Find the grain of each table
Write down what one row represents. A transaction table, a customer table, and a monthly summary can all describe the same business at different levels. Joining them without that distinction can multiply records or introduce misleading totals.
Ask for a usable dictionary
Column names alone may not explain units, category codes, time zones, or missing values. Request definitions and a small example of each important field. Determine whether an empty cell means unknown, not applicable, withheld, or truly absent.
Check coverage and repeated records
Compare the sample with your required period and regions. Review repeated identifiers and distinguish legitimate revisions from accidental duplicates. Establish how corrections, late records, and changing schemas will be documented.
Document the permitted project
A useful commercial discussion explains both the analytical objective and the planned processing. Ask which source permissions support those uses, what restrictions remain, and whether sensitive fields have been transformed. Technical completeness and licensing suitability are separate acceptance decisions.
Review checklist
- Document types or table schemas
- Languages, regions, and date coverage
- Volume by category
- Ground truth and sensitive-data requirements
- Intended use and licensing scope