Definition: Data seeding is the practice of loading a system with prepared data before tests run, so tests start from a realistic, known state. For large systems, seeding may target only the data a test actually needs, while the rest of the system is represented by much smaller placeholder data.
So what? Building realistic data from scratch can take too long for an agile iteration. Targeted seeding lets testers run meaningful tests quickly while keeping setup and storage costs low.
Examples: Seeding only the partitions a customer query will read with realistic data, while mocking the rest. Bulk-loading pre-built data through a fast replication feature before switching to normal imports.
