METHODOLOGY · hi-1.0.0

How the verdicts are computed

Every brand gets one categorical verdict: Strong disclosed record · Proven & steady · Weaker disclosed record · Adverse public-record signals · Too new to judge / Too little disclosed history · Filing too old to rate · Not enough disclosure. It is built from four independent sources — the FDD's Item 20 outlet record, the SBA loan outcomes of the brand's franchisees, the auditor's opinion on the franchisor (Item 21), and verified public records — and every brand page shows those sources side by side under the verdict, with one sentence saying which one decided it. The rules apply in this order; the first that matches sets the verdict:

1. Verified distress. Verified bankruptcy or closure events → Adverse public-record signals.
2. Hard warnings, never gated by size or age. Going-concern language in the latest FDD's audit opinion, or franchisee SBA loans that charged off at least 1.5× as often as franchise loans approved in the same years (on 30+ seasoned resolved loans) → Weaker disclosed record. Thin evidence of trouble is still trouble.
3. No usable filing → Not enough disclosure.
4. Freshness. If the newest FDD we hold is more than 24 months old (measured from the end of its filing year) → Filing too old to rate. Stale filings are also left out of every ranking.
5. Evidence gate. A positive verdict needs at least 50 franchised units, 3 disclosed years, and 25 franchised units at the start of that window — a system that grew from 3 units to 42 clears a units-today test on the strength of the growth itself, but its clean record describes how few units were exposed. Failures take years to surface (median SBA time-to-default is about 61 months). Below the gate: Too new to judge when there is evidence the system is young, otherwise Too little disclosed history.
6. Operating record. An operating index from Item 20 (the growth, exit-mix, turnover and promise-keeping factors below — not transparency): 75+ → Strong disclosed record, 55+ → Proven & steady, below → Weaker disclosed record. A Steady record that is both shrinking and exiting units far faster than comparable systems is Weaker disclosed record. Caps to Steady: a bankruptcy history in Item 4; franchisee loans charging off clearly more often than same-year loans (≥ 1.2×, statistically clear); an exit rate at least 1.5× the sub-sector median. Lift to Steady: an operating index of 40+ whose franchisees' loans charged off at 0.6× or less of the same-year rate, statistically clear.

The reason shown with each verdict is generated from the same numbers that set it, so it can't claim a cap that didn't apply. Loan records “below expected” (0.8× or less, clear) are shown as a strength but only lift a verdict under the rule above. Franchise loans used as the comparison: approved FY2020 or earlier, 16.1% of which charged off overall.

The factors under the verdict

The headline exit rate

The exit rate shown on brand pages and in rankings is (terminations + non-renewals + ceased operations — other reasons) ÷ franchised outlets at the start of the fiscal year, from the latest disclosed Item 20 year. Transfers (an outlet sold to a new franchisee) and reacquisitions (an outlet bought back by the franchisor) are not counted as exits. A brand is compared with the median system in its sub-sector (or its sector when the sub-sector has fewer than 5 systems with current filings and 10+ franchised outlets) — a pooled rate lets one big system speak for everyone. Sector pages also show a pooled rate in which no system carries more than 20% of the weight. A rate of 50% or more is used only when the same filing's unit counts corroborate it or we have checked the PDF. A growth percentage is shown only from a starting base of 25+ units; smaller systems show the counts. It is not a failure or closure rate: non-renewals can be voluntary, and the FDD does not state why any outlet exited. Unusual one-year movements (conversions, non-traditional sites, re-filings) can inflate it, and we flag the ones we have checked.

Whether a brand publishes an Item 19 earnings claim is shown as a fact on every page but does not affect the verdict or any warning list. Royalties above 20% are revenue-share arrangements, not a percentage-of-sales royalty, and are shown as disclosed but never ranked, averaged or modeled.

Citability rules

• ▸ Every score links to its source filing and the specific Item it draws from.
• ▸ When a brand doesn't disclose (e.g. no Item 19), we say “Not Disclosed” — we never estimate or impute.
• ▸ Figures are machine-extracted from the filings. We correct errors against the source PDF when we find them or they are reported, but no figure has been individually checked by a person, which is why every figure links to its source and every page carries a way to report an error.
• ▸ The formula is stable and versioned. When it changes, the version increments and history is retained.

Sources & records

Every dataset feeding a number on this site, with its citation, refresh cadence, and known limitation:

SBA loans are matched to brands by the franchise code SBA records on each loan. Each code is assigned to one brand through a code-to-brand map, built by matching the SBA franchise name for the code to our brand name (exact, or whole-word prefix) and then reviewed by hand; every loan carrying a mapped code goes to that brand. A code not in the map is left unmatched, so a brand's loan count can be incomplete. When two brand records are merged as one system, their loans are pooled.

