How we score, where the data comes from, what we don't know.

• Franchise Disclosure Documents (FDDs)

  Brand-specific filings franchisors must give prospective franchisees under FTC rule 16 CFR 436. Items 1, 3, 5, 6, 7, 11, 12, 17, 19, and 20 are the financial and contract sections we extract. Source PDFs come from state regulator portals (CA, MN, NY, ND, RI, VA, WA, WI, plus NASAA/state-administrator records).

  Coverage · 3,070 FDD-extracted brands, resting on 3,024 filings · FDD years 2017–2026

• Item 19 earnings — what the average is per

  An Item 19 average is not always an average per outlet. Franchisors choose their own unit of analysis and disclose it in the table header: most count outlets, but a substantial minority count franchisees, territories, vehicles, trucks, offices or regions. Where a franchisee holds several outlets, a per-franchisee average is mechanically larger than a per-outlet one — by 1.6x to nearly 6x among the brands we have read — without the underlying business being any stronger.

  We record the franchisor's own noun rather than converting between them, because a converted figure is one the filing does not state. Where a brand's average is not per outlet, the figure is still shown — it is true and disclosed — and carries a label saying what it is averaged over. It is excluded from rankings, sorts, category averages, comparison winners and threshold scores, because an ordering is itself a claim that the numbers are commensurate, and no label beside a sort position repairs that.

  One exception, stated so it is not mistaken for an oversight: where the franchisor counts franchisees but the system holds roughly one outlet per franchisee, the average is a per-outlet average in all but name, and we treat it as comparable. Separately, where a filing never states its unit of analysis we do not guess one — an unread denominator is left unlabelled rather than assumed to be outlets, and the size of that unread population is what we are working down.

• SBA loan disclosures

  Public records released under FOIA, covering Small Business Administration loans guaranteed under the 7(a) and 504 programs. We join these to FDD brand names so you can see real charge-off rates per brand. 3,014 brands have SBA loan history.

  Coverage · 126,612 7(a) loans · 16,915 504 loans · 3,014 brands with lending records

  Charge-off rates are computed against resolved loans (charged off + paid in full), not total originations. Loans still being repaid are excluded from the denominator because their final outcome is unknown. This produces a more accurate picture of actual loss experience but means rates may differ from raw origination-based calculations.

  Sub-program note: 7(a) totals include all sub-programs (Standard 7(a), SBA Express, CAPLines, Export Express, and Community Advantage). SBA Express loans carry a lower guarantee percentage and historically show higher charge-off rates than Standard 7(a). 504 totals include both regular purchase and refinance loans. Lender-level charge-off rates reflect each lender's full portfolio mix across these sub-programs.

  Charge-off rate = charged-off loans / (charged-off + paid in full), cumulative since 2000. Loans still active or in disbursement are excluded from the denominator.

• Brand logos

  Logos are sourced from Google's favicon service using each brand's website domain. Where no favicon exists, a monogram tile renders the brand initials.

How the score is built

Under the hood, the model computes a deterministic internal risk score from 5–100 where lower is better. We invert it for display into the public Verdict Score (0–100, higher is better) shown in the scale above — the two are the same measurement on opposite axes. The steps below describe that internal risk score: it is a deterministic, multi-layer model, so the same data always produces the same score. It blends six independent risk dimensions, then calibrates against real-world loan outcomes and applies hard floors.

Layer 1 — six risk dimensions

Each dimension is scored 0-100 from the FDD data, then combined weakest-link style (a weighted average blended roughly 55/45 with the single worst dimension) so one serious red flag carries weight rather than being averaged away:

• Financial health (~22%): going-concern and audit status, franchisor net income and net worth (Item 21)
• Unit economics (~20%): revenue-to-investment ratio and cash-on-cash return (Items 7 + 19)
• Unit growth (~16%): net unit growth and franchisee turnover (Item 20)
• Scale (~15%): system size and years franchising — a larger, longer-running system is more proven
• Legal (~15%): litigation count and bankruptcy disclosure (Items 3 + 4)
• Transparency (~12%): whether Item 19 earnings are disclosed and financials are audited

Layer 2 — outcome calibration

• SBA 7(a) charge-off rate — the only outcome-validated signal in the model. Low rates on ≥10 loans lower the score; high rates raise it. Requires ≥10 loans for standard tiers, ≥5 for extreme rates.
• FDD staleness — older filings carry more uncertainty and nudge the score up. A 2017 FDD is less certain than a 2025 one.

Layer 3 — hard floors & caps

• SBA hard floor: very high charge-off rates floor the grade regardless of clean FDD language.
• Going-concern cap: an auditor's going-concern doubt caps the letter grade.
• Mega-system representativeness guard: proven 2,000+ unit systems are exempt from mild small-sample floors, so a 24-loan SBA sample doesn't drag down a McDonald's or Papa John's.

The internal risk score is clamped to 5–100 (lower = better), then inverted into the public Verdict Score and mapped to a letter grade at fixed cutoffs: A (70–100), B (46–69), C (38–45), D (28–37), F (0–27). These are the same bands shown in the scale above. The cutoffs are fixed score thresholds, not a percentile split, so the number of brands in each grade is whatever the data produces rather than a fixed share. Approximate weights are shown; exact contribution varies per brand because some signals are absent (e.g. no SBA data means the calibration step is skipped and the dimensions carry the score).

