Translation memory is a database that pairs previously translated segments — sentences or phrases — with their source text so identical or similar content can be reused instead of retranslated.
Why Translation Memory Matters
Translation memory matters because it turns every past translation into a reusable asset instead of a one-off cost. Without it, linguists and AI systems retranslate the same product names, legal boilerplate, and UI strings from scratch every time, risking inconsistent terminology across documents. With TM in place, a sentence translated once stays consistent across every later document, campaign, or product update.
Translation memory plays two distinct roles, and they should not be conflated. For professional translation services, TM matches directly reduce price: a 100% in-context match (identical text in identical surrounding context) is free, a 100% basic match is billed at roughly half the standard per-word rate, and fuzzy matches between 50-99% similarity get a scaled discount; anything below a 50% match is charged as a new translation. For AI translation, by contrast, translation memory is guidance, not a discount lever: the AI references TM to keep terminology and style consistent with prior approved translations, but AI translation is always billed by subscription word-count tier — Free, Basic, Pro, Enterprise — regardless of how many segments match the TM.
The savings compound with volume on the professional-services side: companies that build a substantial TM report 30-50% cost reduction on professional translation within six months, purely from segment reuse — a saving specific to that pricing model, not to AI translation.
A Concrete Example
A software company translating monthly release notes into German might repeat a line like “This update includes performance improvements and bug fixes” in nearly every release. The first time, a linguist translates it and the segment is saved to TM. The next ten times it appears, TM recognizes it as a match: under professional services, that segment is billed free (in-context match) or at a reduced rate (basic match); under AI translation, the match doesn’t change the bill, but it does keep the German wording identical release after release.
How Translation Memory Works in Taia
In Taia, translation memory builds itself automatically: every segment translated or corrected in the CAT editor or by a professional linguist is saved to TM, organized by language pair and project. Teams can run separate TMs by client, department, or content type, and TM is available starting on the Pro plan, with TMX import/export for moving memories in from tools like Trados or MemoQ. See Taia’s translation memory product page for the full picture, including how TM works alongside glossary and style guide.