{
  "version": 1,
  "assetId": "conversion-vs-ai-output-matrix-2026-07",
  "generatedAt": "2026-07-20",
  "license": "CC0-1.0",
  "synthetic": true,
  "containsPersonalData": false,
  "contentTarget": {
    "path": "/compare/pdf-to-excel-vs-ai-extraction",
    "status": "asset-ready-live-noindex",
    "indexingBoundary": "This public matrix does not authorize an index change. The live comparison page remains noindex until named ownership and approved English-native review are recorded."
  },
  "source": {
    "sampleId": "commercial-invoice",
    "url": "/samples/document-library/commercial-invoice.pdf",
    "sha256": "cb430ae7bad1f4270a9594fb0da0d6eb1792beb9240ca76c8fefa31b1e0af84b"
  },
  "conversionOutput": {
    "benchmark": {
      "id": "2026-07-19-paddlevl16-v1",
      "provider": "paddle_vl",
      "model": "PaddleOCR-VL-1.6",
      "resultsUrl": "/benchmarks/conversion-quality/2026-07-19-paddlevl16-v1/results.json"
    },
    "actualWorkbook": {
      "url": "/benchmarks/conversion-quality/2026-07-19-paddlevl16-v1/commercial-invoice-actual.xlsx",
      "sha256": "2f6f7e65d30c5c544783b540fd51ada197d8316b3b8424f164ab7bbcb96fc36d"
    },
    "referenceWorkbook": {
      "url": "/samples/document-library/commercial-invoice-reference-output.xlsx",
      "sha256": "b296b96b6fb65e177c0516af52f684dddc1d08f21b1a8d413d08f384eb443dde",
      "meaning": "Declared expected rows for the synthetic source. It is not a model output."
    },
    "result": {
      "expectedCells": 24,
      "matchedCells": 24,
      "expectedRows": 3,
      "matchedRows": 3,
      "headerSequenceMatched": true,
      "cellExactMatchRate": 1,
      "rowExactMatchRate": 1
    },
    "evidenceBoundary": "The actual workbook and metrics come from one versioned benchmark run on one synthetic selectable-text invoice. They do not establish a general conversion accuracy rate."
  },
  "fieldExtractionShape": {
    "status": "plain-language-fields-and-declared-reference-preview",
    "fieldKeys": ["line_description", "hs_code", "quantity", "line_total_usd"],
    "maximumPreviewRecords": 3,
    "publicPresentation": "Field labels and review instructions are shown in natural language. No machine-readable extraction schema is offered for download.",
    "valueOrigin": "The named-field side uses declared synthetic source values to illustrate field shape. It is not an AI model output, historical customer output, or a promise about a real document."
  },
  "matrix": [
    {
      "dimension": "Input source",
      "conversion": "The same public synthetic commercial-invoice PDF.",
      "fieldExtraction": "The same public synthetic commercial-invoice PDF."
    },
    {
      "dimension": "Delivered form",
      "conversion": "A versioned actual XLSX benchmark workbook retaining the detected six-column line-item table.",
      "fieldExtraction": "Plain-language field requests with a bounded three-record declared reference preview."
    },
    {
      "dimension": "Visible data shape",
      "conversion": "Line, description, HS code, quantity, unit price, and line total across three rows.",
      "fieldExtraction": "Only requested fields: line description, HS code, quantity, and line total USD."
    },
    {
      "dimension": "Review work",
      "conversion": "Review table boundaries, columns, number formats, and every source row.",
      "fieldExtraction": "Review field meaning, repeated records, missing values, and ambiguous labels against the source."
    },
    {
      "dimension": "Evidence boundary",
      "conversion": "One synthetic benchmark result, not a product-wide accuracy claim.",
      "fieldExtraction": "A declared field-and-row preview, not a measured AI extraction result."
    }
  ],
  "reviewChecklist": [
    "Open the shared source PDF beside the actual conversion workbook and inspect all six columns and three source rows.",
    "Treat the named-field side as natural-language setup guidance and a declared preview, then use the signed-in product workflow for a real extraction.",
    "Confirm the requested field meaning, repeated-record boundary, missing values, and ambiguous labels before use downstream.",
    "Do not compare only the two output shapes; test representative documents from the intended workflow and quality range.",
    "Do not generalize this one synthetic document to scans, handwriting, different languages, altered layouts, or customer documents."
  ],
  "claimBoundary": [
    "This is a same-source, one-document synthetic output matrix, not a general product accuracy, cost, speed, or capability comparison.",
    "The conversion side is one versioned benchmark run; the field-extraction side is natural-language guidance and a declared reference preview rather than an AI model output.",
    "Neither output is presented as universally better, guaranteed, or sufficient without source review."
  ],
  "requiredPublicationGates": [
    "Assign a named accountable owner for the comparison page and its 30/60/90-day reviews.",
    "Record approved English-native review with reviewer, date, and evidence before indexing.",
    "Keep the live comparison page noindex until its publication record and index-release evidence are complete."
  ]
}