,497 (86%)13.1%
199745,25038,598 (85%)12.5%
199842,17435,806 (85%)13.5%
199943,50836,911 (85%)14.1%
200043,71837,510 (86%)14.8%
200142,91737,192 (87%)14.6%
200251,61645,017 (87%)13.5%
200367,18358,883 (88%)15.5%
200481,04570,940 (88%)18.6%
200595,78483,919 (88%)24.7%
200697,14285,915 (88%)32.4%
200799,42287,555 (88%)37.1%
200869,27360,811 (88%)30.7%
200941,08736,030 (88%)15.0%
201046,59639,133 (84%)9.4%
201153,21944,365 (83%)7.1%
201243,94037,507 (85%)6.5%
201345,97138,517 (84%)6.3%
201451,80543,636 (84%)6.7%
201563,17951,811 (82%)7.3%
201663,85650,494 (79%)8.0%
201762,20946,540 (75%)— too young to read
201860,18841,458 (69%)— too young to read
201951,76830,608 (59%)— too young to read
202042,10422,120 (53%)— too young to read
202151,63719,013 (37%)— too young to read
202247,42311,676 (25%)— too young to read
202357,2208,443 (15%)— too young to read
202470,1534,842 (7%)— too young to read
202577,9561,387 (2%)— too young to read
202626,41816 (0%)— too young to read

Why this is here

A free, transparent reference built directly from public SBA data, with explicit observability and maturity controls — a benchmark to score any distress signal against. (Commercial SBA-loan datasets exist; the differentiator here is the honest denominator, the censoring of immature cohorts, and the cross-register entity links.) The full loan-level dataset — with dated vintage-capture labels and those entity links — is licensed direct; contact us to evaluate it.

Aggregated and non-nominative: individual borrowers and lenders are not named on this page. Descriptive counts only — no conclusions.