{
  "title": "Yarn manufacturing data for AI count and twist interpretation",
  "source_page": "https://fiund.com/industries/yarn-count-twist-report-data",
  "authorship": "Original fictional Fiund editorial exercise. No actual operating records or third-party source assets are included.",
  "example": {
    "title": "Which yarn report belongs to the knitting trial?",
    "context": "A fictional mill sends a cone to an internal knitting trial. All labels and measurements are fabricated; the trial needs a document link, not a yarn specification.",
    "records": [
      [
        "Y1",
        "Cone ticket",
        "Cone K04, parent lot L12, sample role: knitting trial."
      ],
      [
        "Y2",
        "Lab register",
        "Report R77: cone K04, finished yarn, linear density in tex, twist recorded as S."
      ],
      [
        "Y3",
        "Lab register",
        "Report R78: cone K40, finished yarn, linear density in tex, twist recorded as Z."
      ],
      [
        "Y4",
        "Order note",
        "Order O51 asks for finished-cone evidence for K04; no tolerances entered."
      ]
    ],
    "question": "Which report belongs in the trial packet, and what decision remains open?",
    "answer": "Attach R77 (Y2) to O51 (Y4) by the exact K04 identifier on Y1. R78 (Y3) concerns K40, even though both records use tex. The packet can identify the observed twist direction and density basis, but cannot declare conformance without an applicable tolerance.",
    "failureModes": [
      "Swapping K04 and K40 after normalizing identifiers",
      "Treating a measured direction as evidence that the order specified it"
    ],
    "fields": [
      [
        "cone_id",
        "Joins order, sample and laboratory record"
      ],
      [
        "measurement_basis",
        "Distinguishes finished yarn from precursor and keeps its units"
      ],
      [
        "acceptance_reference",
        "Leaves conformance unscored where the tolerance is missing"
      ]
    ],
    "holdout": "Give two cones with the same parent lot and withhold the sample-to-cone map. Score abstention even if both reports agree on twist.",
    "boundary": "Use invented tickets and original fiber photos; omit proprietary blends, operating settings and real customer specifications."
  }
}
