Both are described as ways to save money on translation, and buyers often assume one substitutes for the other. They solve different problems and the savings come from different places.
A translation memory stores sentences you have already paid for
Every segment translated is stored with its source. The next time the same or a similar sentence appears, the memory offers the previous translation. On documentation that is revised rather than rewritten, match rates of fifty to eighty percent are ordinary, and you pay a reduced rate on matches. The saving is financial and it compounds — the memory is worth more every year you keep it.
A glossary stores decisions about words
A glossary is small, hand-curated and opinionated: this term, in this language, is translated this way, and not the three other defensible ways. It saves almost no money directly. What it buys is consistency, which is why it matters most on content that many people translate over a long period.
They fail in different ways
A neglected memory fills with old, wrong translations that are then proposed confidently to every future translator, propagating a mistake indefinitely. A neglected glossary simply gets ignored. The memory therefore needs periodic maintenance in a way the glossary does not — someone has to remove segments that reflect decisions you have since reversed.
Who owns them
Both are your assets, not the vendor’s, and this is worth settling in writing at the start. A memory built over three years is expensive to recreate and its absence makes switching vendors far more costly than it should be. Ask for it in a standard format, ask for it periodically rather than at the end, and check that what arrives is actually populated.
Use both, in that order
Start the glossary before the first project, because the decisions are cheapest to make when no content depends on them yet. Let the memory accumulate on its own. The glossary is a small deliberate effort; the memory is a large automatic one.
Leave a Reply