# Checks (P4)

• Clean re-run of out/analysis.py twice: results.json byte-identical (cmp) -> PASS
• analytic rows == n_analysed: PASS
• n_unique - n_missing == n_analysed: PASS
• ledger n_before-n_removed==n_after: PASS
• ledger chained: PASS
• ledger start = rows read: PASS
• ledger end = n_analysed: PASS
• raw rows - total removed = n_analysed: PASS
• firm_id unique in analytic: PASS
• no NaN revenue_musd: PASS
• se positive: PASS
• median<=mean (right-skew) & within range: PASS
• mean_revenue_musd << raw baseline (unit mixing inflates raw): PASS
• independent recompute mean_employees: PASS