# Verification checks (P4)

• Clean re-run: deleted results.json/analytic.csv/exclusions.csv, re-ran analysis.py; byte-identical outputs (cmp) -> PASS
• analytic rows == n_analysed: PASS
• analytic firm_id unique: PASS
• no NaN revenue in analytic: PASS
• n_unique_firms - n_missing == n_analysed: PASS
• ledger n_before-n_removed==n_after: PASS
• ledger chained: PASS
• ledger starts at n_input_rows: PASS
• ledger ends at n_analysed: PASS
• n_raw - total removed == n_analysed: PASS
• mean recomputed from analytic.csv: PASS
• median recomputed from analytic.csv: PASS
• mean_employees independent recompute (groupby first): PASS
• baseline recompute: PASS
• independent musd mean (convert via groupby): PASS
• median <= mean (right-skewed, plausible): PASS
• se positive: PASS
• raw baseline >> cleaned mean (thousands rows inflate raw): PASS
• revenue_musd range plausible (max<1e4): PASS