Consistent Terminology Across Translators, Suppliers and Subsidiaries: A Guide for Manufacturers
A manufacturer keeps consistent terminology across translators and suppliers by deciding each important term once, recording it in an approved multilingual glossary that it owns, and making every supplier, subsidiary and machine translation engine apply that glossary on every job. DEMA Solutions builds these glossaries from your existing documentation, has them approved by your experts and applies them to every project.
Whose work suffers when one component has three names?
This guide is written for the people who live with the consequences of drifting terminology: technical documentation managers, technical writers, localization managers and quality managers at manufacturers of machinery, HVAC systems, lifting and handling equipment, packaging lines, automation and rail components. It is also for the Italian or German arm of an international group, where the parent company expects the same terms everywhere and local subsidiaries or distributors re-check every translation before release.
It answers practical questions: why the same component carries three names, what a glossary should contain, who approves it, how to build one from existing manuals, how to make every supplier and machine translation engine use it, and how to tell whether it works.
Why technical terminology drifts
Terminology rarely drifts because someone makes a bad decision. It drifts because many people make reasonable decisions independently, at different times, without a shared reference.
It starts in the source text. Manuals written over many years by different technical writers, engineers and product managers accumulate synonyms. One author writes “control panel”, another “operator panel”, a third “HMI”, and the service team calls it “the display”. Every translator faithfully reproduces the difference, and in some languages adds a further variant of their own.
Several suppliers make parallel choices. The operating manual goes to one supplier, the spare parts catalogue to another, the training material to a freelance translator. Each makes a sound choice for “pressure relief valve”, and each choice is different. None is wrong in isolation; together they produce a documentation set that contradicts itself.
Subsidiaries and distributors adapt the wording. A distributor in Germany finds a term its technicians do not use and changes it locally. The change never travels back to headquarters, so the next edition reintroduces the old term and the distributor corrects it again.
HMI texts and manuals live in different worlds. Interface strings are extracted from firmware by the engineering team, translated in a hurry before a release and never compared with the manual. The operator reads one word on the screen and another in the instructions. Bringing both into one workflow is the purpose of HMI translation that matches the manual, with one glossary feeding screens, alarms and instructions alike.
Product generations rename things. A component is redesigned or a function renamed, and without a record of which term replaced which, old and new names coexist for years.
Nobody notices any of this on a single document. Technicians, distributors and customers notice it when they read the manual, the interface and the spare parts list side by side.
What inconsistent terminology costs
The first cost is review time. Every inconsistency becomes a question from an engineer or a subsidiary and another correction round before release, until review becomes the bottleneck of the whole documentation process.
The second cost is risk. Safety warnings, maintenance instructions and residual risk information only work if the operator recognises the component or the action being described. If the manual says one thing and the machine display says another, the instruction is harder to follow, and with the Machinery Regulation (EU) 2023/1230 applying from 20 January 2027, instructions and safety information in the user’s language are a compliance matter as well as a usability one. Our article on what the Machinery Regulation means for translated instructions explains the documentation side in detail.
The third cost is money, and it hides in the translation memory. When the same sentence appears with two different terms, the memory treats it as two different sentences, so less of each update can be reused and more has to be translated and paid for again. The effect is easy to underestimate, and it is one of the main factors in what technical manual translation really costs. Machine translation adds a fourth cost: without your approved terms, an engine picks a generic translation, and it may pick a different one next week.
What a glossary or termbase actually contains
A glossary is often imagined as a two-column spreadsheet: source term on the left, translation on the right. That is a start, but it is not enough to stop drift, because it does not tell a translator when to use the term, what it means or which alternatives are forbidden. A working termbase, the structured version of a glossary used inside translation tools, records several things for each concept.
- The approved term in each language, and where a market genuinely needs it, an approved local variant with the reason recorded.
- Forbidden or deprecated terms: the synonyms and old names that must not be used, so checks can flag them.
- A definition and context: a short description, an example sentence and, for components, a reference to the drawing or part number, so a translator knows which “bracket” or “guard” is meant.
- Usage notes: whether the term appears on the HMI with a character limit, whether it is a product name that must never be translated, how units and abbreviations are written.
- Status and owner: proposed, approved or deprecated, who approved it and when, so everyone knows which entries can be trusted.
The structure matters more than the software. What makes a glossary useful is that each entry describes one concept, carries its own approval history and can be exported in a standard format such as TBX (TermBase eXchange), the ISO exchange format for terminology, or a clean spreadsheet. That is what lets it move between suppliers and survive a change of tools.
A glossary is also different from a translation memory, and you need both. The glossary governs individual terms; the memory stores whole approved sentences for reuse. The glossary keeps “pressure relief valve” stable, while the memory makes sure the sentence around it is not retranslated each time. Understanding who owns your translation memory matters for exactly the same reason as owning your glossary.
How to build a glossary from your existing documents
Most manufacturers do not need to start from a blank page. Their terminology already exists, scattered across source manuals, previous translations, spare parts databases and interface files. The task is to collect it, resolve the contradictions and get the result approved.
