Report cards › How we grade
How the report cards grade
Every grade is arithmetic over public records, by the method below: no guesses and no model. Each grade shows the method version it was computed with. Company report card · Food report card · Corrections
Company report card
Method v1.1, Oct 23, 2026. Window: The last 5 complete years and the current year so far. Weights: by indicators with data; overall at most one letter above the lowest category (1.1).
Letters
| Letter | Score |
|---|---|
| A | 85 and above |
| B | 70 to 84 |
| C | 55 to 69 |
| D | 40 to 54 |
| F | below 40 |
Every score is rounded to a whole number (halves up) before its letter is chosen: 84.5 is 85, an A.
From indicators to grades
- A cause's score is the plain average of its indicators that have data.
- A category's score is the average of its graded causes and its related indicators, shown only when at least 2 indicators in it have data, or every indicator of the category that applies to the company, when fewer than 2 apply (Not covered ones do not apply).
- The overall grade is the average of the category scores, each category counting as many times as it has indicators with data, so a category resting on one fact counts less than one backed by many. It is at most one letter above the company's lowest category letter (a company with an F in any category gets a D at best, shown as 54). It is shown only when at least 4 categories and 8 indicators have data; otherwise the card says "Not enough public data for an overall grade".
- "Your grade" (members) uses the same arithmetic over the causes in your profile only, when at least 2 indicators behind them have data. It is computed when the page loads and never stored.
- "No records found" means the company is in the source's scope and has none: that is data. "Not covered" means the source does not reach this kind of company: never scored.
- Allegations (open labor charges, pending discrimination suits, consumer complaints) are shown as allegations and never lower a grade. Only resolved outcomes count.
- Counts are taken per 1,000 or 10,000 employees or per $1 billion of revenue; without that figure an indicator shows the raw count and is not scored.
Indicators
- C-LAB-1 Serious workplace-safety violations (Economic systems, wealth and labor)Inputs: OSHA violations of type serious, willful or repeat, plus fatality/catastrophe inspections, per 1,000 employees.Rule: 100 at 0; minus 20 for each 0.5 per 1,000 (pro rata); willful or repeat counts double; floor 0.
- C-LAB-2 Wage theft found (Economic systems, wealth and labor)Inputs: Wage and Hour Division back wages and civil penalties per 1,000 employees.Rule: 100 at $0; 80 under $100; 60 under $1,000; 40 under $10,000; 20 under $100,000; else 0.
- C-LAB-2-OT Overtime back wages found (Economic systems, wealth and labor)Inputs: The overtime part of C-LAB-2: Wage and Hour Division overtime back wages per 1,000 employees.Rule: As C-LAB-2.
- C-LAB-3 Union-rights findings (Economic systems, wealth and labor)Inputs: NLRB unfair labor practice cases against the employer that ended in a settlement, an issued complaint or a Board/ALJ order. Open charges are shown and not scored.Rule: 100 at 0; minus 15 per finding per 10,000 employees; floor 0.
- C-LAB-4 CEO pay ratio (Economic systems, wealth and labor)Inputs: Proxy statement, Item 402(u).Rule: 100 at 50:1 or lower; 0 at 500:1 or higher; straight line between.
- C-LAB-5 Median employee pay (Economic systems, wealth and labor)Inputs: Proxy statement, Item 402(u) median employee total pay, compared with the HHS poverty guideline for a household of four.Rule: 100 at 3 times the guideline or more; 0 at 1 time or less; straight line between.
- C-LAB-6 Mine-safety violations (mining companies only) (Economic systems, wealth and labor)Inputs: MSHA assessed violations per 1,000 employees.Rule: Same shape as C-LAB-1.
- C-ECO-1 Effective income-tax rate (Economic systems, wealth and labor)Inputs: XBRL income-tax expense divided by pre-tax income, 3-year average.Rule: 100 at 21% (the federal statutory rate) or above; 0 at 0% or below; straight line between.
- C-ECO-2 Political money and lobbying (shown, not graded) (Political institutions and democratic restructuring; shown, not graded)Inputs: Corporate PAC giving (FEC); lobbying spend and bills lobbied (Senate LDA), joined to the CSWH bill register's causes.Rule: Not scored. Shown as amounts and lists (decision D5).
