Use random generation for exploration
Fresh records are useful when you want to notice assumptions: a layout may work for a short city but fail for a longer one, or a parser may incorrectly treat every postal code as a number. Generate several examples for the countries your application supports. Inspect the actual fields instead of assuming that every country offers identical coverage or the same postal format.
Save the case that exposed the issue
Once a record reveals a useful boundary, export it and place a reviewed, minimal fixture in your test repository. This tool does not expose a seed that recreates a record on demand. Calling the generator again is therefore not a reliable way to reproduce a failure. Your fixture should carry an explicit scenario name, the relevant address values, and the expectation that makes the test meaningful.
- Keep one short address for a normal submission path.
- Keep one long locality or recipient value for layout coverage.
- Keep one country transition for conditional field behavior.
- Create deliberate invalid variations in the test suite instead of mislabelling them as valid samples.
Review exports before importing them
JSON and CSV are different export views and should not be assumed to contain identical columns. Open the downloaded file, confirm the fields your importer requires, and map them by name. Preserve postal codes as strings so leading zeros survive. If a record contains a synthetic name, example email, or other profile fields that your address test does not use, remove those fields from the committed fixture.
Keep the test boundary explicit
Run these records in local development or an isolated staging environment. The .test email addresses do not receive messages, and generated card examples do not provide a payment service. Stub those integrations or use their approved sandbox fixtures. Local saved records are stored in the current browser rather than an account with cloud synchronization; export important cases before clearing site storage or changing devices. A small reviewed collection is easier to maintain than a large unexplained dump.