Known answers
Each dataset comes with claims, a SQL query and the band its result must fall in. Those are your assertions.
Free datasets for testing
Test data with known answers, free to download.
Every dataset comes with claims a test can assert, and the same download gives the same bytes every time. Larger volumes and other seeds on request.
Realistic test data for Postgres and SQLite
Hand-made rows have no distribution, so the slow query on a skewed table is fast on the fixture. A production snapshot has the distribution and the personal data. Random generators break the joins and change on every run.
Known answers from a fixed seed
Every id, timestamp and attribute is a pure function of the seed and the event’s lineage. The same download is the same bytes, today and after the next refresh. The rows are right-censored like live data. The claims serve as the assertions. One says seven carts in ten are abandoned, with the query that counts them and the band the count must fall in.
Get the test data
Download the SQLite file for a fixed fixture, or the CSV bundle to seed whatever your tests load, and record the manifest line beside the test. For other scales or seeds, leave an email above.
What you get
Same bytes every time
A dataset is named by its seed, scale and start date and carries a fingerprint. Pin the line, not a directory of CSVs.
Edges like live data
The rows are right-censored, with open carts and invoices still in dunning, the edges that date logic must handle.
Any volume on request
Ten customers or ten thousand, with the same distributions and joins. Tell us the size you need.
The datasets
Other use cases