Translation quality assessment guide 

Quality standards and quality assurance come from the manufacturing and service industries. They're representing necessary evaluation systems to ensure the rollout product is error-free. The point is to meet quality standards in performance, design, reliability, and maintainability.

The software industry adopted the same approach for its products making the QA department one of the most important ones.

Yet, quality control for other products, like content, translation, etc. is often overlooked. Assuming that isn't something that needs additional investments besides the regular ones.

Imagine having only people writing codes and handing that straight away to users. Just to have them find all the bugs. Not a pleasant feeling, isn't it?

Translation works the same. Quality control is needed, as quality translation can make a huge difference when rolling out a product or campaign. It can be your deal breaker, and honestly, you want it to prevail for the better.

Why is translation quality important? 

To put it out as simply as possible: what isn’t measured can’t be managed. You can improve a few things when you're performing translation quality assurance. The translation quality is one of them, but there's also a localization success. This directly means your conversion and revenue will be improved.

Low quality of the transition can produce a domino effect and cause a product rollout to go wrong and damage more than just one campaign. It can get stuck with the brand for a long time and cause market failure and big budget loss.

For example, in 2020, an Australian retail brand Kmart made a Mother's Day campaign. That campaign had a tagline #mamaste - something that was imagined as a salutation for all mothers. The tagline was created as a wordplay from the words mama and namaste. With English-speaking countries, the campaign was an exit. But, the problems began when the campaign reached Spanish and Portuguese-speaking geolocations. What Kmart copywriters didn't know was that mamaste in Spanish is a vulgar slang for “did you perform oral intercourse.” Needless to say, the campaign wasn’t so successful in conversion. Additionally, this vulgar slang stuck to the brand to this day.

This backfire could have been avoided if the brand had a localization specialist. Or an additional professional translator to double-check the wording in the campaign. That’s just one of the many reasons translation quality assurance should be a part of your localization workflow.

What is the quality of the translation? 

Translation quality can be tricky to define, as it depends on a variety of things. Grammar and spelling are just the surface of an iceberg.

Multidimensional quality metrics include an understanding of

• ☞ the industry requester is operating in,

• ☞ brand voice of requester,

• ☞ quality of the original text,

• ☞ the purpose of the translated content, and

• ☞ the channel where it’ll be communicated.

Moreover, who is doing the transition, and how involved are the QA translators in all these sectors can also affect quality standards.

How do you evaluate translation quality, then? 

A professional translator is the first thing you'll need for translation quality assurance. Preferably a native speaker that understands deeply the industry the given company is working within. They should be familiar with the quality stage the company operates within.

On the other hand, it's your responsibility to inform the QA translator on

• ☞ where do you plan to position the text,

• ☞ what's the primary purpose of the text

• ☞ introducing the original text.

So be sure to add details to your request such as

• ❓is it informational documentation that should help a user to go through its onboarding easier,

• ❓ is it a copy for a landing page that should convince a potential user to convert?

This should help the QA translator recognize if any wording or formatting exceeds the required quality standard.

Lastly, don't involve the QA translator in the translation process itself. Although it’s possible that one person is a translator, editor, and QA translator, try not to go that way. The efficiency, speed, and quality of the translation will be improved drastically if those are three separate persons. Remember that one person can have around five effective work hours daily and translate around 2000 words during one day. And this is an optimum for the good days.

When it comes to processes, place the translation quality assurance after the translation and editing are done. But do it before the translation is implemented into the product or desired channel of communication.

If you have in-house employees like localization specialists, assign this process to them. If not, delegate the quality control to other third-party language service providers.

What are the qualities of a good translation? 

To evaluate a translation, you’d need some rules collected in the document that’ll serve as the Book of Quality Standards. The goal is for the evaluator to have a certain reference point for non-linguistic improvements that are important for the brand's tone of voice and text's purpose. Something like a Brand Style Guide that you need for designers, content marketers, and copywriters.

But, what should you put in? What quality needs to be reached?

Quality stands for the general excellence of the level or the standard. Quality assurance ensures that the product reaches the highest standardized level.

It includes:

1. observance of the weak points and

2. suggestion of possible improvements so the product can reach the next level of standards.

Quality standards for translation are a little bit tricky, as Geoffrey S. Koby (KSU), Paul Fields (LinguaTech), Daryl Hague (BYU), Arle Lommel (independent scholar), and Alan Melby (LTAC) say in their paper Defining translation quality.

They stress how you need to differentiate quality in terms of quality management and general usage, which tends toward a transcendent perspective. They agree to disagree about the definition and implications of quality standards and about the easiness of objectivity.

However, the authors highlight five important points when it comes to defining translation quality on any occasion:

1. who is the customer the translation is done for? A translator needs to know who the requester is - in which field and level of quality are they operating, what is the core of their business and how they brand themselves.

2. agreeing on the expectation of what’ll qualify as a translation. Usually, it should include a certain level of accuracy and fluency. Accuracy is a bilingual concept that refers to the correspondence between the original and translated text. At the same time, fluency is a monolingual concept that refers to parts of the text, such as grammar, spelling, and cohesion.