- C-ECO-3 Antitrust and consumer-protection actions (Economic systems, wealth and labor)Inputs: DOJ Antitrust Division and FTC cases resolved against the company.Rule: 100 at 0; minus 25 each; floor 0.
- C-EQ-1 Discrimination findings (Race, identity and social justice)Inputs: EEOC suits resolved by consent decree or settlement, tagged by basis (race, sex, sexual orientation/gender identity, disability, age, religion, national origin).Rule: 100 at 0; minus 20 each, scaled per 10,000 employees; floor 0.
- C-EQ-1-RACE Discrimination findings: race (Race, identity and social justice)Inputs: C-EQ-1 suits on the race basis.Rule: As C-EQ-1.
- C-EQ-1-SOGI Discrimination findings: sexual orientation and gender identity (Race, identity and social justice)Inputs: C-EQ-1 suits on the sexual orientation or gender identity basis.Rule: As C-EQ-1.
- C-EQ-1-SEX Discrimination findings: sex (Race, identity and social justice)Inputs: C-EQ-1 suits on the sex basis (pregnancy included).Rule: As C-EQ-1.
- C-EQ-2 Women in leadership (Race, identity and social justice)Inputs: Share of women on the board and among named executive officers, from the proxy statement.Rule: 100 at 50% or more; 0 at 0%; straight line between.
- C-EQ-3 Publishes its EEO-1 workforce data (Race, identity and social justice)Inputs: Yes or no, with a link to the company's own page (checked by hand each year).Rule: Yes = 100, no = 0.
- C-ENV-1 Federal environmental enforcement (Climate, energy and environmental policy)Inputs: EPA ICIS FE&C federal enforcement cases with penalties against the parent's facilities.Rule: 100 at 0; minus 10 per case; a penalty over $1 million counts double; floor 0.
- C-ENV-2 Facilities in serious non-compliance now (Climate, energy and environmental policy)Inputs: Share of the parent's ECHO facilities flagged as significant violators.Rule: 100 at 0%; 0 at 25% or more; straight line between.
- C-ENV-3 Toxic releases (Climate, energy and environmental policy)Inputs: TRI on-site releases (pounds) per $1 billion of revenue.Rule: Banded by cut-points set from the first full run and fixed for method v1.0: 100 at zero, then 80, 60, 40 and 20 up to each cut-point, 0 above the last.The cut-points are set from the first full run and then fixed for this method version; until then this indicator is shown, not scored.
- C-ENV-4 Greenhouse-gas emissions (Climate, energy and environmental policy)Inputs: GHGRP direct emissions (tonnes CO2e) per $1 billion of revenue; shown as "reporting ended" if EPA ends the program.Rule: Banded as for C-ENV-3.Cut-points 3296.063, 19879.184, 115832.083, 502571.971, set Oct 28, 2026 from the first full run (496 companies).
- C-ENV-5 Fossil-fuel business (Climate, energy and environmental policy)Inputs: SIC code in oil and gas extraction, coal mining, refining or pipelines, or a GHGRP fuel-supplier report.Rule: Involved = 0, not involved = 100.Graded based on the company's line of business.
- C-ENV-6 Pollution near overburdened communities (Climate, energy and environmental policy)Inputs: Share of the parent's TRI/GHGRP facilities in areas above the national 80th percentile for low-income or people-of-color share (ECHO facility demographics).Rule: 100 at 0%; 0 at 50% or more; straight line between.Graded based on the company's line of business.
- C-ENV-7 Violations on or near tribal lands (Race, identity and social justice)Inputs: ECHO facilities on or near tribal lands (Facility and Tribal Spatial), joined to C-ENV-1.Rule: Scored like C-ENV-1.
- C-GOV-1 Immigration-enforcement contracts (Immigration and border policy)Inputs: USAspending obligations from ICE and CBP.Rule: 0 if detention or removal services, or over $10 million a year; 50 if any other; 100 if none.Graded based on the company's line of business.
- C-GOV-2 Prison and detention contracts (Criminal justice, policing and public safety)Inputs: USAspending obligations from the Bureau of Prisons and the U.S. Marshals Service for detention.Rule: As C-GOV-1.Graded based on the company's line of business.
- C-GOV-3 Defense dependence (Foreign policy, military and international affairs)Inputs: Department of Defense obligations as a share of revenue.Rule: 100 at 0%; 0 at 25% or more; straight line between.Graded based on the company's line of business.
