{
  "kind": "synthetic-acceptance-worksheet",
  "solution": "shelfsentra",
  "title": "A small shelf-check example",
  "notice": "These four records are invented to show an acceptance check. Each observation checks one SKU across all its assigned facings in a defined shelf area, with all required reference views assessable. They are not camera detections produced by ShelfSentra or results from a store.",
  "columns": ["Shelf check", "Staff observation", "Candidate result", "Expected evaluation label"],
  "rows": [
    ["A", "Checked SKU absent", "Stockout alert", "Correct alert"],
    ["B", "Checked SKU present", "Stockout alert", "False alert"],
    ["C", "Checked SKU absent", "No stockout alert", "Missed stockout"],
    ["D", "Checked SKU present", "No stockout alert", "Correct absence of alert"]
  ],
  "interpretation": [
    "This sample contains two staff-confirmed stockouts within the defined shelf area, but only one was flagged. It also contains two alerts, one of which was false. Counting alerts alone would hide both problems.",
    "The four labels are one correct alert, one false alert, one missed stockout and one correct absence of an alert. Four observations illustrate the labels; they do not estimate store-wide performance.",
    "Define each check by SKU, shelf area, observation time and applicable planogram. Staff check all assigned facings for that SKU in the defined shelf area, and all required reference views must be assessable. An empty facing alone does not establish absence when another assigned facing contains the SKU. Keep an obstructed view in a separate unassessable category rather than counting it as a correct detection or proof of availability.",
    "If recovery time is measured in a real evaluation, define its start and end events, such as task creation and staff-confirmed correction. That interval is not automatically the full duration of the stockout."
  ]
}