38%
of SBA loans have no published outcome
35,640 of 93,493 loans. Every franchise default rate published anywhere is computed over the rest — ours included, at 9.76% across the 45,407 that finished.
We show our denominators.
Two public records — the disclosure documents franchisors file with states, and the SBA’s loan performance file — read as literally as we can manage. No franchisor pays us, and none can pay to appear, rank higher, or have a figure removed.
What the data actually says
783 → 452
make an earnings claim; far fewer say what an outlet earns
Of 972 brands whose filing we have read, 783 make a financial performance representation — and 331 of those disclose something other than an average revenue: the top quartile alone, a gross margin without the revenue underneath, a range with no centre, or system-wide totals rather than per-outlet figures. Each is a lawful Item 19 and none of them tells you what a unit takes.
3
state an average we will not compare
These give an average but describe what it averages ambiguously — most often per territory, where a territory may hold several outlets. Dividing an investment into a figure like that produces a return wrong by however many outlets that is, which is how a brand page here once showed a 250% annual return. They are marked and kept out of comparisons rather than averaged in, leaving 449 a model can be built on.
93,493
loans across every franchise the SBA names
5,617 distinct systems, 963 matched to a brand here. Matching is exact on a normalised name — a loose match would put one company’s failures on another’s page.
1,186
filings read, covering 972 brands
MN and WI registrations, 2015–2026. A machine reads each one and is told to copy, never to compute.
Where we refuse to guess
20
finished loans before we print a rate
Two failures in four is 50% — and also four people. McDonald’s has six finished loans and none charged off: a perfect record and no evidence. Its page says so.
8
peers before a brand gets an industry position
Below that we name the brands and print their rates without ranking them. One filing either way would move a brand a whole quarter.
0
composite scores
A single rating would blend loan outcomes with filing terms, and only 972 of 1,439 brands have both. It would be mostly missing and still look confident.
≠
Empty is not zero
Blank means nobody read it. Zero means the filing states none. Conflating them lets an unchecked brand look cheaper than one that disclosed honestly.
What this is not
467
of 1,439 brands have no filing read
The catalogue is shaped by what the loan record names and what two states publish — not by what is worth buying. Every page says which of the two it can speak from.
4
have been checked by a person
The filings are read by a machine. Every figure links to the document it came from, so you can check any of it yourself.
—
Not advice, and not a forecast
A good loan record is a fact about other people’s outcomes, not a prediction of yours.
↻
A filing describes the year it was filed
Franchisors change fees. Each page shows the year its figures came from; read the current FDD before signing anything.