Manual overrides

0 brands have a manually reviewed score override. These are cases where the automated model produces a misleading grade due to data anomalies (e.g., name collisions in SBA matching, unusual FDD structures, or known extraction errors). Each override is documented internally with a justification. We aim to reduce overrides as data quality improves.

SBA-only brands

1,255 brands have SBA 7(a) loan data but no FDD on file. Their Verdict score is based solely on lending outcomes: charge-off rate is the primary signal, with adjustments for loan volume confidence and recent performance trends. These brands are marked “SBA data only” throughout the site.

Brands4,325

Brand pages the site actually publishes: is_published AND merged_into IS NULL. Excludes unpublished drafts and rows merged into a canonical duplicate.

Brands graded3,070

Published, unmerged brands whose Verdict letter the site actually SHOWS: rating IS NOT NULL and not sba_only (the owner decision that lending-only brands show lending facts, no grade — GRADED_RATING_SQL). The column holds a rating on sba_only rows too, which is why this used to equal published_brands and print '4,327 brands graded' beside 3,071 graded pages (R22, 2026-09-12).

Lending-data-only brands1,255

Published, unmerged brands with SBA lending history and NO FDD on file (data_quality = 'sba_only'). They browse behind a filter, carry no grade, and are the difference between '4,327 brands' and '3,071 with an FDD'.

Brands with an FDD on file3,070

Published, unmerged brands whose row NAMES a source FDD document (pdf_filename). This is a count of BRANDS, not documents. 3,072 of these are `verified` — the brand, the compiled index and the file on disk all agree.

FDD filings3,024

Distinct source FDD documents behind those brands (distinct lower(pdf_filename)). Lower than fdd_brands because 48 brands are covered by a document filed jointly with another brand. THIS is the number to quote as "filings" or "documents".

Brands with an FDD year3,070

Published, unmerged brands with fdd_year recorded. A YEAR is not a DOCUMENT: 19 brands carry a year with no file named. The unscoped version of this query (no published/unmerged filter) returns 3,188 and is what methodology used to print.

With a revenue figure1,405

Published, unmerged brands publishing an average gross-sales figure from Item 19 (avg_gross_sales IS NOT NULL). Gross sales are not profit.

With matched SBA history3,014

Published, unmerged brands with at least one matched SBA loan from a named lender (sba_total_loans IS NOT NULL AND sba_unique_lenders > 0).

SBA 7(a) records126,612

7(a) franchise loans in the FOIA extract, rolled up BY LENDING BANK (sba_banks). Per-brand charge-off is 7(a)-only; 504 is real-estate.

SBA 504 records16,915

504 franchise loans in the same FOIA extract, rolled up BY CDC (sba_cdcs).

SBA loan records143,527

Combined 7(a)+504, from the lender rollups. Each loan is counted once. THIS is the public SBA figure — it is the 143,527 on the homepage.

Owner contacts278,585

Franchisee owner contacts from FDD Item 20, summed over PUBLISHED, unmerged brands (sum(contact_count) WHERE contact_count > 0). The unscoped sum is 279,885 and equals count(*) FROM contacts; the difference is contacts sitting on rows that have no page. Personal due-diligence only — not for telemarketing or automated calls (TCPA).

Brands with contacts2,545

Published, unmerged brands with at least one Item 20 owner contact.

Categories18

Rows in the categories table.

Verified against the document3,072

Of fdd_brands, those where all three sources agree: the brand names a document, the file is on disk, and text has been extracted from it. The 2 exceptions are year disagreements (skrimp-shack-food-truck, tapioca-express).

Item 19 declined1,007

Filings we hold whose Item 19 EXPLICITLY makes no financial performance representation. Item 19 is voluntary, so these rows are correctly empty — counting them as an extraction gap inflates the backlog roughly threefold.

Item 19 with an FPR1,596

Filings we hold whose Item 19 makes a financial performance representation with real figures. Larger than revenue_brands (1,086) because an FPR is not always a stated average — some list per-outlet figures only, and deriving a mean from those would be fabricating a number the filing does not state.

Live counts as of 2026-09-30; figures without a live query are measured from the source-of-truth build and carry its date. Where a figure below is smaller than one you have seen elsewhere on the internet for this corpus, the difference is the definition, not the data — a brand is not a filing, and a loan attributed to two brands is still one loan.

Strong is 80 points or more, partial 50 to 79, thin below 50. The points are not shown on brand pages: the site keeps one public 0-100 number, the verdict score.

All financial data is free to browse. There is no subscription. The $49 unlock applies only to the franchisee phone numbers (and AI- generated validation questions) for a single brand. One-time purchase.

• Free: investment ranges, royalty rates, Item 19 revenues, unit growth, SBA charge-off rates, ratings, contract terms, litigation summaries, comparison and matrix views.
• $49 / brand: every current and former franchisee that brand's FDD lists in Item 20 (the count varies by brand), plus AI-generated questions to ask owners during validation calls.

FDDs are filed annually with state regulators. We re-extract a brand when its filing year changes. SBA 7(a) data is refreshed quarterly when SBA publishes new FOIA datasets. The data ticker on the home page reflects the current totals; each brand page stamps the FDD year it was extracted from.