A translation memory (TM) is a database that stores every sentence you've ever translated, paired with its source. When the same sentence — or a similar one — appears in a future project, the TM surfaces the existing translation so the linguist doesn't start from zero.

Most people think of TM as a cost-saving tool. It is — but that's the least interesting thing it does.

What TM actually saves

1. Consistency across releases

Without a TM, every release is a fresh start. The linguist who translates "Cancel subscription" in v2.1 might use a different term than the linguist who translated the same phrase in v1.8. Over 20 releases and 15 languages, those inconsistencies compound into a fragmented user experience.

A TM enforces consistency automatically. When the same string appears again, the linguist sees the approved translation and either reuses it or consciously decides to update it — with a record of why.

2. A paper trail for term changes

When your legal team decides that "Terms of Service" should become "Terms of Use" across the product, a TM gives you a searchable record of every instance of the old term, in every language it was translated into. Without it, you're grepping through spreadsheets.

3. Onboarding speed for new linguists

When a new linguist joins a project, the TM gives them immediate access to every translation decision the previous linguist made. They can see how the team handles ambiguous terms, whether the product uses formal or informal register, and what abbreviations are accepted.

How TM fits into a real workflow

  1. New project: Source strings are run against the TM. Exact matches are auto-populated. Fuzzy matches (70%+ similarity) are surfaced as suggestions.
  2. Linguist review: The linguist accepts, modifies, or rejects each suggestion. New translations are added to the TM.
  3. Next release: Only new or changed strings need translation. The TM handles the rest — with consistency guaranteed.

When TM doesn't help

TM works best for repetitive, structured content: UI strings, help articles, product documentation. It's less useful for creative content like marketing copy, where tone and context change between campaigns. For those, transcreation is the right approach — and the TM stays out of the way.

Want to set up a translation memory for your product? Start a project — we'll build the TM from your existing translations and use it from day one.