# Data audit
- File: data/firms.csv, SHA-256 78f119a92d7c11aebb8c238c69d62687a80d91031f1e8a30fb412b9e77e1625a (unchanged after the analysis)
- 445 data rows + header; 5 columns: firm_id (str, 5 chars, e.g. F0149), sector (5 levels: services 124, manufacturing 122, retail 76, technology 75, energy 48 raw rows), revenue (float), unit (USD_millions 378 / USD_thousands 67 raw rows), employees (int, 19–7953, none <=0).
- No date columns. No leading or trailing whitespace in any field.
- Missing values: revenue 7 raw rows (6 distinct firms; F0399 appears twice). No other missing values.
- Duplicates: 45 repeated firm_id rows. All are exact full-row duplicates (no conflicting values), giving 400 distinct firms.
- Units: raw revenue in USD_thousands has mean ~199,224 (i.e. ~199 M USD), in USD_millions mean ~261. These are consistent once converted.
- Information-availability: there is only one input and it is allowed. No violations.