- C-FIN-1 Consumer-finance enforcement (Healthcare, education and social welfare)Inputs: CFPB enforcement actions. Complaint counts are shown, not scored.Rule: 100 at 0; minus 25 per action; floor 0.
- C-HLT-1 Private health-insurance business (Healthcare, education and social welfare)Inputs: SIC code 6324 (hospital and medical service plans).Rule: Involved = 0, not involved = 100.Graded based on the company's line of business.
C-LAB-5 uses the 2026 HHS poverty guideline for a household of four in the 48 contiguous states and the District of Columbia: $33,000 (U.S. Department of Health and Human Services, ASPE: 2026 Poverty Guidelines).
Causes
27 causes are graded, 1 shown without a grade and 55 not graded, each with its reason. Wherever a company lobbied on a bill CSWH's bill list tags with a cause, the card lists it under that cause (shown, not graded).
Economic systems, wealth and labor
- Democratic socialism: Not gradedNot graded: no company record measures this. Employee-ownership plans in Form 5500 data are a possible later addition and were not checked
- Wealth redistribution: Graded (C-ECO-1)
- Wealth taxes: Graded (C-ECO-1)
- Public ownership of industries: Not gradedNot graded: no company record measures this. Employee-ownership plans in Form 5500 data are a possible later addition and were not checked
- Stronger labor unions: Graded (C-LAB-3)
- Living minimum wage: Graded (C-LAB-5, C-LAB-2)
- Guaranteed employment: Not gradedNot graded: no company record measures this. Employee-ownership plans in Form 5500 data are a possible later addition and were not checked
- Shorter workweek: Graded (C-LAB-2-OT)
- Worker ownership: Not gradedNot graded: no company record measures this. Employee-ownership plans in Form 5500 data are a possible later addition and were not checked
- Anti-corporate concentration: Graded (C-ECO-3)
- Corporate responsibility: Not gradedNot graded yet: this cause was added after method v1.0; which company indicators count for it is a method v1.1 decision
Also counted in this category: C-LAB-1, C-LAB-6 (worker safety).
Healthcare, education and social welfare
- Medicare for All: Graded (C-HLT-1)
- Elimination of private health insurance: Graded (C-HLT-1)
- Tuition-free public college: Not gradedNot graded: no national open data ties companies to this. Corporate-landlord data exists only city by city
- Student debt cancellation: Graded (C-FIN-1)
- Public or social housing: Not gradedNot graded: no national open data ties companies to this. Corporate-landlord data exists only city by city
- Universal rent control: Not gradedNot graded: no national open data ties companies to this. Corporate-landlord data exists only city by city
- Universal childcare: Not gradedNot graded: no national open data ties companies to this. Corporate-landlord data exists only city by city
- Guaranteed basic income: Not gradedNot graded: no national open data ties companies to this. Corporate-landlord data exists only city by city
- Free or subsidized public transit: Not gradedNot graded: no national open data ties companies to this. Corporate-landlord data exists only city by city
- Tenant power: Not gradedNot graded: no national open data ties companies to this. Corporate-landlord data exists only city by city
- Broad debt cancellation: Graded (C-FIN-1)
- Food safety: Not gradedNot graded on the company card: food brands are graded on the food report card
Race, identity and social justice
- Structural racism: Graded (C-EQ-1-RACE)
- Reparations for slavery: Not gradedNot graded: employer health-plan coverage is not public in any dataset
- Equity-based policies: Graded (C-EQ-3)
- Transgender rights: Graded (C-EQ-1-SOGI)
- Public funding of gender-affirming care: Not gradedNot graded: employer health-plan coverage is not public in any dataset
- Reproductive autonomy: Not gradedNot graded: employer health-plan coverage is not public in any dataset
- Indigenous sovereignty: Graded (C-ENV-7)
- LGBTQ+ antidiscrimination: Graded (C-EQ-1-SOGI)
- Socialist feminism: Graded (C-EQ-2, C-EQ-1-SEX)
- Gender-affirming care for minors: Not gradedNot graded: employer health-plan coverage is not public in any dataset
Immigration and border policy
- Abolish ICE: Graded (C-GOV-1)
- Broad immigration amnesty: Not gradedNot graded: no public record links companies to this cause
- End immigrant detention: Graded (C-GOV-1)
- Decriminalize migration: Not gradedNot graded: no public record links companies to this cause
- End mass deportations: Graded (C-GOV-1)
- Immigrant access to public services: Not gradedNot graded: no public record links companies to this cause
