Executable

Does this page say the coffee is organic?

ALS-COFFEE-1.2-CERT-001

This question is asked as a test and produces a result for a real page.

Assertion

organic_claim_stated is_stated true

Applicability

applies when
the product is roasted coffee offered to consumers in any format
signal
product_type or title indicates coffee

Accepted evidence

  • product_description — an explicit organic statement naming the certifying agent e.g. Certified organic by Oregon Tilth.
  • structured_data — the same statement inside the JSON-LD product description
  • product_faq — the same statement in FAQ structured data

Explicitly insufficient evidence

What does not count, and why. Published because the omission is where a check quietly becomes a claim.

  • the bare word "organic"
    An operation selling under the US small-operation threshold may lawfully call its product organic without certification, so the word alone is evidence of nothing checkable.
  • "organically grown", "organic practices", "beyond organic", "transitional"
    None of these sits inside the four regulated label tiers, and "beyond organic" specifically is marketing that positions itself as superior to the only checkable category.
  • "made with organic ingredients"
    That is a distinct and lower labelling tier with a seventy per cent threshold which may not use the seal, and a naive check scores it as equivalent to organic.
  • a rendered organic seal image with no agent named
    Image content is not page text and the seal resolves to nothing on its own; the certifier's name is the field the rule requires precisely because the seal is unverifiable without it.
  • a farm's or exporter's certificate offered while the roaster is uncertified
    Only a certified handler may use the seal, so the register must resolve to the roaster; this is the most legitimate-looking substitution in coffee because the farm genuinely is certified.
  • an EU control-body code such as XX-BIO-NNN treated as identifying the seller
    The code identifies the control body that audits, never the operator audited, so a check that resolves the code has verified nothing about this seller.
  • "certified organic since 2011"
    A certification history is not a current status, and the register publishes surrendered, suspended and revoked precisely because real certificates stop being real.

Conflict rules

  • When the description claims organic and the shipping policy page also contains the word — then read only product surfaces; a document about orders is not evidence about this product precedence: product_description > structured_data
  • When an organic claim and a contradicting statement both appear — then the engine reports the contradiction and does not pass the row, which is the correct fail-closed behaviour

What a pass licenses

A pass establishes: That readable product text on this page states an organic claim of some form, quoted verbatim where a clean sentence exists.

A pass does NOT establish: That any operation is certified, that a certifying agent was named, that the claim resolves in any register, or that the claim is about the coffee rather than the packaging.

Discrimination

Measured fail rate: 92.0% (92 of 100 adjudicated)

95% interval: 85.0% – 95.9%, against a 15-85% target band (wilson95 vs target band)

Verdict: Not discriminating — the whole 95% interval lies outside the target band, so almost every store answers the same way. This is evidence for a retirement decision and is not the decision.

Declared bias in the instrument

A 95% interval bounds SAMPLING error and nothing else. These are the known biases in the measuring instrument for this row, declared rather than netted — where two point in opposite directions and neither is quantified, combining them into one signed number would be a guess wearing a direction.

  • deflates fail rate, 2pp
    Measured false-pass rate of this engine on this entry, from the row-by-row audit of the same run (experiments/v3-2/verdicts.json): 2 of this entry's 8 pass rows were confirmed wrong. Correcting them moves rows from pass to fail, so the measured fail rate is BELOW the true one.
  • inflates fail rate, UNQUANTIFIED
    The bounded semantic tier was DISABLED for this run: experiments/v3-2/run_standard.ts sets PRODUCT_TEST_SEMANTIC=0. It applies only to requirements of kind `claim` (src/server/productTest.ts#applySemanticTier). Its GRANT path promotes a claim the lexical term list missed; with the tier off, a page that states this claim in wording the vocabulary does not carry is recorded as a fail, so this path pushes the measured rate ABOVE the true one. UNQUANTIFIED: the audit read PASS rows only, and fail rows carry no false-negative information, so no artifact in this repo bounds it.
  • deflates fail rate, UNQUANTIFIED
    The same disabled tier, its OTHER path. A semantic VETO withdraws a lexical match whose sentence is about something else; with the tier off those matches stand as passes, so this path pushes the measured rate BELOW the true one. Declared separately and NOT netted against the grant path above: neither is measured, so nothing establishes which dominates, and a single signed claim about the tier would be a guess wearing a direction.