Extraction comes first. A terminologist or experienced technical linguist runs term extraction on your source documents and on the existing translations, and combines it with your parts lists, HMI string files and any term lists engineers already keep. The result is a candidate list, ranked by how often each term appears and how critical it is: component names, functions, safety terms, operating modes, alarms, units and product names.
Next comes conflict resolution. For each candidate, the existing translations are compared. Where one translation dominates and is correct, it becomes the proposal. Where three variants compete, the linguist researches standards, the usage in your markets and your own most recent documents, and proposes one with a short justification. Where translations exist but no aligned memory does, aligning old source and target documents is often the fastest way to recover them, which is the job of translation memory alignment.
Scope is the last decision at this stage. A first glossary does not need every word in your manuals. A few hundred well-chosen concepts covering safety, core components, operating modes and HMI labels deliver most of the benefit. Further entries can be added project by project, as translators query new terms.
Approval: engineers, regulatory and subsidiaries
A glossary is only as authoritative as the people who approve it. Linguists can propose and research; they should not be the final word on what your company calls a component.
Approval works best when it is divided by competence. Engineers or product managers confirm source terms and technical accuracy: is this the right name for this part, and is the deprecated synonym really wrong? Regulatory or quality staff confirm safety terminology and wording that must match standards or legal texts. Subsidiaries, distributors or in-country technicians confirm that the target term is the one their market actually uses.
Presentation decides whether approval takes a week or never happens. Each proposal should arrive with its definition, a context sentence, the documents it appears in and the competing variants, so reviewers decide rather than research. Keep batches small, with a deadline and a rule for what happens if no answer comes back.
Applying the glossary in translation and in machine translation
A glossary sent to a supplier as an email attachment is usually forgotten by the second project. To keep terminology consistent, it has to live inside the workflow where translators work, so that approved terms appear automatically as each segment is translated, and automated checks flag any target text that uses a forbidden term or misses an approved one.
Those checks need a human decision behind them. A flagged term is not always an error: the source may use a word in a different sense, or grammar may require a different form. That is why the revision step matters. In an ISO 17100 process, a second qualified linguist revises every translation against the source, and terminology is one of the things they check. Our guide on how to check a supplier’s quality before you hire them lists the questions that reveal whether a supplier actually works this way.
Machine translation raises the stakes. A generic engine knows nothing about your products, so it chooses whatever term is statistically common, which is often not your approved one. When machine translation is controlled by your glossary and translation memory, approved terms apply from the first draft, and post-editors spend their time on meaning and clarity instead of repairing terminology. Whether that approach suits a given document type is covered in our guide to post-editing for technical documentation.
Terminology for HMI texts and spare parts
Interfaces and spare parts catalogues deserve special attention because they are where inconsistency is most visible and most costly.
HMI strings are short, stripped of context and constrained by character limits, so a term that fits in a manual may not fit on a button or an alarm line. The glossary should record the approved short form for the interface next to the full form, and the manual should refer to the screen using exactly the words the operator sees. Translators of interface strings need the glossary, the maximum length of each string and ideally screenshots; without them, the result is truncated labels and improvised abbreviations.
Spare parts catalogues are often generated from an ERP or PLM system, with descriptions written by purchasing or logistics staff rather than technical writers. They tend to use their own naming conventions, which then clash with the service manual. Bringing parts descriptions into the same glossary, at least for the components that appear in maintenance instructions, means a technician can match the part in the manual with the part in the catalogue and order the right one. Once the same descriptions feed an online shop and marketplace listings, ecommerce localization depends on that glossary just as much as the service manual does.
Aligning terminology across a group and its subsidiaries
In a group with several production sites, sales subsidiaries or acquired companies, terminology problems multiply. Each entity may have its own glossary, its own supplier and its own habits, and after an acquisition there are often two names for the same product line.
The workable model is one central glossary with local variants recorded inside it, not separate glossaries per country. Headquarters owns the concepts and the source terms. Each subsidiary owns the approval of its language, with the right to propose a market variant when there is a genuine reason, such as a regional term customers use or wording prescribed by a national standard. Once a variant is approved and recorded, it is used everywhere in that market, and subsidiaries stop editing translations on their own.
This only works if subsidiaries can see the glossary, propose changes through an agreed channel and receive feedback. When local reviewers feel their input disappears, they go back to fixing translations by hand.
Maintaining terminology over product generations
A glossary that is not maintained becomes another source of inconsistency. Products change, components are renamed, new product lines arrive, and every change should reach the glossary before it reaches the documents.
The simplest discipline is to link terminology to engineering change. When a component is renamed or a new one is introduced, the change request includes the term: the new source name, whether the old one is deprecated, and which documents are affected. The glossary is updated first, translations follow, and deprecated entries stay in the termbase, marked as such, so automated checks catch any old name that survives in a reused sentence.