- Eliminate visa caps and quotas: Not gradedNot graded: no public record links companies to this cause
- Open borders: Not gradedNot graded: no public record links companies to this cause
Criminal justice, policing and public safety
- Police abolition: Not gradedNot graded: no public record links companies to this cause
- Defund or reallocate police budgets: Not gradedNot graded: no public record links companies to this cause
- Prison abolition: Graded (C-GOV-2)
- End cash bail: Not gradedNot graded: no public record links companies to this cause (bail-bond firms are rarely SEC filers)
- Drug decriminalization: Not gradedNot graded: no public record links companies to this cause
- End mandatory minimums: Not gradedNot graded: no public record links companies to this cause
- End qualified immunity: Not gradedNot graded: no public record links companies to this cause
- Prisoner voting rights: Not gradedNot graded: no public record links companies to this cause
- Civilian oversight of police: Not gradedNot graded: no public record links companies to this cause
- Gun violence: Not gradedNot graded yet: this cause was added after method v1.0, and no company record has been chosen for it
Climate, energy and environmental policy
- Green New Deal: Graded (C-ENV-4, C-ENV-3)
- Fossil fuel phaseout: Graded (C-ENV-5, C-ENV-4)
- Public ownership of energy: Not gradedNot graded: no public record links companies to this cause
- Environmental justice: Graded (C-ENV-6, C-ENV-1, C-ENV-2)
- Climate reparations: Not gradedNot graded: no public record links companies to this cause
- Restrict new fossil fuel projects: Graded (C-ENV-5, C-ENV-4)
- Public transit electrification: Not gradedNot graded: no public record links companies to this cause
- Ecosocialism: Not gradedNot graded: no public record links companies to this cause
Foreign policy, military and international affairs
- Anti-imperialism: Not gradedNot graded: no public record links companies to this cause
- Major defense spending cuts: Graded (C-GOV-3)
- Close overseas military bases: Not gradedNot graded: no public record links companies to this cause
- Palestinian self-determination: Not gradedNot graded: the UN OHCHR database of businesses involved in settlements is not used until its reuse terms are checked (decision D11)
- End military aid to Israel: Not gradedNot graded: the UN OHCHR database of businesses involved in settlements is not used until its reuse terms are checked (decision D11)
- Opposition to economic sanctions: Not gradedNot graded: no public record links companies to this cause
- Reduce U.S. military alliances: Not gradedNot graded: no public record links companies to this cause
- Anti-interventionism: Graded (C-GOV-3)
Political institutions and democratic restructuring
- Abolish Electoral College: Not gradedNot graded: no public record links companies to this cause
- Multiparty democracy: Not gradedNot graded: no public record links companies to this cause
- Abolish U.S. Senate: Not gradedNot graded: no public record links companies to this cause
- Supreme Court restructuring: Not gradedNot graded: no public record links companies to this cause
- Publicly funded elections: Shown, not graded (C-ECO-2)
- Voting rights for noncitizens: Not gradedNot graded: no public record links companies to this cause
- New socialist constitution: Not gradedNot graded: no public record links companies to this cause
- Parliamentary-style government: Not gradedNot graded: no public record links companies to this cause
- Government overreach: Not gradedNot graded yet: this cause was added after method v1.0, and no company record has been chosen for it
- Elections 2026: Not gradedNot graded yet: this cause was added after method v1.0, and no company record has been chosen for it
Radical structural and revolutionary positions
- Abolition of capitalism: Not gradedNot graded: no public record links companies to this cause
- Revolutionary transformation: Not gradedNot graded: no public record links companies to this cause
- Abolition of private profit in essential services: Graded (C-HLT-1)
- Nationalization of major corporations: Not gradedNot graded: no public record links companies to this cause
- Worker-controlled government: Not gradedNot graded: no public record links companies to this cause
- Dismantling existing security institutions: Not gradedNot graded: no public record links companies to this cause
Version history
- v1.0 (Oct 23, 2026): First published method: the indicators, thresholds and equal weights proposed on 2026-10-23 and approved as recommended (decisions D1-D14). To be revisited after the first full run with real numbers, as version 1.1 with a public change note (D12).
