Translation quality: How to achieve better outcomes while lowering costs

Translations are the golden ticket to lucrative international markets. However, this comes at a price: Global corporations can easily spend hundreds of thousands to millions of dollars a year on translations. 

Our project managers, localization consultants and solution architects analyzed thousands of projects and pinpointed the root of the problem: Translation budgets and processes are often planned based on “gut feeling”. 

The result? Skyrocketing translation costs due to bloated processes. Or poor translation quality because savings are made in the wrong places. 

In this blog post, we’ll explore: 

• How to define professional translation quality
• How to measure translation quality
• How you can improve translation quality IMMEDIATELY

1. What is translation quality?

Good translations convey the content and style of the source text, while respecting cultural sensitivities and considering the intended use of the target text.

Translation quality is closely linked to business success: Poor translations can quickly lead to PR disasters, legal repercussions, or angry customers (find out about the risks of poor translations). 

Quality – is it just a matter of taste?

There are some things about the taste of a pizza that we can agree on: Is the dough crispy or burnt? Is it made with sun-ripened San Marzano tomatoes or cheap tomato sauce? Some things, on the other hand, come down to personal preference: Do you prefer New York-style or original Neapolitan? With olives or mushrooms on top?

And it’s the same for translations. On the one hand, there are objectively measurable criteria for assessing translation quality, such as: Does a text contain spelling mistakes? Have technical terms been translated correctly? Other aspects, however, are subjective. These include the style of a text or the desired degree of cultural adaptation. 

Another important consideration is that different situations require different levels of quality. Staying with our food analogy: We don’t go to Michelin-starred restaurants every day – usually a simple meal that fills us up is enough. 

A translation that is fit for purpose is referred to as “good enough”. Instructions for a coffee machine don’t need to be inspiring, they just need to be comprehensible. Other texts – a sales contract, a brochure for an art exhibition or your own website – will have higher translation quality requirements. 

In short, what translation quality means is highly dependent on what YOU want to achieve with your content. 

In the next section, we’ll look at how you can define translation quality for your company. 

2.    Defining translation quality

Use the following checklist to define what level of translation quality you require for your business purposes: 

Communications context 

• Who is your target audience? (what are their cultural, linguistic and demographic characteristics)
• What is the purpose of your translations? (to inform, persuade, sell, etc.)
• Does the text need to be adapted for the target market?
• Where will the translations be published? (e.g. website, software, advertising campaign)

Brand & image 

• How do you define your brand personality?
• How would you like to address your target audience? (e.g. casually, formally, humorously)
• Are there any guidelines for your corporate language? (specific language or style rules)
• Are there certain topics in your industry that need to be handled sensitively? (e.g. taboo topics, legal restrictions)

Text conventions

• Are there any special formatting or layout requirements? (e.g. length restrictions in software texts)
• Are there any industry regulations that need to be observed? (e.g. when preparing technical documentation)

Collaboration 

• How will the review process work?
• Do translators need to be provided with reference material (e.g. style guidelines, background information)?

3.    How to measure translation quality

Now that everyone involved knows what the quality requirements are – from translators to project managers to development teams and subject-matter experts. 

The question becomes how to measure translation quality consistently. 

This seems like an impossible task when a company translates into dozens of languages and a wide variety of teams on different continents are involved. This is where internationally recognized standards for measuring translation quality come in handy. 

We’ll discuss two options here: human and automated quality assurance. 

ISO 5060: The new gold standard for translation quality 

ISO 5060 offers a comprehensive system for classifying translation errors that can be customized for specific scenarios

Published in 2024, ISO 5060 is the first international standard for conducting objective and systematic evaluations of translations. 

ISO 5060 categorizes translation errors into seven main categories: 

• Terminology
• Accuracy
• Linguistic conventions
• Style
• Locale conventions
• Audience appropriateness
• Design and markup

All error types can be customized and weighted by severity. This allows for a tailored definition of quality for different text types. 

Example 1: A spelling mistake on a low-traffic webpage may be annoying but is unlikely to have serious consequences. However, if the same mistake appears on a 10-meter-high billboard, it’s a different matter. 

Example 2: A factual error in a blog post should be weighted lower than a false statement in a legally binding contract. 

Error categories and list of translation errors in an LQA (Linguistic Quality Assurance) form 

MTQE: Risk management for AI translations

Many companies today translate millions of words a year – a figure that is only set to grow. Without AI tools like Google Translate, it would be impossible to manage such a large volume of text. That’s why there are automatic processes for checking translation quality, also referred to as MTQE – Machine Translation Quality Estimation. 

In MTQE, a machine uses an algorithm to check machine-generated translation output. Translated text sections (usually individual sentences) are given a score that indicates how good or bad a translation is. 

Business use cases 

1. Benchmark for translation quality: Which engine delivers the best translations: DeepL, Google Translate or ChatGPT? Using established metrics such as BLEU, COMET, TER and chrF, a score can be calculated for all MT outputs, which can be used to compare the respective performance based on your texts. 
2. Smart quality assurance: When translating huge volumes of text, high-quality outputs that do not require further human post-editing can be automatically identified. This means your reviewers can focus on correcting low-quality content, which results in enormous cost savings. 
3. Optimization of translation workflows: MTQE allows companies to identify AI-translated content that can be published without further processing. The review process can thus be customized as needed. 

