A translation memory (TM) is a database that stores previously translated source and target segments as paired units, allowing translators and CAT tools to reuse them when identical or similar text appears again. In translation and localization documentation, the TM is central to project preparation, match-rate analysis, and billing, because it influences consistency, turnaround time, and costs while reducing repetitive manual work.
In daily translation and localization work, the TM appears in CAT tool projects, pre-translation analyses, and invoice breakdowns that report fuzzy matches, repetitions, and new words. Project managers use it to estimate savings and enforce consistency across updates, while translators see retrieved matches in the editor for reuse or revision.
A frequent error is assuming every 100% or fuzzy match from a TM is correct for the current context, because stored segments may come from an outdated product version or a different register. Blindly reusing them can propagate errors, damage terminology choices, or create inconsistent UI text, so matches still need human review.
For creative and consumer content the failure mode is flatness, not inaccuracy — and flatness passes an accuracy check. Terms here also carry technical constraints (reading speed, character budgets) that are part of the brief.
Angel Translation locks terms like this into a project terminology base before expert review begins, so the same source term resolves to the same target term across every document in a submission — and stays consistent in the next revision. See the 8-step AI + expert workflow.
Send us the files. Acknowledged in 4 hours, quoted in 24.
Request a Quote →Content Globalization solutions →