# Contract (P0)

Question: Build a documented analytic dataset (one row per firm, revenue in millions USD) from data/firms.csv and report basic aggregates. Estimands: (1) number of distinct firm_id; (2) number of distinct firms with unknown revenue; (3) n firms used in revenue aggregates; (4) mean and (5) median revenue in millions USD over firms with known revenue; (6) mean employees over all distinct firms; baseline: mean of raw revenue column over non-empty rows, no cleaning. Meaningful assuming duplicate rows are assembly errors (one row per firm intended) and unit correctly states each row's currency scale. Outputs: out/results.json (task_id, n_unique_firms, n_missing_revenue, n_analysed, mean_revenue_musd, median_revenue_musd, mean_employees, mean_revenue_raw, feasibility, feasibility_reason, data_issues, n_input_rows, random_seed); out/analytic.csv (firm_id, sector, revenue_musd, employees; n_analysed rows); out/exclusions.csv (step, rule, n_before, n_after, n_removed, rationale; chain from rows read to n_analysed); out/report.md (6 sections); out/analysis.py; out/claims.json. Information rules: only data/firms.csv; revenue in millions USD everywhere; missing-revenue rows excluded from revenue aggregates and listed in the ledger. Invalidity: conflicting duplicate records with no resolution rule; unknown units; unparseable revenue. Would make result partial.