Airbnb, an online marketplace for international homestays, uses MTQE to monitor the quality of user-generated content. The content on Airbnb’s website is first translated by a proprietary MT system and then analyzed using MTQE. Afterwards, translators review content that has been flagged as critical. This ensures that visitors to the Airbnb portal get the best possible user experience, while the company saves a considerable amount on translation costs. 

4.    Translation workflows in the age of AI

Human vs machine

Traditional translation workflows are no longer suitable nor sustainable in today’s world. There are two reasons for this: 

1. The amount of content that international companies need to produce has increased dramatically – and with it, the potential translation costs. 
2. As a result, companies are increasingly seeking to automate the translation process using machine translation. 

When it comes to translating huge volumes of text in a short time frame, human translators are often too expensive, too slow, and sometimes the quality simply varies too much. But machines are not the perfect solution either: They produce inexplicable errors and fail at dealing with content that is stylistically challenging. 

One thing is clear: Humans or machines alone are not the answer here; a symbiosis is what is needed. 

Think of workflows as modules

A translation workflow involves multiple steps that need to take place at a specific time, making it more complicated to execute than one might think.

For a long time, the idea of having multiple people participate in the translation process the “second set of eyes” principle was treated as gospel in the translation industry. 

This multi-step process is known as TEP (short for Translate, Edit, Proof) and has been advocated by industry standards such as ISO 17100. 

However, multiple review cycles do not always mean higher quality. Over-correction can even negatively impact the overall result – as the saying goes, “too many cooks spoil the broth”, not to mention the ever-increasing costs. 

The fact of the matter is that in 2025 there will no longer be a single ideal translation workflow. 

That’s why Milengo develops tailored solutions for our clients based on their business objectives. Below is an overview of the most important concepts from which we create a translation workflow. 

Step 1

1. Human translation: The classic. A throughput of 300 words per hour can be expected here. 
2. Machine translation: When using machine translation, it is important to clarify whether you want the machine to be trained on your texts in advance so it can better imitate your corporate language. You then need to consider whether the translations should be published without being checked or whether they should be edited by a human in a second step (post-editing). 

Treat marketing translations with extra care! Consider whether your content requires cultural or stylistic adaptation (transcreation). It may also be useful to perform international search engine optimization for webpages and blogs. 

Step 2

Post-editing: Have you used machine translation? If so, you’ll need to decide whether the text needs to be edited to a human level of quality (full post-editing) or whether it just needs to be comprehensible, even if it is not stylistically polished (light post-editing). You can find more tips on post-editing processes in our Hands-on Guide. 

Editing: A review of the translation’s content and style. This includes identifying translation errors and improving the text flow. 

Proofreading: The text is checked for formal correctness. This includes spelling, grammar, punctuation and more. Content errors are not corrected, as only the target text is checked. 

Step 3 and/or 4

Technical review: The content of the translation is reviewed by subject-matter experts – for example, by engineers or your legal department. This is important when it comes to sensitive or complex technical topics. 

Layout check: For brochures, print magazines and web pages, a final check of the translation in the finished design eliminates any remaining errors. 

User testing: This involves testing whether a translation actually fulfills its function in the context of localized software, and everything is in the right place. 

Quality control 

Translation quality should be assessed at regular intervals using a standardized method to check whether it meets requirements (see section “How to measure translation quality”). 

Continuous improvement 

Close collaboration between the client and the language service provider is essential in order to improve translation quality in the long run. Milengo customers, for example, can provide feedback on translations directly via the LanguageDesk customer portal. This feedback is then incorporated into all future translations. 

High quality translations at a low price – is it even possible?

Even with AI, good translations can still cost a lot of money. The good news is that companies can make significant savings if they take a smart approach to translation quality. Here is a summary of our most valuable tips: 

• Identify areas where translation quality that is “good enough” is sufficient to avoid unnecessary additional spend. 
• Design integrated and automated translation workflows (for detailed steps, take a look at our Guide to Translation Management). 
• Define quality standards in a style guide and share it with all stakeholders to avoid costly misunderstandings and revisions. 
• Professionalize your translation management processes with a proven translation management system that includes translation memories and multilingual glossaries. 
• Regularly measure translation quality and adapt your processes if necessary. 

Need an all-in-one solution?

You’re in the right place! Milengo is a leading language service provider offering tailored translation services. We use translation management systems with language AI, comply with the ISO 5060 standard, and strictly adhere to our customers’ style guides and terminology: 

For technical and manual documents: TextGuard ensures error-free translations tailored to your company and your target audience – and costs 80% less than traditional translations. It is ideal for clear multilingual documentation that leaves no room for misunderstanding. 

For marketing content and materials: TextFlow is your ticket to effective communication in a multicultural world. It gives you multilingual content that strikes exactly the right note with your customers – and costs 50% less than traditional translation services.