- v1.1 (Oct 23, 2026): After the first real run (19 of 25 companies graded A): the overall grade weights each category by how many indicators have data, so a category resting on one yes/no fact no longer counts as much as one backed by many records; and the overall grade is at most one letter above the company's lowest category grade (Scott 2026-10-23, 'Weight + cap').
Sources
- Company registry: SEC EDGAR, Wikidata and GLEIF (U.S. Securities and Exchange Commission; Wikidata; GLEIF)Source last updated Oct 27, 2026 · refreshed every 7 daysThe largest U.S. SEC filers by revenue (XBRL frames), with Wikidata (CC0) for the QID, aliases, subsidiaries and employee counts, GLEIF (CC0) for LEI parents and children, and each 10-K's Exhibit 21 subsidiaries.
- NLRB unfair labor practice cases (National Labor Relations Board)Source last updated Oct 26, 2026 · refreshed every 7 daysCharges against employers (CA cases) from NLRB's Recent Charges and Petitions Filings report and case pages; the case search is not used (robots.txt disallows /search/). NLRB Bulk Data Policy (read 2026-10-23): "We ask, however, that bulk data extractions be scheduled for non-peak hours (between 11pm EST - 6am EST)"; "automated programs not run more than 5 concurrent requests from the same IP, or the total requests exceed 20 per 200ms." NLRB records are U.S. government works (17 U.S.C. 105). Disclaimer: "While we try to keep the information timely and accurate, we make no guarantees."
- SEC EDGAR XBRL financial data (U.S. Securities and Exchange Commission)Source last updated Oct 28, 2026 · refreshed every 7 daysIncome-tax expense and pre-tax income from each 10-K's XBRL data.
- SEC EDGAR proxy statements (DEF 14A) (U.S. Securities and Exchange Commission)Source last updated Oct 27, 2026 · refreshed every 30 daysCEO pay ratio and median employee pay (Item 402(u)) and the board's women share, read by fixed patterns; statements the patterns cannot read go to the hand-entry queue.
- SEC EDGAR annual reports (10-K) (U.S. Securities and Exchange Commission)Source last updated Oct 27, 2026 · refreshed every 30 daysThe employee count stated in the 10-K's human-capital section.
- OSHA inspections and violations (U.S. Department of Labor, Occupational Safety and Health Administration)Source last updated Oct 27, 2026 · refreshed every 7 daysAll employers OSHA or a state plan reaches. Read from the complete datasets DOL publishes for download (data.dol.gov data catalog) and, for MSHA, MSHA's Open Government Data files; no key is needed. DOL data are U.S. government works (17 U.S.C. 105); DOL's copyright page could not be read from this network (403), so a separate data licence is not verified.
- Wage and Hour Division compliance actions (U.S. Department of Labor, Wage and Hour Division)Source last updated Oct 27, 2026 · refreshed every 30 daysConcluded compliance actions since fiscal year 2005, updated quarterly; WHD publishes no page per case, so rows link to the dataset. Read from the complete datasets DOL publishes for download (data.dol.gov data catalog) and, for MSHA, MSHA's Open Government Data files; no key is needed. DOL data are U.S. government works (17 U.S.C. 105); DOL's copyright page could not be read from this network (403), so a separate data licence is not verified.
- MSHA assessed violations (U.S. Department of Labor, Mine Safety and Health Administration)Source last updated Oct 27, 2026 · refreshed every 7 daysMines only; matched through MSHA's own record of each mine's controller. Read from the complete datasets DOL publishes for download (data.dol.gov data catalog) and, for MSHA, MSHA's Open Government Data files; no key is needed. DOL data are U.S. government works (17 U.S.C. 105); DOL's copyright page could not be read from this network (403), so a separate data licence is not verified.
- EEOC litigation press releases (U.S. Equal Employment Opportunity Commission)Source last updated Oct 27, 2026 · refreshed every 7 daysLawsuits the EEOC filed and resolved, read from its newsroom list (no dataset of lawsuits exists; the newsroom search is not used because robots.txt disallows it). Charges settled before a lawsuit are left out. EEOC releases are U.S. government works (17 U.S.C. 105). EEOC disclaimer (read 2026-10-23): "While we try to keep the information timely and accurate, we make no guarantees."
