Fraud patterns your rules should catch.

Generate card and account transactions with normal spending and realistic fraud mixed in, labelled, so you can test detection rules, models and investigation tools safely.

Generate data free All capabilities

Free trial for every new account. No credit card.

The example

One card's week, compromised at 2 a.m.

What to look for in the screenshot above. These are the details that usually break in hand-made or random test data.

Normal spending first: groceries, fuel and coffee near home, one or two transactions an hour.
Then a burst abroad: high-value card-not-present purchases minutes apart, 7,962 km from the last use.
Velocity and distance features are precomputed per transaction, as a fraud model would see them.
Fraud labels sit on the burst only; the card's next normal purchase back home is clean.

What you get

Realistic fraud typologies

Card testing, account takeover, gift-card cash-out and cross-border bursts.

Features included

Velocity, distance from last transaction, merchant category and time of day.

Labelled outcomes

Fraud labels and dispute reasons for supervised models and rule tuning.

PCI-safe

Card numbers are synthetic and never belong to real cardholders.

How it works

From idea to dataset in three steps

Describe or upload

Describe your data in plain English, or upload your SQL schema for PostgreSQL, MySQL / MariaDB, SQL Server and SQLite.

Review the plan

Check the tables, columns and rules the AI proposes, and adjust anything you like.

Generate and download

Choose how many rows you need and download a ready-to-run SQL script, CSV, Excel (one sheet per table) or JSON.

Questions

Can I control the fraud rate?

Yes. Make fraud rare (like real portfolios) or common for testing recall.

Are card numbers valid?

They pass format and Luhn checks but are synthetic, and are shown masked where appropriate.

Can I add disputes and chargebacks?

Yes. Add disputes and chargeback tables that reference the fraudulent transactions.