3. level of accuracy and fluency. This is more a concern of whether the translated text should improve flaws of the sourced text or should it be on the same level as the original text regarding quality.

4. understanding the purpose of the text and the target audience. Language service providers should be aware of this information as the level of quality is highly dependent on it.

5. there is no objective standard (yet). This can lead to confusion, as there is no explicit definition, but it can also open a door for improvement. It's part of the field where scholars and academics involved in transition studies could work together with professional translators from various fields to reach an objective book of the standards. Until that happens, make sure to create yours to give your localization workflow a better direction.

How can I improve my translation quality? 

Okay, so we need standards to be able to define what is a high-quality translation. But what are we measuring here? A number of grammar issues? Misspelled words? What should you focus on, and in which order?

Let’s go step by step.

What is the main point of translation? 

A German linguist and translation studies scholar Juliane House defines translation in her book Translation quality assessment as the “ result of a linguistic-textual operation in which a text in one language is re-contextualized in another language.”

If we go by this definition, a translation process includes two integrated processes:

1. a cognitive procedure: happens in the translator’s brain
2. a social, cross-cultural, and cross-linguistic practice: that happens when the reader is consuming the translated text.

The last one is crucial when the sourced and targeted languages come from two totally different cultures. Translating word for word won't reach the goal on these occasions.

The goal is to get the same information, context, and understanding in both languages. Sometimes this means using different phrases or combinations of words than we would in literal translation. Simply because languages function differently in terms of linguistics and grammar.

Machine translation can come quite handy sometimes as a quick solution, but they can't understand cultural and social context that changes grammar rules or linguistic usage per se. So while you can use machine translation output as a starting point when there's an enormous quantity to be translated in a minimum time frame, we highly recommend you still have a translator and editor review that translation before the quality evaluation.

As for the first point, we certainly can't track and measure cognitive procedures in a person's brain casually. At least not yet. However, we certainly can analyze its product - the translation.

Three-Step Model for Assessing Translation Quality 

A remarkably effective method for assessing translation quality

Dennis Brown, a managing director of Pacific International Translation, suggests a three-step model for a practical translation quality assessment that can help you improve your translation. Hence, it comes out completely error-free.

The three-step model consists of the following:

1. Evaluating the translation for accuracy of meaning 

This should be the first step, and the reviewer should be focused solely on meaning - whether every phrase is translated from the source language accurately into the targeted language without any losses or redundancy.

At this point, the focus is only on the quality of the meaning of the content:

• is the used phrase accurate
• is every word well spelled,
• are there any typos, and |
• is the vocabulary consistent?

Brown suggests doing it either in the old fashion manner - dividing source and translated texts into phrases in two tables, printing it, and evaluating it row by row.

Or, use a translation management system and its advantages to get it easier and faster. This tool will save time for everyone in the process, as it automatically divides text into phrases according to the directions in the localization file. So both the translator and reviewer can use a translation editor to translate, compare, review and check the context in one unique table. These tools also have translation memory and glossary features that can help you and the translator to stay consistent in every language.

However you choose to do it, he highlights the importance of checking short chunks only - no more than seven to ten words.

2. Evaluating the translation's quality of expression 

Here, you’re looking if the translation sounds natural, without any possibility for misinterpretation and no awkward reading parts. This is also the step when you’re double-checking if the tone and voice of the text are good.

This is the phase where a localization specialist or a QA translator says: “Mamaste won’t work for an Iberoamerican market. It’s the same word as the one used in rude slang”.

Logically, no academic level of language knowledge can help you realize this. It’s necessary that that person is native and is able to notice all shades of language and cultural differences. That’s why it is necessary to do content localization in every market, not just language translation.

The point of this step is to minimize any possible ambiguity. You don’t want tag lines or even sentences open to misinterpretation. It's a sign of poor translation.

For example, you’d need to know that mamar in Spanish actually means to breastfeed, and no mames would simply mean don’t do the breastfeeding. If you say that in Spain, of course. However, a tag that says no mames in Mexico would produce a backfire… or at least a ton of laughs. In Mexican slang no mames is used in crude slang for don’t kid with me, no way, etc.

Again, translation memory and glossary features in translation management software can help you to remember all good and bad words that should be used or completely avoided.

3. Technical/formal check 

This is the part where you check if the layout and the form of the text are proper and if they fit the placeholders.

This can be the trickiest part of the translation evaluation if you're not equipped to see where and how the user would consume it. So this third step is usually done when the product is already designed before the last approval. This stage can turn out pretty late in the process to change anything significant or pretty costly.

Again, the best translation management tools can help you here with two options. One is in-context editing. This feature enables a preview of every text string in the predicted surrounding. This can help you gain context and understanding of formation and actions. The other ones are Adobe or Figma features . These will allow designers to pull translations into the designing environment and give feedback during the translation process so everyone can act on time.

Improve the quality of the translation 

With the right translation team, the right translation management tools any translation project can be successful in high quality. It doesn't matter if you're a translation agency, a freelance translator, or a SaaS company that wants to localize its product - high-quality translation should be your ultimate goal.