- EPA ICIS federal enforcement and compliance cases (ECHO) (U.S. Environmental Protection Agency)Source last updated Oct 27, 2026 · refreshed every 7 daysFormal administrative and judicial federal cases, updated weekly by EPA. EPA: "the data quality before November 2000 has not been assessed and should be considered unknown". Cases are found by the defendant's name; cases against a subsidiary not in the company's list are not found.
- EPA ECHO facility compliance and demographics (U.S. Environmental Protection Agency)Loaded · refreshed every 7 daysThe company's TRI and GHGRP facilities as ECHO lists them now. ECHO no longer publishes a national-percentile flag for the area around a facility (EJScreen was withdrawn in 2025), so C-ENV-6 cannot be computed from it.
- EPA Toxics Release Inventory (Envirofacts) (U.S. Environmental Protection Agency)Loaded · refreshed every 30 daysFacilities report each year for the year before; the newest reporting year is used. The parent is the one each facility reports (EPA's standardized parent name where EPA has one).
- EPA Greenhouse Gas Reporting Program (Envirofacts) (U.S. Environmental Protection Agency)Source last updated Oct 28, 2026 · refreshed every 30 daysEach facility reports its highest-level U.S. parent companies with their ownership shares; a company is given its share of a facility's direct emissions. The newest year in Envirofacts is used (2023 on 2026-10-23). EPA proposed in 2025 to end reporting for most source categories.
- USAspending.gov federal awards (U.S. Department of the Treasury (USAspending.gov))Source last updated Oct 28, 2026 · refreshed every 7 daysContracts obligated by ICE and CBP, by the Bureau of Prisons and the U.S. Marshals Service (detention), and by the Department of Defense, to recipients whose name is one of the company's names, and to their child recipients. Amounts are obligations per fiscal year as USAspending reports them.
- FEC committee totals (corporate PACs) (Federal Election Commission)Source last updated Oct 28, 2026 · refreshed every 7 daysCorporate PACs connected to the company: what they gave to federal candidates' committees per two-year cycle. Committee and organization data only: individual contributors are never loaded.
- Senate LDA.gov lobbying reports (U.S. Senate Office of Public Records (LDA.gov))Source last updated Oct 28, 2026 · refreshed every 7 daysQuarterly lobbying reports (LD-2) naming the company as client: the amount reported and the bills named. Shown, not graded; the forms do not say for or against. Data accessed from LDA.gov on the date shown."Senate Office of Public Records cannot vouch for the data or analyses derived from these data after the data have been retrieved from LDA.gov"
- CFPB enforcement actions and consumer complaints (Consumer Financial Protection Bureau)Source last updated Oct 28, 2026 · refreshed every 7 daysEnforcement actions resolved by an order or judgment count; pending actions and consumer complaints are shown as allegations, never scored. The complaint database is not a statistical sample.
- DOJ Antitrust Division and FTC cases (U.S. Department of Justice and Federal Trade Commission)Source last updated Oct 28, 2026 · refreshed every 7 daysAntitrust Division press releases and FTC cases naming the company. Resolved outcomes (guilty pleas, settlements, consent decrees, final judgments, closed FTC cases) count; lawsuits still pending are shown as allegations.
LDA.gov: "Senate Office of Public Records cannot vouch for the data or analyses derived from these data after the data have been retrieved from LDA.gov"
CSWH is not affiliated with these agencies. No agency seal or logo is used.
Food report card
Food report cards: how we grade
Method v1.0, Oct 23, 2026
We grade food brands on three kinds of public record: FDA recalls, FDA warning letters, and ingredients that a federal or state law has revoked, banned or flagged.
Every grade is arithmetic over the records listed on the card, with the method version it was computed with. No model or person adjusts a grade.
A grade is CSWH's assessment under this published method, based on the records listed. It is not a safety rating, and it says nothing about any one product's health effects.
Recalls are a history. For recalls happening now, see FDA and USDA.
For current recalls see FDA and USDA. FDA: https://www.fda.gov/food/recalls-outbreaks-emergencies/recalls-foods-dietary-supplements · USDA: https://www.fsis.usda.gov/recalls
What is graded
A brand under one brand owner, as FoodData Central lists them (for example Cheerios, owned by General Mills). Products listed with no brand name form the owner's own listing.