Loan outcomes are compared with loans of the same years. When a loan was made matters more than whose franchise it financed: across every SBA franchise loan, roughly a third of those approved in 2006–08 charged off, against 6–8% of those approved in 2010–14, and recent loans look safe mainly because they haven't had time to fail. So the headline is a loss ratio: a brand's charge-offs among its resolved loans approved at least 6 fiscal years before the data cut (FY2020 or earlier), divided by the charge-offs expected if each of those loans had failed at the rate of all SBA franchise-flagged loans approved in the same year. 1.0× is exactly as expected. Both sides count only resolved loans (charged off or paid in full). One sample floor applies everywhere — brand pages, rankings, state pages, the risk model: 30 seasoned resolved loans; below it no ratio is shown. A difference is called clear only when a one-sided Poisson test puts it at p < 0.05. The all-time raw charge-off rate is still shown as evidence, never as the score. Median time to default is shown only with 10+ defaults. SBA codes that are not franchises a buyer can purchase — fuel supply agreements with refiners, insurance agency appointments, vehicle dealerships, direct selling — are excluded from rankings when we hold no FDD for them.

Coverage

1,261 brands tracked; 850 with an extracted FDD; 372 rated on a current filing with enough Item 20 history. The 850 brands with an FDD account for 1,043 extracted filings; the rest have a page because of their SBA loan record. Nearly all filings come from Minnesota's registry, which skews the sample toward brands that register there. We have no complete count of active US franchise systems, so we do not publish a coverage percentage. Cross-corpus findings live on the research page with per-study sample sizes attached.

Distress Index

A separate composite (di-1.0.0) of individually verified bankruptcy and mass-closure events, each weighted by severity and by recency (a signal loses about half its weight every three weeks). Court dockets are listed on brand and distress pages as public record but do not move the index: many are a franchisor suing a former franchisee, and a lawsuit's existence says nothing about who is right. SBA charge-off rates are shown separately and are not part of this index. An event is published only after we verify the brand's own corporate entity is involved. No bankruptcy or closure event is published today, so the index reads 0 for every brand. That means nothing has been verified, not that nothing is wrong.

FDD Risk Score

A fractional logit scorecard (unweighted), FDD + SBA loan-record features; label = vintage-adjusted charge-off rate over seasoned loans (v3), trained on the 313 systems with at least 5 seasoned resolved SBA 7(a)/504 loans. The label is each brand's vintage-adjusted charge-off rate (its loss ratio × the pooled rate of all seasoned franchise loans), so the model learns brands, not loan eras; the inputs are figures the FDD itself discloses plus three drawn from the federal loan record behind its franchisees. The trained scorecard then scores every system with an extracted filing (622 when the model was last trained) — including brands whose franchisees never borrow through SBA, which is the point: the FDD is public years before loan outcomes are.

Accuracy is reported cross-validated, never in-sample. Under 5-fold cross-validation, measured against the 152 brands whose own record is itself reliable (≥30 seasoned resolved loans), Spearman rank correlation is 0.58 and mean absolute error is 5.6 percentage points. Across all 313 labels including very thin cohorts those figures are 0.35 and 11pp — lower mostly because a six-loan brand's own rate is itself uncertain by ±15pp, so it is a noisy yardstick rather than evidence the model is worse. Quintile calibration (modeled → observed): 7.9% → 8.2% · 13.7% → 14.8% · 16.2% → 15.6% · 19.7% → 20.1% · 29% → 31%.

Each brand also carries a confidence tier reflecting the evidence behind its score: 189 insufficient · 328 directional · 105 measured. Only “measured” brands — 30+ seasoned resolved loans, the same floor as everywhere else — get a point estimate; others get a band name, not a number. When a brand's own loan record clearly contradicts its modeled score, the page leads with the observed record.

The score shown on brand pages is the percentile of the modeled rate across all scored systems (bands: under 30 lower risk, 30–54 moderate, 55–79 elevated, 80+ higher), always displayed with its top drivers and, when the brand has a seasoned loan record, the observed loss ratio beside it. It is deliberately a linear scorecard, not a black box — every score decomposes into the disclosed figures that produced it. A score is context for reading the filing, not a substitute for it.

Editorial

Rankings are algorithmic and sourced. Rankings are not for sale. Every verdict is our opinion, computed from the disclosed figures above. Corrections: every page links to its underlying filing; if a number is wrong, the filing is the arbiter — email the brand, the page, and the figure to the address on the Terms page and we will re-check it, normally within two business days.