How to choose a translation agency: 6 key questions to ask

Choosing a translation agency in a business context is not about “buying words”. In multilingual projects, quality depends as much on language skills as on organisation: project management, terminology consistency, quality checks, security and a defined strategy for integrating (or not) AI-based translation.

This guide is intended for marketing, product, legal, regulatory, localisation and procurement teams who need to select a language service provider that can deliver at scale without compromising on standards. Here is a short checklist: 6 key questions (plus 1 bonus question) for experienced professionals.

Why a B2B checklist changes everything

The right checklist should not be limited to evaluating “language quality”. It should highlight a true alignment, one that ensures the agency fully meets your exact needs.

A multilingual project is a production chain: briefing, file preparation, terminology, translation, revision, QA, internal validations, then delivery (often in specific formats). The more languages and people involved, the greater the risks: inconsistencies, delays, repeated reviews and hidden costs.

How do agencies ensure quality? How do they measure it? How do they coordinate your projects and data?

When establishing your checklist, you can rely on resources and standards such as ISO 17100: requirements for translation services.

The 6 key questions when selecting a translation agency

1. What are your fields of specialisation?

Your first filter: the actual level of sector-specific experience. An agency that translates “everything” may be suitable for simple content, but when dealing with technical, legal, regulatory, medical or financial content, sector-specific expertise is essential to mitigate risks, including accurate terminology, compliance and traceability.

You can (and should) start by asking the key question: “Which companies in the sector have you worked with in the past?”.

Then ask for details and evidence by inquiring about:

• The type of content
• The volume
• The languages

Other aspects to explore include glossary creation, style guide and the management of regulatory information monitoring.

Why it matters: genuine sector-specific expertise minimises terminology and compliance errors, which can be expensive (and sometimes very expensive) to correct.

2. Are the translators native speakers of the target language? How are they selected?

In a B2B context, the question is not simply whether translators are native speakers or not. The key question is how the agency ensures accuracy, consistency and continuity. We therefore recommend asking two essential questions:

• Do translators translate exclusively into their native language?
• How are they tested, and who validates their skills in terms of subject matter, style and accuracy?

Do not settle for a general answer: ask about the process (tests, onboarding, evaluations, revalidation, replacement management) and the degree of oversight (reviewers, lead linguists, terminology managers).

Why it matters: because quality depends as much on the linguists’ expertise as on the agency’s ability to secure consistent, well-chosen teams.

3. How do you manage terminology and language consistency?

This is often the top concern in multilingual programmes: a single product, concept and message... but variations can occur from one batch to another, or even from one language to another. A good agency ensures terminology management, including glossary creation, validation, dissemination and maintenance, as well as a clean and usable translation memory.

Questions to ask:

• Who proposes the terms and who validates them? (you, the agency, or co-validation)
• How are decisions tracked, how does the agency avoid inconsistencies between countries/variants? (e.g., fr-FR vs fr-CA)
• How are the memory and glossary returned to you if you switch provider?

Why it matters: because terminology consistency prevents endless back-and-forth, repeated corrections and the risk of your message losing impact across markets.

4. How do you protect the confidentiality of documents?

In a business context, translation often involves sensitive information, such as strategy, contracts, pricing, technical data and sometimes personal data). You must therefore evaluate the agency as a link in your subcontracting chain, considering access, storage, retention and contractual rules.

Ask for factual answers:

• Where are files stored, for how long, and who has access to them?
• How are rights managed (principle of least privilege, project-based access)?
• What is the agency’s approach to subcontracting and what contractual commitments are in place (NDA, clauses, traceability)?

From a compliance perspective, you can refer to the GDPR (official overview on EUR-Lex) and CNIL: obligations and best practices for subcontractors. For high-level requirements, also check alignment with ISO/IEC 27001: information security management.

Why it matters: because a confidentiality breach or poorly managed subcontracting can create legal, reputational and contractual risks far beyond the translation project.

5. What is your quality assurance (QA) system and how do you make it measurable?

QA (Quality Assurance) goes beyond simple spellchecking. In a B2B setting, it must cover compliance with the brief and style guide, terminology consistency, numbers and units, product references, and, if necessary, format constraints (tags, variables, formatting).

The “measurable” aspect is key. Request a breakdown of errors template (minor/major/critical), an example QA report or quality dashboard and the approach to continuous improvement (how the agency prevents the same errors from recurring). At this level, the existence of documented processes (rather than “implicit” ones) is an excellent indicator of maturity.

Why it matters: because without metrics and continuous improvement, quality depends on people... and becomes unstable as volumes increase.

6. What is your approach to AI translation?

The aim here is not to take a stance “for” or “against”. The goal is to understand when AI is used, its limits and the degree of human control involved. A mature agency explains its approach based on the type of content: creative marketing, legal, technical, regulatory, etc.

Key points to require:

• Transparency (the agency clearly states whether machine translation is used)
• The pricing applied
• Eligibility criteria (which content goes through this workflow and which is excluded)
• Security (what guarantees are in place and if a third-party tool is involved)

The ISO 18587 standard on human post-editing of machine translation can serve as a reference for structuring post-editing processes. For more information, you can read our article on AI translation rates.

Why it matters: because poorly managed AI usage can harm brand integrity, introduce meaning errors and even pose confidentiality risks depending on the tools used.

Bonus question for experienced professionals: what CAT tools do you use?

This question is not essential for an initial selection, but it provides valuable insight into the provider’s capacity to ensure consistency at scale. CAT (Computer-Assisted Translation) tools facilitate the management of translation memories, terminology, automated quality checks, and, in some case, integrated workflows (TMS, CMS connectors, QA).

If you have some knowledge in the field of language management, you have probably heard of CAT tools. If not, you can consult our guide to quickly understand what CAT tools are.

What matters is not necessarily the “brand” of the tool —though using leading tools like Trados or MemoQ is recommended— but rather how the tool is used and integrated.

Why it matters: because well-configured tools turn translation into a scalable process, enhancing reliability, speed and predictability.

Conclusion: choosing a translation agency is choosing a system

To select a B2B translation agency that can handle multilingual projects, you need to evaluate the organisation as much as the language: specialisation, linguist selection, terminology consistency, confidentiality, measurable QA and a defined AI strategy.

These 6 questions (plus 1 bonus question) provide a solid foundation for comparing providers without making mistakes.

If you wish to discuss your volumes, languages, constraints and expected level of quality, you can visit the website of our translation agency.