Every language service provider will tell you they deliver high-quality translations. But if you are comparing suppliers, reviewing translated content internally, or checking whether a project has been completed to the standard you expected, that claim is not enough.
You need a practical way to assess translation quality.
That does not always mean building a complex scoring system or asking your team to become translation specialists. In many cases, it means knowing what to look for, understanding which mistakes matter most, and judging the translation against the purpose of the content.
A legal contract, a product manual, a marketing campaign and an internal HR memo should not all be judged in exactly the same way. Accuracy matters in every case, but tone, style, terminology, cultural fit and layout may carry very different weight depending on the project.
So how do you measure translation quality in a way that is fair, useful and realistic?
Translation quality is not just about “correct words”
A good translation is not simply a word-for-word version of the original text. It needs to carry the same meaning, serve the same purpose and work for the target audience.
That sounds simple, but quality can quickly become subjective.
One reviewer may prefer a translation that follows the source text very closely. Another may prefer a version that reads more naturally in the target language, even if the sentence structure changes. A subject-matter expert may focus heavily on terminology, while a marketing manager may care more about tone, rhythm and emotional effect.
This is why translation quality assurance needs structure. Without it, feedback can become a collection of personal preferences rather than a useful assessment of whether the translation is accurate, appropriate and fit for purpose.
Start with the purpose of the translation
Before assessing the quality of a translation, ask what the content is supposed to do.
Is it intended to inform, persuade, instruct, protect, train, sell, support or comply with regulation?
A translation for internal reference may only need to be clear and accurate. A website localisation project needs to sound natural to local users and support conversion. A medical or technical document may need strict terminology control. A marketing campaign may require transcreation rather than literal translation.
This matters because the same change can be a problem in one context and a good decision in another.
For example, a translator adapting a slogan for a local market may need to move away from the literal source text to preserve the original impact. In a technical manual, the same kind of freedom could create confusion or risk.
The first step in translation quality assessment is therefore not asking, “Is this a good translation?” It is asking, “Is this a good translation for this purpose?”
What should you check when assessing translation quality?
Most translation quality assurance processes look at several core areas. The balance between them will vary by project, but these are the main things to consider.
| Quality area | What to check | Why it matters |
|---|---|---|
| Accuracy | Does the translation preserve the meaning of the source? | Prevents mistranslation, confusion and risk |
| Terminology | Are specialist terms translated correctly and consistently? | Essential for technical, legal, medical and financial content |
| Fluency | Does the translation read naturally in the target language? | Improves clarity, readability and user trust |
| Style and tone | Does the language fit the audience, brand and purpose? | Especially important for marketing, e-learning and customer-facing content |
| Consistency | Are repeated terms, phrases and formatting handled consistently? | Important for large projects, recurring work and multi-file content |
| Grammar and spelling | Are there typos, grammar issues or punctuation errors? | Basic quality marker that affects professionalism |
| Cultural fit | Does the content work for the target market? | Helps avoid awkward, confusing or inappropriate messaging |
| Formatting and layout | Are numbers, dates, tags, links, tables and layout preserved correctly? | Prevents publishing errors and practical usability problems |
This kind of checklist is often more useful for non-specialist reviewers than a dense formal model. It gives people a shared language for discussing quality without turning every review into a debate about personal style.
Use error categories, not vague feedback
One of the most useful principles from formal translation QA models is the idea of categorising errors.
Instead of saying “this translation is poor”, a reviewer should identify what kind of issue they have found. Is it a mistranslation? A terminology error? A grammar mistake? An omission? A formatting problem? A tone issue?
This makes feedback far more actionable.
For example:
- “This sentence is wrong” is not very useful.
- “The translation changes the meaning of the original sentence” is better.
- “This is a major accuracy issue because the translated sentence reverses the intended meaning” is much more useful.
Once errors are categorised, they can also be weighted by severity.
A missing comma and a mistranslated legal obligation are not equal. A typo in an internal draft may be minor. A wrong dosage instruction, product specification or contractual term may be critical.
A simple severity scale can help:
| Severity | Meaning | Example |
|---|---|---|
| Minor | Does not seriously affect meaning or usability | Small punctuation issue, slightly awkward wording |
| Major | Affects clarity, professionalism or correct understanding | Incorrect terminology, unclear sentence, inconsistent key phrase |
| Critical | Creates serious risk, wrong meaning or unusable content | Mistranslation of a legal, medical, safety or financial point |
This approach keeps the review focused on risk and impact, not just preference.
Translation QA models: when should you use them?
There are established translation quality assessment frameworks, including MQM, LQA models and industry-specific systems. MQM, for example, is designed to help analyse translation quality through error types and severity levels, and can be applied to human translation, machine translation and AI-generated translation. (themqm.org)
These models can be very useful for large-scale localisation programmes, ongoing vendor assessment, regulated content, machine translation evaluation or projects where you need measurable quality data over time.
However, they can also be too complex for a buyer who simply needs to review a project sensibly.
For most organisations, the best approach is to borrow the logic behind these models without overcomplicating the process:
- Define what quality means for the project.
- Review a representative sample, or the full translation if needed.
- Categorise issues by type.
- Rate issues by severity.
- Separate objective errors from personal preference.
- Decide whether the translation meets the agreed standard.
This gives you a structured quality assessment without forcing every reviewer to become a localisation quality specialist.
