What is linguistic quality assurance and how to create a process

How many errors does your translation have? No one cares.

For most people, translation errors sound pretty petty. And who can blame them? It seems nerdy. Full of grammar rules. Unimportant. 

It’s here that the challenge of translation quality becomes clear. On the one hand, poor translations = bad user experience, brand erosion, costly/embarrassing mistakes, and potential legal issues. On the other, when localization professionals talk about building a review step to avoid these scenarios, they’re greeted by the rolled eyes of their colleagues.

Note: This article won’t be defining translation quality. What we will be doing is taking a look at common translation challenges so that you can build a linguistic quality program that works for your customers. 

What matters isn’t the language – it’s the user that consumes the language

For people working in localization, it’s obvious why translation quality matters – it’s the key that unlocks great multilingual customer experience. The problem? Most internal stakeholders don’t view it this way. 

The key to getting people to care is refocusing our attention on the primary objective of a linguistic quality program: providing a great customer experience for end users around the globe. 

Expert corner:

Nataly Kelly, VP of Localization at HubSpot has a great way of reframing how we talk about translation quality:

Instead of saying: “Machine translation alone won’t work for this scenario, because translation quality will suffer.” 

We should say: Machine translation alone won’t work for this scenario, or customer experience will suffer.”

Deprioritizing translation quality = deprioritizing customer experience. That’s how you should frame it when talking to your colleagues in customer support, product, marketing, and sales. 

Improving how you communicate with others on key topics will allow you to accomplish your localization goals faster and tackle challenges when they inevitably arise. 

The most common translation challenges 

Some of the common challenges we come across when working with companies new to the localization journey are:

• Establishing shared goals with internal stakeholders. People across the company (content designers, product, and go-to-market teams) don’t share the goal of producing high-quality localized products, so translation quality is an afterthought. 
• Defining quality. There are objective ways to measure the accuracy and fluency of a translation. Many companies struggle with defining language quality, so they don’t know whether translations are good or bad. 
• Including customer feedback in quality evaluations. The number of mistakes made in a translation judged by a linguist is less important than your customer’s experience. Many companies find it difficult to include customer feedback as part of their equality evaluation (unless they are called out on social media). 
• Process. They have no clearly defined processes for collating bugs and feedback from internal stakeholders, customers, and partners to ensure critical issues are fixed by launch. 
• Systems. There are no tools and automations to support languages in maintenance, make continuous improvements, and launch new languages efficiently. 
• And a whole lot more…

Defining and categorizing translation errors

While it may seem obvious what translation errors are, we think there’s room to define what we mean.

Broadly speaking, a translation error is any lack of consistency between the source and target text. Errors can be divided into objective or subjective mistakes:

Subjective errors:

• Literal translations are used to describe a word-for-word translation of the source text. They sound unnatural and don’t convey the intended meaning. 
• Mistranslations are incorrect translations that mislead readers. 
• Undertranslation is giving less information in the translation than is in the source text.
• Overtranslation is giving more information in the translation than is in the source text.
• Omission means dropping a word or words from the source text in the translation.
• Cultural nuance not being taken into account. Great translation requires an extensive understanding of local regulations, complex terminology, and cultural context.
• Style guides not being respected. A glossary and style guide are vital components of the language assets that you will build with your language partners to ensure consistency of niche terminology, style and voice, and incorporate target audience information.

Objective errors:

• Missing translations
• Spelling, grammatical, and syntax errors
• Glossary & terminology inconsistencies 
• Spacing, typos, capitalization, hyphenation, and so on

At Slack, they have three bug categories for translation errors: functional, linguistic, and won’t fix:

• Linguistic: These bugs are in the realm of localization. Most often they are typos, grammatical errors, or misplaced placeholders. But when you’re translating meaning rather than words, linguistic issues also include not respecting style guides and brand voice. 

• Functional: These issues include errors within variables, cut-off sentences, or words or sentences showing up in the wrong language, and these will need to be fixed by the engineering and design teams.

• Won’t fix: This includes issues with placeholders and text rendering in English (or the source language instead of the chosen locale). It’s important to categorize these because you don’t want to waste time and money on reporting a missing translation if it’s just a bug. 

Defining the categories for translation errors and educating stakeholders is important so that when errors occur, the right person can escalate the issue appropriately.

And then what? Building the linguistic quality process, team, and systems

Once you’ve made a commitment to creating a great multilingual customer experience and have clearly categorized translation errors, you can build the processes for error handling, review, and continuous improvement. 

