{
  "example_kind": "original synthetic manual planning case",
  "slug": "forecast-interval-coverage-ledger",
  "title": "Audit which outcomes fall inside forecast ranges",
  "scenario": "Synthetic planning case: Four invented duration ranges have inclusive endpoints: A2–4 actual 3; B3–5 actual 6; C4–8 actual 5; D2–3 actual 3.",
  "layout": "evaluation",
  "dataset": {
    "heading": "Inspect the invented case records",
    "headers": [
      "Case",
      "Inclusive range days",
      "Actual days",
      "Inside?",
      "Width days"
    ],
    "rows": [
      [
        "A",
        "2–4",
        "3",
        "Yes",
        "2"
      ],
      [
        "B",
        "3–5",
        "6",
        "No",
        "2"
      ],
      [
        "C",
        "4–8",
        "5",
        "Yes",
        "4"
      ],
      [
        "D",
        "2–3",
        "3",
        "Yes, at endpoint",
        "1"
      ]
    ]
  },
  "derivation": "Inside count=3 of 4. Coverage=3/4=75%; mean width=(2+2+4+1)/4=2.25 days. Coverage alone ignores how wide the ranges are.",
  "result": "Three of four outcomes fall inside the displayed ranges, or 75% observed sample coverage. Preserve widths and the missed case; four invented outcomes do not establish future coverage.",
  "boundary": "The guide checks interval membership and width rather than a percentile of completed-item history.",
  "limitations": "The ranges are supplied predictions without a verified probabilistic model. No confidence or prediction-interval calibration is established.",
  "sources": [
    "forecast"
  ],
  "tasks": [
    [
      "Record the endpoint convention",
      "Evaluation owner",
      "Both ends are inclusive; D’s actual three is therefore inside."
    ],
    [
      "Retain misses and widths",
      "Reviewer",
      "B is outside and all four interval widths remain visible."
    ],
    [
      "Limit the confidence statement",
      "Planning lead",
      "The 75% figure is descriptive of this toy set and is not a calibrated next-task probability."
    ]
  ],
  "faqs": [
    [
      "Could very wide ranges improve coverage?",
      "They can contain more outcomes while becoming less useful; retain width and the decision purpose."
    ],
    [
      "Does 75% mean the next forecast is 75% reliable?",
      "No. A tiny synthetic sample cannot establish that future probability."
    ]
  ],
  "method_references": [
    {
      "id": "forecast",
      "title": "Forecasting: Principles and Practice — accuracy",
      "url": "https://otexts.com/fpp3/accuracy.html",
      "scope": "Genuine held-out forecast and point-error evaluation context. Binary scoring examples use their own explicit toy definitions; no real task model is validated."
    }
  ]
}