Brands with at least 5 products in USDA FoodData Central's Branded Foods data, plus brands members ask for (decision D2).
Windows: The last 5 years, counted back from the day the cards were made. Ingredients: The brand's products as currently listed in FoodData Central (discontinued and non-U.S. products left out).
F-REC-1 Recalls by severity (40%)
FDA recall events (openFDA food enforcement reports) matched to the brand. One event counts once, however many products it covers; its class is the most serious of its records.
- Points per event: Class I 10, Class II 4, Class III 1
- 0 points: 100
- 1 to 4 points: 80
- 5 to 10 points: 60
- 11 to 20 points: 40
- 21 to 40 points: 20
- more than 40 points: 0
F-REC-2 Recall reasons (shown, not graded)
The reason FDA gives, grouped by a fixed keyword list: pathogen (Listeria, Salmonella, E. coli and others), undeclared allergen, foreign material, other.
Shown as a breakdown; not scored.
F-ING-1 Tier A ingredients: revoked or banned (25%)
Share of the brand's products (with an ingredient statement) that contain an ingredient revoked for food by FDA, or banned in all food by an enacted state law that no court has blocked.
100 at 0%; 0 at 20% or more; a straight line between (for example 10% gives 50).
- 0%: 100
- 10%: 50
- 20% or more: 0
F-ING-2 Tier B ingredients: being phased out or needing a warning (15%)
Share of the brand's products that contain one of the six dyes FDA is working to remove by the end of 2027 (Red 40, Yellow 5, Yellow 6, Blue 1, Blue 2, Green 3), an ingredient on the Texas SB 25 warning-label list, or one on the West Virginia HB 2354 list.
100 at 0%; 0 at 60% or more; a straight line between (for example 30% gives 50).
- 0%: 100
- 30%: 50
- 60% or more: 0
F-ING-3 Tier C: IARC classifications (shown, not graded)
Ingredients named in a statement that correspond to an agent IARC classifies in Group 1 or 2A. Each is cited singly, quoting IARC's own name for the agent, with a link to its monograph (decision D8).
Shown only; never scored.
F-ENF-1 FDA warning letters (20%)
Warning letters from FDA food offices (the Human Foods Program, CFSAN, the human and animal food operations divisions, and import divisions writing about food) to the brand owner.
100 at 0 letters; minus 30 for each letter; never below 0.
- 0 letters: 100
- 1 letter: 70
- 2 letters: 40
- 3 letters: 10
- 4 or more: 0
F-PLG-1 FDA dye pledge (shown, not graded)
The brand owner's entry in FDA's tracker of pledges to remove petroleum-based dyes: Complete or In progress.
Shown as a note; not scored (it already shows as fewer dyes in F-ING-2).
From indicators to a grade
The weighted average of the scored indicators that have data: recalls 40%, Tier A ingredients 25%, Tier B ingredients 15%, warning letters 20%. When the warning-letter list could not be read at all, the other weights are used alone.
The score is rounded to a whole number (halves up) before its letter is chosen: 84.5 is 85, an A.
No grade at all unless the brand has ingredient data (at least one product with an ingredient statement) and has been checked against FDA recall data.
- A: 85 to 100
- B: 70 to 84
- C: 55 to 69
- D: 40 to 54
- F: 0 to 39
Matching recalls to brands
Store-brand and co-packer recalls are shown under the recalling firm only. They are never put on a national brand without a UPC match.
Warning letters and dye pledges are matched to the brand owner by the same name rule.
No fuzzy or guessed match ever counts.
- Exact: A UPC in the recall's product description or code information matches a product's GTIN in FoodData Central. On that product's brand.
- Strong: The recalling firm is the brand owner (the same name once legal words such as Inc or LLC are set aside, or a cited alias) and the brand's name appears in the product description. On that brand.
- Firm only: The recalling firm is a brand owner but no brand is named. On the owner's page, never on a brand.
- Other: A recall that names a brand but comes from another firm. Goes to our review queue; counts only after a person confirms it.
Ingredients
Deterministic, against the CSWH watch list's synonyms: for example "FD&C Red No. 40", "Red 40", "Red 40 Lake", "Allura Red AC" and "E129" all count as Red 40.