⚠️ THIS VERDICT MAY NOT BE ACTED ON AS A RETIREMENT, and the document says so rather than working around it. The rule `bias_exceeds_margin` blocks a retirement whose margin to the band edge is smaller than a declared bias pointing at that edge, and an unquantified bias blocks it outright. This is a claim row measured with the semantic tier off, which is exactly such a bias. The entry therefore stays `executable` and keeps running. A measured `not_discriminating` verdict is EVIDENCE for a retirement decision; it is not the decision. Recomputed by `standards/coffee/issue_v1_2.ts` from experiments/v3-2/standard_run3.jsonl through the shipped `verdictFor` in standards/discrimination.ts; no re-run, no network, no spend. Cross-checked against experiments/v3-2/discrimination.json and standards/coffee/v1.0/fitness.json.

Authored expectation: no prediction, confidence low. The numeric band this standard used to carry is not published at this grammar version.

Organic is common in specialty coffee but far from universal, and the engine's dictionary matches the bare word as well as the certified phrasings, so the row will pass wherever the word appears at all. The band is wide because the pass condition is looser than the question a buyer is asking.

The numeric fail-rate band this reasoning was authored with is NOT carried at grammar 1.2 and is not reproduced here. Authored bands held 1 of 10 against this category's own sample and every miss was HIGH, so a direction derived from the band would inherit the error that condemned it (SCHEMA.md §9.3 step 3). The band survives in ALS-COFFEE v1.1 (standard_hash f8ec2780f60c38931913e5b6cd37506500c8462709209de7180ba6691d6137e7), which is byte-frozen and still served.

Known gaps

  • "A gift bundle of our house blends." with an organic phrase in the shipping policy's page title — false_pass
    Claim rows do not exclude the shipping-policy surface and the policy fold-in carries SEO chrome, so a claim about the product can be proven from a document about orders. Found on a real tea store and being fixed concurrently.
  • "Love that it's fragrance-free!" — a customer in Portland. — false_pass
    A review quote credited as the store's own statement. Applies to every claim row, and coffee pages carry heavy review content.
  • "The tube is BPA-free." — false_pass
    A container property credited to the contents. The coffee analogue is an organic-cotton or compostable-bag claim crediting the coffee.
  • "Most cheap versions are made from thin stamped steel." — false_pass
    The comparative veto tolerates only one word between "most" and its verb, so "Most supermarket blends are organic" would credit a competitor's property to this product.

For a shopper

The word organic on its own can be lawful without any certification for a small seller, so what matters is whether a certifying agent is named. You can check a named agent and the roaster's legal name against the USDA Organic INTEGRITY Database yourself, free, and see whether the certificate is current.

For a merchant

If a certifying agent certified the operation that packs this coffee, name that agent in product-page text alongside the legal operation name, so a reader can resolve the pair in the public register rather than taking a seal image on trust.

Grounding

  • US National Organic Program, 7 CFR 205.303 (legal_instrument)
    that where a packaged product is labelled organic the certifying agent's name must appear on the information panel, which is why a named agent rather than a seal is the evidence a reader needs
  • USDA Agricultural Marketing Service, published organic enforcement settlements (regulator)
    that organic misrepresentation is a real and enforced failure mode in coffee specifically, including a roaster that settled for a fifteen-thousand-dollar civil penalty for representing products as organic without certification
  • CCOF, guidance on using the Organic INTEGRITY Database (certification_body)
    that the register is described by a certifying agent authorised under the national organic programme as the official place to check whether an operation is certified, which is the intended verification route rather than the agent's own word