{
  "constructed": true,
  "plannedBeforeObservation": {
    "question": "Does this bounded observed comparison justify a next draft change?",
    "conditions": [
      "A",
      "B"
    ],
    "postAgeDays": 7,
    "metric": "entered impression events",
    "randomized": false,
    "causalInference": false,
    "keep": "same declared account/metric scope and observation age",
    "unknown": "topic, distribution and audience differences remain; no private viewer joins"
  },
  "rows": [
    {
      "id": "A1",
      "condition": "A",
      "value": 80
    },
    {
      "id": "A2",
      "condition": "A",
      "value": 100
    },
    {
      "id": "A3",
      "condition": "A",
      "value": 120
    },
    {
      "id": "B1",
      "condition": "B",
      "value": 85
    },
    {
      "id": "B2",
      "condition": "B",
      "value": 105
    },
    {
      "id": "B3",
      "condition": "B",
      "value": 900
    }
  ],
  "expected": {
    "medianA": 100,
    "medianB": 105,
    "meanA": 100,
    "meanB": 363.3333333333333
  },
  "missingExample": {
    "id": "B2",
    "value": null,
    "status": "unavailable; not zero"
  },
  "conclusion": {
    "en": "The entered B median is five events higher, while its mean is dominated by one 900-event record. This tiny unrandomized set cannot establish that the changed writing choice caused a gain. Keep a next drafting decision separate from a universal performance claim.",
    "de": "Der eingetragene B-Median liegt fünf Ereignisse höher; sein Mittelwert wird vom 900-Ereignis-Eintrag geprägt. Diese kleine nicht randomisierte Menge belegt keine Wirkung der geänderten Schreibentscheidung. Eine nächste Entwurfsentscheidung bleibt von einer allgemeinen Leistungsbehauptung getrennt."
  }
}