Why an import preview should be treated as a proposed change, not just parsed data

A CSV import screen can parse a file correctly yet still leave users unclear about what clicking Import will actually do. The author argues that a useful preview must show which rows match which records, what would change, and which record versions the plan relies on. Rejected and ambiguous rows should be surfaced individually rather than hidden in a generic error count. Because records can change between preview and commit, teams must pick an explicit policy, such as rejecting stale plans, revalidating conflicts, or applying unaffected rows, and enforce it on the server with atomic version checks. Retries after lost responses also require a clear operation-identifier contract, and test cases should cover duplicates, ambiguous matches, and partial application.