Why Bad Translation Isn't Always the Translator's Fault

Summarize this post with: 8 min read

A bad translation is often caused by the process around the translator, not the translator. The usual culprits are unclear or badly written source text, missing context, no brief about audience and purpose, no glossary, unrealistic deadlines, text split into fragments, and content that was never designed to be translated. Fix those inputs and the same translator will usually deliver far better work.

Key takeaways

Translators work from what they are given. Ambiguous source text produces ambiguous translations.
Context is the single biggest quality lever: screenshots, page URLs, notes on where a string appears.
A short brief (audience, tone, purpose, literal or adapted) prevents most “wrong style” complaints.
Glossaries and style guides stop terminology drifting between translators and projects.
Websites and apps need to be built for translation: no text in images, no sentences glued together from fragments, room for longer text.
When quality is still poor after all that, then look at the translator or the tool.

What counts as a bad translation?

Before assigning blame, it helps to separate the kinds of problems. A useful framework is MQM (Multidimensional Quality Metrics), an open system for classifying translation errors. Its main categories include:

Accuracy: the meaning is wrong, added or missing.
Fluency (linguistic conventions): grammar, spelling and punctuation errors.
Terminology: the wrong term for your product or industry, or inconsistent terms.
Style: the wrong tone or register for the audience.
Locale conventions: dates, numbers, currency, addresses in the wrong format.
Audience appropriateness: content that is offensive or confusing in the target culture.

Only some of these are purely the translator’s responsibility. Terminology, style and audience problems very often trace back to missing information. Accuracy problems often trace back to the source text.

Reason 1: The source text is unclear

Translators can’t translate what the author didn’t say. Common source problems:

Ambiguity. “We’ll update the plan” could mean a subscription plan or a project plan.
Wordplay and idioms that only work in the source language.
Inconsistent terms. If the English site says “workspace”, “project” and “account” for the same thing, the translation will too, or the translator will guess.
Long, complex sentences with several clauses and pronouns whose reference is unclear.
Errors in the original, which a careful translator will either copy or have to ask about.

What to do instead: edit the source before translation. Plain, consistent, well-structured text is cheaper and faster to translate and produces better results in every language.

Reason 2: No context

This is the most common cause of bad website and app translations. Translators often receive a spreadsheet of strings with no idea where each one appears. Is “Close” a verb on a button or an adjective? Is “Order” a noun in a menu or an instruction? Is “Free” about price or availability? In many languages each option is a different word.

What to do instead: provide screenshots or page links, add notes to ambiguous strings, and where possible let translators work directly on the page. A visual editor that shows each sentence in place removes most context errors.

Reason 3: No brief about audience and purpose

A translator who doesn’t know who the reader is has to guess the tone. Is this for teenagers or for procurement managers? Should it be formal Sie or informal du in German, vous or tu in French? Is it a legal text that must be literal, or marketing copy that should be adapted freely?

Brief every project with:

the audience (who they are, what they already know);
the purpose (inform, sell, instruct, comply);
the tone and formality level;
whether to translate literally or adapt freely;
anything to avoid (competitor names, sensitive topics, specific terms);
the target market, not only the language (Spanish for Spain or for Mexico).

Reason 4: No glossary or style guide

Without a glossary, every translator makes their own terminology choices, and every machine translation run may choose differently too. Your product’s key feature can end up with three names in one language. Without a style guide, date formats, capitalisation and tone vary from page to page.

What to do instead: keep a glossary of key terms, their approved translations and do-not-translate words (brand names, product names). Add a one- or two-page style guide per language. Free public examples such as the Microsoft localization style guides show what to cover. Our guide to building a localization glossary walks through it.

Reason 5: Content not built for translation

Some problems are baked into the website or product before any translator sees it:

Text inside images, which can’t be edited without a designer.
Sentences assembled from fragments in code (“You have” + number + “items in your” + “cart”). Word order and grammar differ between languages, so the pieces can’t be translated correctly on their own.
Plurals handled as “item(s)”, when many languages have several plural forms.
Fixed-width layouts. Translations are often longer than English, and the W3C’s article on text size in translation cites IBM guidance that very short strings can expand by 200–300%. If the button doesn’t fit, the translator may be forced to shorten the text until it no longer makes sense.
Hard-coded formats for dates, currency and numbers.

What to do instead: treat internationalisation as part of development. Whole sentences with placeholders, proper plural handling, flexible layouts and locale-aware formatting.

Reason 6: Unrealistic deadlines and budgets

A professional translator typically needs time to research terms, ask questions and revise. When a large project is due tomorrow, those steps are the first to go. The same happens when the budget only covers translation, not revision by a second person. The international standard for translation services, ISO 17100, treats that second-person revision as a required step, not a luxury.

What to do instead: plan translation into release schedules, prioritise what really needs human translation now, and use machine translation plus review for the rest.

Reason 7: Nobody answers the translator’s questions

Good translators ask questions. If nobody on your side answers, or answers arrive after the deadline, they have to guess. Many “mistakes” are guesses the translator flagged and nobody read.

What to do instead: name one contact person per project, answer queries quickly, and add the answers to the glossary or brief so the same question doesn’t come back.

Reason 8: Last-minute changes to the source

When the original text changes after translation has started, some pages end up translated from an old version. The result looks like a translation error but is really a version-control problem.

What to do instead: freeze the source before translation, or use a tool that detects changed text and sends only the new or edited sentences for translation.

When it really is the translator (or the tool)

Sometimes quality problems persist even with clear source text, context, a brief and a glossary. Signs that the translator or engine is the issue:

meaning errors in simple, unambiguous sentences;
grammar and spelling mistakes a native speaker would not make;
the glossary is ignored;
the same issues come back after feedback.

At that point, ask a second native speaker to review a sample, compare against your MQM categories, and change the translator or add a review step. For machine translation, compare engines on your own content, and see our tips on improving automated translation.

How ConveyThis removes the most common causes

On a website, ConveyThis addresses several of these causes directly:

Context: the visual editor shows every translation on the actual page, so translators and reviewers see exactly where each sentence appears.
Terminology: a glossary (on paid plans) sets approved translations and do-not-translate terms for every language.
Consistency: translation memory reuses approved wording across pages.
Collaboration: team roles let you invite translators and reviewers per language, so the people who know the market check the result.
Changing source text: ConveyThis detects new and changed text on your pages and translates it, so language versions don’t drift out of date.

You can also import existing translations as CSV or TMX files. Read more on the translation quality page and the features page, or see which plan fits your site.

For a checklist of mistakes to avoid on your side of the process, read 8 common translation mistakes.

Frequently asked questions

Why is my translation bad even though I hired a professional? Most often because of missing context, an unclear brief, ambiguous source text or no glossary. Check those inputs first, then review a sample with a second native speaker.

What should I send a translator along with the text? The audience and purpose, the tone, the target market, a glossary, screenshots or links showing where the text appears, and the name of someone who will answer questions.

How can I tell whether a translation is good if I don’t speak the language? Ask a second native speaker to review a sample against clear criteria (accuracy, terminology, style, locale conventions), and check that glossary terms appear as agreed.

Can machine translation be better than a human translator? For speed and volume, machine translation is far ahead. For nuance, brand voice and high-stakes content, human translation or human review is still the safer choice. Most businesses combine the two.

Want your translators to see every sentence in context? Create a ConveyThis account and invite them to the visual editor.