Open-data wants automated checks that sample CSVs land and silver schemas match in a dev catalog. What test type fits?

Open-data teams that need automated checks that sample CSVs land and silver schemas match in a dev catalog need integration tests. Integration tests exercise the ingest/landing path with sample files and assert resulting silver schemas (and often row constraints) in a non-production catalog—proving connectors, expectations, and transforms work together without touching live public tables. That dress rehearsal is what protects open-data publishes from schema drift.

Unit tests that land sample CSVs and assert silver schemas in a non-production/dev catalog fail by test type because unit tests isolate small logic with mocks rather than exercising the connected ingest-to-silver path in a dev catalog; using integration steps does not make the check a unit test.

Contract tests that land sample CSVs and assert silver schemas in a non-production/dev catalog fail because contract tests verify a producer-consumer interface agreement, not functional landing and silver-schema correctness together in a dev catalog.

Performance tests that land sample CSVs and assert silver schemas in a non-production/dev catalog fail because performance tests measure latency, throughput, or behavior under load, not functional proof that sample CSVs landed and silver schemas match in a dev catalog.

Exam caveat: unit vs contract vs performance vs integration are different test types even when they describe the same landing steps. Sample CSV → silver schema in dev = integration. Do not call a single contract test an integration test. Keep integration fixtures small but representative of real municipal CSV quirks (encoding, nulls, extra columns).

Operational check: in a non-prod/dev catalog, run the integration suite that lands sample CSVs and asserts silver schemas; gate merges on green results; do not relabel that functional schema check as unit, contract, or performance tests for dev-catalog validation.