Whole words only: "red" alone never matches, and "Red 40" never matches "Red 400".
A synonym never spans two listed ingredients ("orange, B vitamins" is not Orange B), and a negated mention ("non-bromated flour", "BVO free") does not count.
Each match names the law or FDA action behind the entry and its effective date. Ingredient rows state the law, never a health claim.
Tier A: revoked for food by FDA, or banned in all food by an enacted state law that no court has blocked. Tier B: the six FDA phase-out dyes, the Texas SB 25 list and the West Virginia HB 2354 list. Tier C: IARC Group 1 or 2A agents (shown only).
Reviewed by hand every month; each entry cites the law or FDA action, its link and effective date. A law under court challenge is labelled as such.
Each ingredient row shows the date of the FoodData Central record. A brand that has changed a recipe can use "Report an error" to point at newer data.
Not used
- USDA FSIS recall data (meat, poultry, egg products): Its API returned HTTP 403 to our server as well as to our office network on 2026-10-23, so its terms and data could not be verified. Under decision D1 it stays out until it is verified from the server; meat, poultry and egg recalls are therefore not on the cards yet.
- Open Food Facts: Not at launch (decision D7): FoodData Central is public domain with no share-alike duty.
- EWG Food Scores: Its terms forbid deriving data from it without written permission.
- EFSA OpenFoodTox: Licensed CC BY-ND (no derivatives); we may link to EFSA opinions but do not use the data.
- School-meal additive laws: They cover food served in schools, not food sold in stores.
- Alcoholic beverages as an ingredient (IARC Group 1): IARC's agent is drinking alcoholic beverages, not an ingredient in a cooked food.
Required wording
- Data provided by the U.S. Food and Drug Administration
- Source: USDA FoodData Central
- CSWH is not affiliated with FDA, USDA or any agency named here.
Corrections and cost
Anyone, including a brand owner, can report an error on any card or row. Reports are answered within 10 business days; accurate facts stay, wrong matches and figures are fixed, and every change goes in the card's public change log.
The data is free and no model calls are made: $0.
Sources
- openFDA food enforcement reports (FDA recalls) — Recall records from 2004 on, with the recalling firm, product description, code information, reason, class, dates and event id. openFDA asks that its data not be used to warn the public about recalls or to track a recall's progress, and says FDA does not update a recall's status after it is classified. So the cards are a history: the status shown is the status reported by FDA at classification, and for current recalls we point to FDA and USDA. Weekly."Data provided by the U.S. Food and Drug Administration"
- USDA FoodData Central, Branded Foods — Branded products with their brand owner, brand, GTIN/UPC and full ingredient statement: the brand list and the ingredient data. The bulk release when FDC publishes a new one (checked weekly); the FDC search API for brands members ask for."Source: USDA FoodData Central"
- FDA warning letters — Letters by company, issuing office, subject and date, each with its letter page. From the warning-letters page's own table data, which has every letter and its link. The page's XLSX download is not used: it gives at most 1,000 rows a file and its pages repeat (checked 2026-10-23: six downloads gave 1,792 different letters of 3,768). FDA notes that later events "may have changed the regulatory status of issues discussed in the letter". Weekly."Data provided by the U.S. Food and Drug Administration"
- FDA tracker of food-industry pledges to remove petroleum-based dyes — Each company's pledge, the products or brands it covers, and its status. Weekly."Data provided by the U.S. Food and Drug Administration"
- Federal and state additive laws (the CSWH watch list) — The laws and FDA actions behind each watched ingredient: FDA's revocations of Red No. 3, BVO and Orange B; the FDA dye phase-out; California AB 418; Texas SB 25; West Virginia HB 2354. Hand review every month.
- IARC Monographs (show-only links) — The group of a single agent, cited with a link to its monograph. Hand review every month.
Corrections
- Anyone, the company included, can report an error on any card or any record: use "Report an error".
- We confirm your email address first (members use their account), and reports are limited per hour and per day.
- We answer within 10 business days (Monday to Friday), from team@changestartswithher.org.
- Accurate facts stay. Wrong matches and wrong figures are fixed.
- A company may add a short response with a link; it is shown under the record.
- Every change made because of a report or a staff decision is listed in the card's public change log.
Write to team@changestartswithher.org with any question about a card.