1Synthetic bank
Ordinary customers get everyday activity: salaries, rent, bills, card spending, ATM use, transfers to friends, and business sales, suppliers and payroll. Planted scenarios get their own customers, mixed into the same data. The threshold defaults to about US$10,000 in local currency. Change it to match your rules.
1Screening dataset
A fictional watchlist of people, companies, banks, vessels and test locations. Payments mix planted matches with legitimate look-alikes and ordinary activity. No live sanctions lists are downloaded.
1Ownership case pack
Evaluation profiles
Curated fictional cases for testing ownership and control detection. Expectations use dated US, EU and UK guidance profiles. They do not assess real companies or give permission to process a payment.
1Trade-finance case pack
Linked companies, banks, invoices, shipments, letters of credit and payment draws. Reference prices, ports and screening restrictions are fictional test fixtures. Goods-classification cases request review; an HS code alone does not establish a legal restriction.
Credit documents export as structured JSON/CSV/SQL records. They are not SWIFT MT700 messages or bank-ready instruments.
1Schema
Accepts SQL CREATE TABLE (PostgreSQL, MySQL, SQLite, SQL Server), TypeScript interfaces, or a JSON sample.
2Settings
3Tables and columnsScenariosScreening casesOwnership casesTrade cases
Small datasets regenerate as you edit. Use Generate for larger runs.
Each case produces registry, ownership, control, designation and payment records. expected_ownership and case_catalog contain the separate answer key. Post-designation disposals with incomplete facts remain review cases.
Compare invoices with the supplied reference prices and business scope, link draws to shipments, and check the fictional screening records. Documented premium prices, installments and transshipment are legitimate controls. expected_trade_alerts and trade_labels form the separate answer key.
Fields to test
Name variants
BICs, vessel numbers and test-city matches use exact identifiers. Name fields also test aliases, spelling, accents and paired Chinese name forms. expected_matches is the separate answer key; keep it away from your screening system.
Network preview
Ownership and control evidence
Arrows run from owner to company. Dashed lines show ended holdings or board-control rights. No automatic restriction means only that the named test profile did not trigger; it is not a payment-clearance result.
4Output
Payment XML targets base pacs.008.001.08 batches. Watchlist exports use OFAC formats with explicit fictional test markers. Network-specific envelopes and usage rules are separate.
Compare your monitoring alerts
Load your system's alerts as JSON or CSV and compare them with this generated dataset's answer key. Files stay in your browser. Maximum 10 MB and 100,000 alert records.
Generate AML data, then choose an alert file.