If CAT tools are the translator's workshop, translation memories and termbases are the two most important tools on the bench. They sound technical and are often mentioned together, but they serve different purposes, and understanding the distinction clearly is the difference between merely using software and genuinely benefiting from it.
Translation memory: reusing complete segments
A translation memory stores pairs of source and target segments, sentence by sentence, as you work. When similar text appears again, the tool searches the memory and offers your previous translation. Matches are described in practical categories:
- 100 per cent matches mean the new segment is identical to one stored previously;
- fuzzy matches are partially similar and require careful adjustment, which is where careless work can introduce errors;
- repetitions within the same document are handled automatically;
- anything below a useful threshold is translated fresh.
The memory's value grows with every project. A long-term client's documentation, updates and new products become progressively faster to translate, while approved wording is preserved exactly. This is also why consistent quality matters: an error stored in a memory can propagate into many future documents until it is noticed and corrected.
Termbase: governing individual terms
A termbase is a structured glossary. Rather than storing sentences, it records individual terms and their approved translations, often with additional information: definitions, usage context, part of speech, forbidden alternatives and references. While you translate, the tool recognises terms in the source and presents approved renderings automatically.
Termbases are where specialist knowledge accumulates. In life sciences, legal and technical work, a well-maintained client termbase ensures that product names, regulated vocabulary and preferred terms stay consistent across dozens of documents and every linguist on the programme. Entries work best when they include enough context to use the term correctly, and ownership should be clear: the client decides what is approved, while the language team maintains the structure and flags terms that appear ambiguous across documents. A termbase treated as a shared, governed asset becomes considerably more valuable than a simple list of word pairs.
How they work together
The tools are complementary. Imagine translating an updated product manual. The translation memory retrieves sentences from previous editions, already polished and approved. Within any new or changed sentences, the termbase guarantees that individual terms follow the client's decisions. One handles whole segments; the other governs the vocabulary inside them. Used together, they remove the two most common sources of inconsistency in professional work.
Maintaining quality in the assets
Memories and termbases repay care and punish neglect. Practical discipline includes:
- only storing reviewed, approved work, never unchecked machine output;
- keeping client assets separate, so one client's terminology never contaminates another's;
- correcting an error everywhere it appears in the memory, not merely in the current document;
- adding new approved terms deliberately, with enough context to use them correctly;
- retiring outdated entries when products change or terminology evolves;
- aligning everyone on the team, including reviewers, to the same assets.
When the assets disagree
Occasionally the tools offer conflicting advice: a termbase suggests one rendering while a translation memory contains another, perhaps approved in an older project. Such conflicts should never be resolved silently by choosing whichever appears first. The termbase generally represents the client's current decision, but where the discrepancy looks significant, the correct action is to raise it with the project team so the older memory can be corrected.
This protects consistency across the whole programme, because a contradictory memory will continue to surface in future documents. Once a decision is made, the assets should be aligned deliberately, outdated entries amended and, where appropriate, the style guide updated. Treating conflicts as evidence to resolve, rather than obstacles to ignore, is what keeps quality assets trustworthy over time.
The economics, in plain terms
Clients benefit from these systems through faster delivery, lower cost on repeated content and measurably greater consistency, which is why agencies maintain them as standard. Translators benefit too: less mechanical repetition, fewer opportunities for inconsistency, and the ability to present themselves as the organised professionals clients prefer.
Perhaps most importantly in an age of automation, well-maintained memories and termbases are part of what keeps human work valuable. They embody a client's accumulated decisions and approved language, a body of knowledge that no generic engine possesses. Working smarter with these two tools is not an advanced trick. It is simply how professional translation is done.