# Data audit

• data/firms.csv SHA-256 78f119a92d7c11aebb8c238c69d62687a80d91031f1e8a30fb412b9e77e1625a (matches bundle.json)
• 445 data rows + header, 5 columns: firm_id (str, 5-char e.g. F0149, no whitespace), sector (5 values: services 124, manufacturing 122, retail 76, technology 75, energy 48 rows), revenue (float, 7 empty), unit (USD_millions 378 rows, USD_thousands 67 rows), employees (int, 19..7953, none missing).
• No date columns (cross-section).
• Duplicates: 45 rows duplicate an earlier firm_id; all 45 are exact duplicates across all columns (0 conflicting). 400 distinct firms.
• Units: USD_thousands values 19142..1005760 (= 19.1..1005.8 M USD), USD_millions 5.742..2057.575 — consistent magnitudes after conversion; no evidence of mislabelled units.
• Missing revenue: 7 rows = 6 distinct firms (F0399 appears twice, both empty): F0048, F0121, F0129, F0269, F0348, F0399.
• No non-numeric strings in revenue/employees, no negatives. No availability-rule violations (single permitted input).