The Realism Engine

Every synthetic data tool has the same problem: columns don’t know about each other. Faker generates a random total of $847.23 and a random subtotal of $312.00 — but those numbers have no relationship.

Misata v0.5.2 introduces the Realism Engine, a post-generation pass that enforces 11 cross-column consistency rules.

The 11 Rules

Monetary Integrity

1. total = subtotal + tax + shipping — Every order’s total is the exact sum of its components.
2. cost < price (30–70% margin) — No product costs more than its selling price.
3. line_total = qty × unit_price − discount — Every line item is mathematically computed.
4. discount ≤ 30% of unit_price — Discounts stay realistic.
5. tax = subtotal × rate — Tax is derived, never random.

Temporal Integrity

6. delivered_at > created_at — Deliveries happen 1–14 days after orders.
7. delivered_at = NULL when status ≠ delivered — Pending orders don’t have delivery dates.
8. updated_at ≥ created_at — Updates can’t predate creation.

Identity Integrity

9. email = f(first_name, last_name) — Emails are composed from the person’s actual name.
10. slug = slugify(name) — URL slugs derive from category names.
11. FK referential integrity — Every foreign key points to a real parent row.

The Result

After applying all 11 rules to our 1,515-row e-commerce dataset, we achieved 100% consistency — every single mathematical and logical relationship holds.

Try It

pip install misata==0.5.2
misata generate --smart —db your_database.db

The --smart flag enables the realism engine automatically.