id: '192'
narrative_anchor_date: '2028-01-01'
test: Create a concise one-metric baseline card for the original February 2023 mid-segment churn signal.
  Include the date, v1 cohort, churn and comparison baseline, who performed the joint and independent
  reruns, and the bounded reason it became the historical before-snapshot.
load_bearing_facts:
- 308
- 301
- 302
expected_tool_calls:
- create_doc
grade:
  type: tool_trace
  config:
    check_version: 2
    today: '2028-01-01'
    semantic_judge_version: 2
    assertions:
    - type: field_equals
      tool: create_doc
      action_id: create_baseline_card
      path: result.ok
      value: true
      check_id: riley_192_00
    - type: field_llm_judge
      tool: create_doc
      action_id: create_baseline_card
      path: args
      criterion: 'The document must identify the observation as February 7, 2023; define the historical
        v1 mid-segment cohort as 50–200 seat accounts; and state exactly that observed monthly churn was
        5.1% versus a 2.1% trailing-six-month baseline. Missing or conflicting dates, cohort bounds, versions,
        rates, or baseline periods must fail.


        The document must say that Riley and Owen reran the cut together and Riley also reran it solo,
        obtaining the same result, and must conclude only that the signal was treated as a valid/clean
        before-snapshot rather than a cohort-definition artifact. It must frame this as the original historical
        investigation baseline, not as a current metric or a later recovery result.'
      check_id: riley_192_01
mock_state: {}