Nobody has ever been excited by a glossary. Terminology management is the least photogenic part of a translation project, it never appears in a pitch deck, and it is almost always the thing that decides whether the finished work reads like one company speaking or like six freelancers who never met. Buyers who understand this get better translations for less money. Buyers who do not tend to pay twice.
The problem is consistency, not vocabulary
Take a manufacturer with a product that has a housing, a casing and a shell, depending on which engineer wrote the document. In English the three words coexist harmlessly because readers infer the meaning. Translate them literally into four languages and you now have twelve terms for one object, spread across manuals, spare parts lists, training material and a support site. Every one of them is defensible on its own. Together they are a maintenance problem that grows every year.
This is why terminology work starts before translation rather than after it. Someone has to decide, in the source language, that the object is called one thing. That decision is usually uncomfortable because it means telling a department that its preferred word lost. It is also the single cheapest quality improvement available to most companies.
What a working glossary actually contains
A term list of two columns is not a glossary, it is a wish. A usable entry has the approved term, the approved equivalents in each target language, a short definition, the context in which it applies, and the forbidden alternatives with a note on why they were rejected. That last field does more work than people expect, because without it every new translator relitigates the same decision.
Keep it small. Two hundred well chosen entries covering product names, safety language, regulated phrases and the words your competitors use differently will outperform a two thousand entry dump that nobody maintains. Terminology management fails from neglect far more often than from insufficient ambition.
Where translation memory fits
The two are frequently confused. A glossary governs individual terms. A translation memory stores previously translated segments so that repeated sentences are reused rather than retranslated. Both live inside the same computer assisted translation tools, and both belong to you rather than to your supplier, a point worth writing into the contract before the first invoice.
The savings from a memory are real but slower to arrive than sales material suggests. The first project funds the database. The second and third start paying it back, particularly in documentation with high internal repetition. Wikipedia's overview of translation memory is a fair technical description of how segment matching works, and it explains why a memory helps a product manual enormously and a marketing brochure hardly at all.
Technical manual translation is where it pays first
If you are choosing where to start, start with the manuals. Technical manual translation combines everything that rewards this discipline: high repetition between versions, terms with safety implications, long document lifespans and a legal requirement in many markets to supply instructions in the local language. A change to one warning label should propagate everywhere in an afternoon, not trigger a rewrite.
Formal quality standards recognise the same priority. ISO 17100, the international standard for translation services, sets out requirements covering process, resources and the handling of project specifications, and the ISO 17100 standard is a reasonable reference point when comparing suppliers who all claim quality without defining it.
Questions worth asking a supplier
Ask who owns the memory and the glossary, and how you would take them elsewhere. Ask whether new terms discovered mid project are added back or simply used. Ask how they handle a term you later decide to change, because the honest answer involves work and the dishonest answer involves the word automatically. PoliLingua's guide to what translation buyers should know about CAT tools covers the same ground from the technology side and is useful preparation before those conversations.
Ask, finally, who reviews the terminology. A linguist can tell you which German word is correct. Only somebody inside your business can tell you which one your customers already use, and that person needs to be named and given the time.
The cost of getting it wrong is rarely a mistranslation
When terminology drifts, the failures that follow do not look like language failures, which is why they are so rarely traced back. A service engineer orders the wrong part because the parts list and the manual disagree. A distributor argues that a warranty clause means something different in their market. A regulator queries an instruction because the wording no longer matches the approved submission. Each of these gets handled by a different department, and none of them files a complaint about translation.
That invisibility is the reason budgets go elsewhere. The savings from consistent terminology show up as problems that never happened, which no dashboard records. The only reliable way to see the effect is to look at query volume: how many questions your translators raise per project, and whether that number falls over time. When it does, the glossary is working.
Starting from nothing
If you have no glossary at all, do not commission a terminology project. Take the last three documents you translated, pull out every term that caused a query, and write those down. That list will be short, specific and immediately useful. Add to it after each project rather than in a planned exercise that never gets scheduled.
Within a year you will have something worth maintaining, built entirely from real problems. Terminology management done this way costs almost nothing and quietly removes a category of error that most companies simply accept as the price of working in several languages.
