SBA 7(a) and 504 Small Business Loans

The U.S. Small Business Administration publishes its full FOIA loan extracts - every 7(a) and 504 loan the agency has approved - as free public data on data.sba.gov. This actor turns that register into a charged-per-record lookup, screening and analysis tool over 2,190,504 loans back to 1991.

Built for: small-business and commercial lenders, credit and market research, franchise and industry analysis, government-contracting research, SBA-lender due diligence, and borrower screening.

What it covers

Two programs, one actor. Pick a corpus:

Program	Loans	FY range	Lenders
7(a)	1,961,455	1991-2026	banks: 5,655 distinct
504	229,049	1991-2026	CDCs: 249 distinct

Each record carries the borrower (name, street, city, state, ZIP), the lender - the originating bank (7(a): name, FDIC/NCUA number, address) or the Certified Development Company (504: name and address) - the third-party lender on 504 loans, the gross approved amount and the SBA-guaranteed amount (7(a)), the approval date and fiscal year, the NAICS code and industry description, the franchise code and name, the project county and state, the business type and age, the term, the jobs supported, and the loan status and outcome - the date paid in full, the charge-off date and the charge-off amount.

Where the money went, and how it ended. 7(a) approvals by borrower state: CA 254,619 / TX 155,470 / NY 129,267 / FL 103,265. 504 by state: CA 45,049 / FL 16,584 / TX 10,666. Loan-status mixes are a genuine risk signal:

Status	7(a)	504
Paid in full (PIF)	1,179,360	114,445
Exempt	297,494	61,266
Cancelled	242,790	32,314
Charged off	220,688	12,169
Committed	21,123	-
Not funded	-	8,516
Closed	-	339

Two things worth knowing. The extract carries no unique loan number, so lookup resolves a borrower by exact name plus location. And the "paid in full" status ships spelled two ways in the source - PIF on the 504 side and P I F on the 7(a) side - so the actor normalizes both to PIF (loanStatusNormalized) before filtering or grouping, and a facet can never silently split.

Typical questions

• "Show every SBA 7(a) loan to a restaurant (NAICS 722) in California that charged off."
• "How much 7(a) lending has Wells Fargo done since 2015?"
• "Which 504 CDCs are most active in Florida?"
• "Find SBA loans to this borrower."
• "How did 7(a) approvals trend by fiscal year, and how do they break down by state and industry?"
• "Which franchises took the most SBA money?"

Inputs

Input	What it does
corpus	7a (default) or 504
borrName / lender	borrower or originating-lender name substring, case-insensitive
borrState / projectState / projectCounty / borrZip	location of the borrower or the financed project
naicsCode / franchiseCode	industry code prefix / SBA franchise code
businessType / loanStatus	legal form / loan status (PIF, CHGOFF, CANCLD, EXEMPT, COMMIT)
approvalFY or fyMin / fyMax	SBA approval fiscal year (exact or range) - narrows which shards are read
amountMin / amountMax	gross approved amount range
approvalFrom / approvalTo	approval date window (YYYY, YYYY-MM or YYYY-MM-DD)
locationId	the extract's internal LocationID, exact
aggregate / groupBy	one summary row per group with a count - borrower state / project state / NAICS / approval FY / lender / loan status / business type
maxResults	cap records. The aggregate mode returns every group in full and is not cut by this cap.

Default run = a tiny aggregate (loans by borrower state) - fast and safe for the daily auto-test, never a surprise full dump. For records, add a borrName / lender / loanStatus / projectState filter or a locationId.

Example inputs

A borrower's loans - a bare name in lookup mode returns that borrower's SBA loans.

{ "mode": "lookup", "corpus": "7a", "borrName": "Anytime Fitness" }

Charged-off restaurant loans in California -

{ "corpus": "7a", "projectState": "CA", "naicsCode": "722", "loanStatus": "CHGOFF", "maxResults": 20 }

Large 504 projects in California -

{ "corpus": "504", "projectState": "CA", "amountMin": 2000000, "maxResults": 20 }

7(a) lending by state, FY2020 and later -

{ "corpus": "7a", "fyMin": 2020, "aggregate": true, "groupBy": "borrState", "maxResults": 50 }

Low cost

From $0.00005 per record, down to $0.00003 at Gold. Billed only for the rows you use, and each record is metered individually.

The SBA register is ~1 GB of flat CSV across six files with no query server behind it - there is no way to ask the SBA for "just charged-off restaurant loans in California". Rather than make every lookup pay for a full-register download, the register is mirrored into fiscal-year shards, and each run streams only the shards its filter needs - memory stays flat, so a query scales with the rows returned. Every refresh is integrity-checked against the register's known shape (per-segment row floors, the expected columns, the loan-status vocabulary and key-column coverage), so a degraded or schema-changed source fails loudly at refresh instead of publishing bad rows.

Example output

mode=lookup, a 504 borrower (abridged):

{
 "platform": "sba-loans",
 "source": "sba-foia-loans",
 "state": "US",
 "corpus": "504",
 "borrName": "Turning Brook",
 "borrCity": "Alpena",
 "borrState": "MI",
 "borrZip": "49707",
 "lenderName": "Michigan Certified Development Corporation",
 "cdcName": "Michigan Certified Development Corporation",
 "thirdPartyLenderName": "Mercantile Bank",
 "thirdPartyDollars": "1767115",
 "grossApproval": "1248000",
 "approvalDate": "2009-10-02",
 "approvalFY": "2010",
 "termInMonths": "240",
 "naicsCode": "623312",
 "naicsDescription": "Homes for the Elderly",
 "projectCounty": "ALPENA",
 "projectState": "MI",
 "businessType": "CORPORATION",
 "loanStatus": "PIF",
 "loanStatusNormalized": "PIF",
 "paidInFullDate": "2012-12-31",
 "jobsSupported": "14",
 "refreshed": "2026-10-13",
 "mode": "record"
}

aggregate=true, corpus=7a, groupBy=borrState - 7(a) loans by borrower state (top):

CA   254,619
TX   155,470
NY   129,267
FL   103,265
OH    87,900

Source

• SBA - 7(a) & 504 FOIA data (data.sba.gov) - the official free public dataset this actor reads.
• Loan data is the SBA's own published FOIA extract; it reflects the agency's record as of the extract's as-of date.