Do not ignore the quality of the source text
Translation quality does not depend only on the translator. It also depends on the source material.
If the original document is unclear, inconsistent, outdated or poorly written, those problems may affect the translation. Sometimes the translator can flag and resolve issues. In other cases, they may need to follow the source closely, especially for legal, technical or regulated content.
This is why source quality should be considered during assessment.
Ask:
- Was the original text clear?
- Were key terms defined?
- Was the source consistent?
- Were there ambiguities that needed clarification?
- Did the translator receive reference materials, glossaries or brand guidelines?
- Were there formatting or file issues that increased the risk of error?
If reviewers do not take this into account, they may blame the translation for problems that started in the source.
Human translation, machine translation and AI-assisted translation
Translation quality assessment has become more important as machine translation and AI-assisted workflows have become more common.
Machine translation can be useful for speed, scale and internal understanding, but raw machine output is not the same as professionally reviewed translation. The quality can vary significantly depending on the language pair, subject matter, source quality, terminology and intended use.
For content that will be published, sent to customers, used in legal or technical settings, or relied on for decision-making, human review remains essential.
When assessing machine-translated or AI-assisted content, pay particular attention to:
- false fluency, where the translation sounds natural but changes the meaning
- terminology errors
- omissions or added meaning
- inconsistent handling of repeated phrases
- formatting, tags, numbers and units
- tone that does not match the brand or audience
- culturally awkward phrasing
A polished sentence is not automatically an accurate sentence. This is one of the biggest risks when reviewing AI-generated or machine-translated content.
Who should review a translation?
The best reviewer depends on the project.
A native-speaking subject-matter expert may be well placed to judge terminology and technical accuracy. A professional editor may be better at reviewing fluency, grammar and readability. A marketing specialist may be needed to assess tone, brand voice and cultural impact.
For high-value or sensitive projects, an independent linguistic review can be useful. This means asking a qualified linguist who was not involved in the original translation to assess the work against agreed criteria.
The key word is “independent”. If the reviewer has a reason to overstate or understate the quality of the work, the assessment may not be reliable.
It is also important to brief reviewers properly. Without clear instructions, reviewers often rewrite according to personal taste. That can create unnecessary changes and make the translation look weaker than it is.
A good review brief should explain:
- who the target audience is
- what the translation will be used for
- whether the tone should be formal, neutral, persuasive or technical
- which terminology must be used
- whether the reviewer should focus only on errors or also suggest stylistic improvements
- how feedback should be categorised
A simple translation quality checklist for buyers
If you are not a translation specialist, you can still carry out a useful initial review by asking the right questions.
Meaning and accuracy
Does the translation preserve the meaning of the original?
Are any sentences missing, added or changed?
Are numbers, dates, names, prices, measurements and legal references correct?
Terminology
Are key terms translated consistently?
Has the translator followed your glossary or preferred terminology?
Are industry-specific terms correct for the target market?
Readability
Does the translation sound natural to a native speaker?
Are sentences clear and easy to follow?
Does the text avoid awkward literal phrasing?
Tone and purpose
Does the translation suit the audience?
Does it match your brand voice?
Would the wording feel appropriate in the target culture?
Formatting and usability
Are headings, tables, bullet points, links and tags handled correctly?
Are dates, currencies, units and punctuation localised properly?
Is the file ready for publication or does it need layout review?
Risk
Would any issue create legal, financial, safety, reputational or user-experience risk?
If so, it should be treated as more than a minor language preference.
How to avoid confusing preference with error
This is one of the hardest parts of translation review.
A reviewer may suggest a different word, phrase or sentence structure simply because they prefer it. That does not always mean the original translation was wrong.
Preference becomes an error only when the original choice is inaccurate, unclear, inconsistent, unsuitable for the audience or contrary to agreed terminology or style guidance.
For example:
- “I would say this differently” is a preference.
- “This term does not match our approved glossary” is an error.
- “This sentence sounds less elegant than I expected” may be a style comment.
- “This sentence changes the legal obligation in the source” is a serious accuracy issue.
A structured QA process helps separate these points. That makes feedback fairer for linguists and more useful for clients.
How Asian Absolute approaches translation quality
At Asian Absolute, translation quality is managed through a combination of specialist linguists, project management, terminology control and review.
For many projects, quality is not created at the final proofreading stage. It starts earlier, with the right translator, a clear brief, good reference materials and an understanding of the target audience.
Depending on the project, quality assurance may include:
- selection of linguists with relevant subject-matter expertise
- use of glossaries and preferred terminology
- translation memory for consistency across projects
- editing and proofreading by experienced reviewers
- formatting and file checks
- final review before delivery
- independent assessment for sensitive or high-value content
This is especially important for projects involving technical, financial, medical, legal, marketing or multilingual content, where a small error can have a much larger impact.
Final thoughts
Translation quality is not measured by one simple question. It depends on accuracy, terminology, fluency, consistency, tone, cultural fit and the purpose of the content.
A good translation quality process does not need to be unnecessarily complex. But it does need to be structured.
Define what quality means for the project. Review against clear criteria. Separate objective errors from personal preference. Consider the severity of each issue. And when the content carries risk, involve qualified linguists who understand both the language and the subject matter.
If you need help assessing translation quality, reviewing multilingual content or setting up a reliable translation QA process, Asian Absolute can help you choose the right approach for your project.