Public data only. Scores are statistical indicators, not accusations. V-Finder does not allege that anyone committed a crime; people in Caught Cases are named only because a public DOJ release names them. How scores work · Report an error or request a correction: python0o@mail.grokbot.com
239 PPP loans whose borrowers were publicly charged, pleaded or were convicted, each linked to a DOJ/U.S. Attorney press release and matched to SBA loan data by name, state and exact amount. Sorted by loan size. Includes the misses.
Warning signs, not accusations. Scores rank a loan against all 11,468,210 PPP loans by traits shared with charged cases. Most high-scoring loans are legitimate, and the people listed here are named only because a public DOJ release names them. Charged is not convicted; read the release.
Known gaps: why 151 charged loans are missed
Fake Schedule C sole proprietors: 60 of 65 missed. The $20,833 pattern matches millions of honest loans; catching these needs IRS data.
Inflated payroll at real businesses: 28 of 61 missed. Only extreme per-job amounts show in public data.
Fake or shell businesses: 21 of 41 missed. Business age is self-reported, and free state registries cover only 5 states.
Ring members: 13 of 28 missed. Shared phones, bank accounts and IP addresses are not public.
Stolen identities (11 of 17 missed), insiders/preparers (4 of 4) and EIDL-linked schemes (4 of 9): not visible in public data.
We tested ring and cluster features and a separate small-loan ranking on the tuning split only. Neither improved held-out recall beyond the error bars, so the score is unchanged. Details: method page. What data would close these gaps:Gap Closer.