Develop and test new systems on data that behaves like production
Built for development and test environments running on structured data, from a single table to relational databases with many linked tables.
Request demoTrusted for software test data by leading organisations
Why the usual test data lets problems through to production
The real data is off-limits
Legal and compliance rules often keep production data out of development and test environments. Where it is allowed, the risk is still high, because every developer and tester can look into sensitive records.
Dummy data misses what breaks your software
Hand-made and rule-based test data looks tidy. Real data has variety, outliers and invalid records, and that is where a pipeline fails. Test without them and the problems show up in production.
The data you need does not exist yet
You want to test specific ranges and scenarios, sometimes ones that never occur in your real data. A performance test needs more rows than the source can supply.
What test data has to deliver, and how BlueGen does it
Data that behaves like production
The full variety and spread of your real data, including the outliers and invalid records your system will meet once it goes live.
A model that learns every variation
BlueGen's model learns from the real data and regenerates the variation it finds, invalid examples included. If the source holds a move registered after a person's death, records like that come back in the synthetic data, so your software meets them during testing.
Data that loads straight into your system
Applications run on multiple linked tables. The data must match your exact data structures, keep referential integrity and follow the business rules in the data. Fields the system expects, such as names and citizen service numbers (BSN), have to be filled too.
Relational generation with rule-based fields
BlueGen generates relational datasets with data structures and referential integrity intact, and each record stays consistent across every table it appears in. Identifiers and other PII carry no statistical meaning, so they are generated from configurable rules alongside the model, in the format your application expects.
Your scenarios, at volume, on demand
Specific ranges and scenarios, including ones missing from the real data, and enough rows for a performance test. Fresh data should be quick to produce and easy to regenerate for every test run.
Conditioning & scaling
With conditioning, you set part of a record and the model fills in the rest, so you can create scenarios beyond what your real data contains. A source of 100,000 rows can become millions with the same statistical properties, as many as your performance test needs.
One million virtual individuals in CBS acceptance testing
Statistics Netherlands allows no microdata on development or test environments. For its new Demographic Register application, BlueGen generated a synthetic population from a model trained on the full register, and it went into acceptance testing in December 2025. In January 2026 it caught the first real software defects in the application, before they reached production.
See what synthetic test data can do for you
- Bring a system you are developing or testing
- See how BlueGen would recreate its data, outliers included