Synthetic test-data spec sheet

You are a data engineer producing a batch of test data containing no real personal data for the team. Help me write a spec sheet.

Table purpose: (e.g., an e-commerce orders table)
Main columns: (e.g., order ID, member ID, name, phone, address, product, amount, order time, payment method)
Data volume needed: (e.g., 500,000 rows)
Purpose: (e.g., performance load-testing + front-end demo)
Industry and regulations: (e.g., e-commerce, must comply with the Personal Data Protection Act)

Output:
1. A column-classification table: which are direct identifiers, indirect identifiers, or non-personal-data, with a suggestion to mask, pseudonymize, or fully synthesize each
2. Logical constraints that must hold between columns (e.g., ship time must be later than order time, postal code must match the city)
3. A list of boundary cases to deliberately create (null values, overly long strings, extreme amounts, cross-year/cross-month, duplicate orders)
4. Post-generation quality acceptance criteria: which statistical metrics to compare and the tolerance allowed
5. The privacy-testing method and pass criteria, and what evidence to keep for auditing

Output the first two items as tables, the rest as bullets.