{
  "recorded_at_utc": "2026-11-01T18:14:15.492433+00:00",
  "stage": "Final pre-outcome method amendment submitted for independent acceptance. No commercial-name collision arithmetic has run. Source framing/encoding and identifier-coverage diagnostics were seen; this is not a preregistration.",
  "supersedes_only": "Parsing specification and primary-versus-secondary name-key designation in protocol-pre-outcome.json. The original protocol and failed attempts remain unchanged.",
  "source_sha256": "16df7e01468ad666465d4b516d17d85c200547478a61bc7310c519ea83179e69",
  "reason": "The official 2020 dictionary declares semicolon and code page1252, but this exact2026 export contains undefinedcp1252 bytes and undocumented quote characters. Independent byte framing checked everyphysical line:12fields. A separate strict standardCSV-per-record check found allbutone record valid and byte-identical to this frame. The conservative analysis excludes ALL quote-bearing required name fields rather than guessing quote semantics, alongside missing/undecodable names. Originalbytes remain private and unchanged.",
  "parser": {
    "record_framing": "For this exactSHA only: split each LF-delimited physical record after removing only its line ending on original semicolonbytes into exactly12fields; fail on anyother width. Require no embeddedCR/LF in fields. This is a byte boundary specification, not a claim that quotes are literal product-name characters. All required commercial/technical fields containingASCIIquotes are unknown for name analysis. Do not strip, repair, unescape or merge source records.",
    "header": "Require the exact 12 named columns in the frozen original order, then access fields by their verified names.",
    "encoding": "Strict cp1252 per field. Undefined byte in a field makes that field uninterpretable. Keep the byte-perfect original private; no errors=ignore, errors=replace, Latin1 reinterpretation or guessed UTF8 repair.",
    "required_fields": "Exact-text registration/cadastroID must be nonblank, strictlycp1252-decodable, unquoted, without source whitespace or malformed form; auditbeforeoutcomes. Commercial name must be nonblank, strictcp1252-decodable and contain noASCIIquote. Technical name follows same rule only for the fixed complete-triple comparison. Anyunknown commercial field contributes no guessed name membership; nonrequired badbytes do not discard an otherwiseinterpretable pair.",
    "missing": "Only empty or all-whitespace fields count as missing. Other placeholder-like strings are not silently converted; schema/identifier audit must flag suspicious nonnumeric or whitespace-bearing identifiers for review before outcomes.",
    "identifiers": "Preserve every nonblank interpretable registration/cadastro field as an exact string, including leading zeros. Do not use process number, raw row or manufacturer count as a registration identity. Audit form anomalies before proceeding.",
    "conflicts": "Retain all distinct known ID-name memberships and all distinct known ID-name-technical triples. Never keep only a first row, impute an unknown name from other rows, or assume multiple labels under one identifier are errors. Report IDs with multiple commercial/technical names separately.",
    "standard_record_crosscheck": "For every physical datarecord, independently strict-parse that single record with standardCSV semicolon/doublequote syntax using a reversible Latin1 bytecontainer only, then require12fields and exact byte equality to physicalframing. Anyexception, changedfield or wrongwidth makes ALL commercial/technical memberships on thatrecord unknown. Its unambiguous strictlydecoded firstfield ID remains in all-IDdenominator. This quarantines the one observed malformed record rather than choosing permissivequote semantics. No commercial/technical key uses an ambiguousrecord."
  },
  "name_keys": {
    "primary": "Literal exact strictlycp1252-decoded commercial field with noASCIIquote, including its punctuation, diacritics, numbers and whitespace. Exclude missing, undecodable or quote-bearing commercial fields. This is a declared interpretable-label subset, not all catalogue labels.",
    "secondary": "The original predeclared conservative key: NFC Unicode, outer trim plus collapse Unicode whitespace, then casefold; retain accents, punctuation, numbers and quotes. Clearly separate its results from literal primary results.",
    "reason_for_designation_change": "Literal primary audit makes the empirical administrative equality test easiest to inspect. Conservative normalization is a separately labeled analytical convention, not official search behavior. This change preceded collision outcomes.",
    "quote_sensitivity": "No quote-inclusive outcome sensitivity is authorized. Publish counts of rows and distinctIDs with quote-bearing requiredfields as unknown coverage. No outcome-driven quote removal or interpretation."
  },
  "denominators": {
    "name_collision": "Distinct known, nonblank, unquoted, strictlydecodable commercial-name keys attached to a usable exactID; collision means a key maps to≥2distinctIDstrings. This percentage describes only this known-key universe. It is not a lowerbound fraction of ALLpossible source labels, because unknown names also affect that denominator.",
    "affected_identifier_share": "Distinct usable exact IDs with at least one confirmed known shared commercial-name key, divided by ALL distinct usable exact IDs in the frozen export. Where any names are unknown/missing, this is a confirmed lower bound, not a complete non-uniqueness prevalence estimate.",
    "technical_comparison": "One fixed set of complete distinct (ID,commercial-name,technical-name) triples. Compare the count of these triple memberships whose commercial key maps to at least two IDs versus whose (commercial,technical) key maps to at least two IDs. Each unit is a triple membership, not necessarily one distinct ID or product; an ID with multiple assignments can appear more than once. Refinement is monotone by construction, not evidence of improved UI search or user outcomes.",
    "unknowns": "Report rawrow, distinctID, knownname membership, missing/undecodable/quote-bearing-field and multiple-name/technical-ID counts. Keep allIDs withunknown names in the overallID denominator and show fullyinterpretable-ID coverage. Do not impute, claim noobservedcollision proves uniqueness or use source rows as distinctdevices."
  },
  "analysis_gate": "Independent source reviewer must accept this exactamendment, then independently compare source/IDconsistency audit. If identity/framing remains unresolved, hold. Main outcomes only after source/required-field acceptance; fullarticle only after independentcalculation and meaningfulincrementalusefulness gates.",
  "unchanged_hard_limits": "No current legal status, physical-device/model/SKU count, patient exposure, safety, clinical risk, counterfeit, quality, physician authorship/review, observed user-error or official-UI performance claims. No public raw rows, names, IDs, corporate fields or source compilation. No provider, native, mail or production actions.",
  "outcome_arithmetic_run": false,
  "full_study_credit": 0,
  "updated_before_outcomes_at_utc": "2026-11-01T18:14:33.648091+00:00"
}