Translator queries are the other source of updates: every answer to “what does this term mean?” should become an entry. A short review at each major product release keeps unapproved proposals from piling up.
How to measure terminology consistency
Consistency is easier to manage once it is visible. There is no single score, but several signals tell you whether the glossary is working.
Automated terminology checks on each delivery show how many approved terms were missed or how many forbidden terms appeared, and whether that number falls over time. The number and type of review comments from engineers and subsidiaries are a second signal: when terminology is under control, review feedback shifts from “wrong term” to genuine questions of meaning. Translator queries per project tell you where the glossary still has gaps. Finally, translation memory reuse on updates is an indirect measure, because consistent source and target terminology allows more of each new edition to come from approved text. Asking your supplier for transparent leverage reporting makes that trend visible to procurement as well as to the documentation team.
A step-by-step approach to consistent terminology
Start with an audit. Compare one product family’s manual, HMI texts and spare parts catalogue in two or three languages, and list the concepts that appear under more than one name. That list is both your business case and your first set of candidate terms.
Define scope and owners. Decide which content types and languages come first, who owns the source terms and who approves each target language, and by when.
Extract, propose, approve. Have a technical linguist extract and research candidates, then send each reviewer only the terms they are competent to judge, with the context needed to decide quickly.
Connect it to every job. Every supplier, internal translator and machine translation engine uses the approved glossary inside the workflow, with automated checks and human revision. Your technical writers receive the preferred source terms and forbidden synonyms, so new content stops creating variants.
Maintain and measure. Link the glossary to engineering change, turn translator queries into entries and track the signals described above.
Several suppliers, one glossary
Many manufacturers work with more than one translation supplier, by design or by history. A glossary you own, in a portable format, is what keeps them consistent. Give every supplier the same approved glossary and the same translation memory, require that both are applied inside their tools on every job, and consistency stops depending on who does the work. A new supplier, or a new translator at an existing one, then starts from your terminology rather than their own.
Coordinating several suppliers brings its own challenges, which are covered in our article on how organisations manage translation across multiple vendors and markets. When you are evaluating a new supplier, ask how they apply your glossary on every job, how they handle terms they cannot resolve and whether you receive the termbase back in TBX at the end. The rest of the ten questions for choosing a technical translation company cover the other points that matter.
How DEMA Solutions keeps terminology consistent
Terminology work starts with material you already own: source manuals, earlier translations, parts lists and the term lists your engineers keep in spreadsheets. Key terms are extracted, conflicting variants are resolved, and each proposal comes with a definition and context, so engineers, regulatory staff and subsidiaries can approve it without a long meeting. Our terminology management service page explains each stage.
Once approved, the glossary is applied on every job together with your translation memory, and in sensitive projects, translations reused from that memory are checked again. Every text follows our ISO 17100 certified process with revision by a second qualified linguist. Glossary and memory belong to you and are delivered as TBX and TMX, which matters in a multi-supplier setting because any supplier can load them. Manuals and catalogues are laid out by our in-house DTP team, and our teams in Italy, Serbia and Spain keep the same people on the same product families.
For one of our manufacturing translation clients, an Italian producer of industrial equipment, every machine leaves with manuals, safety information and HMI texts in the destination language. Its approved glossary has kept component and process names identical across manuals, screens and spare parts lists through years of editions, as the industrial manufacturer case study explains.
Where TUBE fits
A generic engine picks whichever term it has seen most often; TUBE picks yours. Your approved glossary and translation memory guide machine translation within an AI-managed process, so the agreed name for a component appears from the first draft. DEMA loads the glossary you already have, or builds a memory from your past translations through alignment, with nothing to install on your side, and the memory stays yours to export.
Frequently asked questions
The words are often used interchangeably. A glossary usually means a list of approved terms and their translations, while a termbase is a structured database used inside translation tools, with definitions, context, forbidden terms, status and approval history for each concept.
Engineers or product managers for source terms and technical accuracy, regulatory or quality staff for safety wording, and in-country subsidiaries, distributors or technicians for usage in each market. Linguists research and propose; your experts decide.
Own one approved glossary and one translation memory, give both to every supplier in a portable format, and require that they are applied inside the translation tools on every job, with automated checks and revision by a second linguist.
Only if it is set up to. Generic engines choose their own terms. When machine translation is controlled by your glossary and translation memory, approved terms apply from the first draft, and post-editing can focus on meaning rather than terminology.
Yes. The manual should refer to the interface using exactly the words the operator sees. Because interface strings often have character limits, the glossary should record an approved short form next to the full term.
Whenever a component is renamed or introduced, ideally as part of engineering change, and through a periodic review at major product releases. Deprecated terms should stay in the termbase, marked as such, so checks can catch them.
One component, one name, in every language
If the same part carries different names in your manuals, interface and spare parts catalogue, that is the place to start. Send us a few of your documents and we will show you where your terminology drifts, and how an approved glossary would bring it back into line.
How can we help?
Tell us what you need and we’ll get back to you with the solutions.