# Contract — F1-dev-01

Question: turn the synthetic firm cross-section data/firms.csv into a documented, analysis-ready dataset (one row per firm) and report basic aggregates. Estimands: (1) the number of distinct firm_id values; (2) the mean and median revenue in millions USD over firms with known revenue; (3) the mean employees over all distinct firms. These only make sense if (a) duplicated rows are true duplicates of one firm, (b) the unit column correctly gives each row's currency scale, and (c) missing revenue is simply excluded (complete-case analysis, no imputation). 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 (a ledger that chains from the raw row count to n_analysed); out/report.md; out/analysis.py; out/claims.json. Information rules: use only data/firms.csv; report revenue in millions USD everywhere; rows with missing revenue are excluded from revenue aggregates and listed in the ledger. Invalidating conditions: conflicting duplicates (same firm_id with different values), unknown units, or a ledger that fails to reconcile.