Process

Establish a quality pillar 

A quality pillar is in charge of maintaining quality standards at scale. Develop a scoreboard for each language (ideally with your language partner) so that all quality evaluations take place objectively, with clearly defined rules instead of opinions. 

For new languages, this pillar will run linguistic quality assurance (LQA) processes, intaking errors and feedback from internal stakeholders, partners, and beta testers.

Note: While it’s important to have objective standards by which quality can be measured, we believe this is less important than understanding what quality means to your customers. Localizing surveys into customer languages and asking them about their experience is a great place to start (and the survey design doesn’t need to be complex). 

Address challenge points for new languages

When launching a new language, do a kickoff overview for cross-functional teams on what they need to know about the market. 

Addressing a lot of potential challenge points (e.g., date and time formatting, or plurals and possessives) and providing teams with some initial knowledge of the language will make it easier to address them as work begins.

When you kick off translations, prioritize areas of your product that are difficult to localize. What are the areas where translations must match across platforms? What are the potential challenges? Translating these first means you have more time to ensure that they are translated well.

Track, measure, report, and improve the areas your customers care about

LQA scores are calculated based on the number of mistakes made in a translation, as judged by a linguist.

Traditional LQA programs use a separate vendor to evaluate the translations of a primary translation vendor. If you don’t trust the quality of translations from your language partner, you can hire a vendor for LQA, or work with a partner you trust. 

Mistakes are still important. But here’s what’s more important: your customer. Obsessively tracking user experience isn’t always straightforward and cheap. However, neither is paying for a third-party LQA program. 

User research using surveys and focus groups is the most useful method of improving what is important. Work with your UX team to include the multilingual customer experience in their research. There are also research partners out there that can help with this. Focus groups are even more insightful when it comes to digging into your customers’ experience if budget allows. 

Team

Once companies have some processes in place, they need the right people to continually refine them and make improvements. Team structure varies wildly from company to company, but the localization process typically involves four primary groups of stakeholders. Here’s how each of them can impact quality: 

1. Designers and software developers: Designing and building your product and ensuring it looks and feels how it should to resonate with your global customers. Developers are also responsible for initiating the localization process by adding string identifiers (i.e., keys) to the product code, as well as fixing bugs and taking full care of the back-end of your digital product.

2. Product, project, and localization managers: Educating all stakeholders in the localization process about your product while also identifying requirements, tools, technologies, and the overall strategy.

3. Copywriters and translators: Preparing all the content that needs to be localized and/or participating in transcreation to ensure it’s well adapted to the target market.

4. QA specialists and reviewers: Establishing a rigorous review process through proofreading and testing in order to deliver the agreed translation quality.

Note: You don’t need all the above stakeholders to have an efficient translation process. However, you do need an owner who can keep track of the end-to-end process across the company. Ideally, this person can be responsible for building relationships and educating internal stakeholders. Cross-functional collaboration is needed more than ever not just between localization teams and design teams, but between all internal teams focused on the product life cycle.

Systems and tools

There are two key elements to building a well-localized product:

• The right people 
• The tools to help these people do their best work (it’s especially difficult to manage a growing group of people without integrated tools and data that flows into a single workspace) 

Usually, translation workflows are managed using a patched-together system that is siloed and relies heavily on good old spreadsheets. 

Mario Pluzny, currently the Localization Program Manager at Twitter, shares the biggest challenge around ensuring quality without the right tools and processes: 

“The quality of global content cannot be assured without everyone looking at the same source. I think that versioning is one of the biggest challenges we face in localization. Even once things are in translation, the English source will often still change. The use of a TMS will allow you to make changes more easily.”

Here’s what the review workflow looks like when using a TMS like Lokalise:

1. Create custom statuses to identify multiple review steps (e.g., copywriting review, subject matter expert review, regulatory review, internal quality checks, etc.)
2. Add internal and external project contributors as team members
3. Enable automated notifications for changes like translations being updated or reviewed, keys added or modified, tasks created or closed, and more
4. Apply filters for new empty keys that require translation, and unverified keys, where the source has been updated and translations require updates
5. Apply appropriate custom statuses to the filtered selection of keys using bulk action
6. Create tasks from the filtered selection of keys
7. Predefine review cycles by creating chained tasks that will trigger once the previous one is completed

The above example is a simplified workflow. It can be optimized further by building a layer of automation on top. The obvious candidates to begin with when automating are repetitive, manual tasks like email/Slack notifications and QA checks. But the possibilities don’t end there.

You can build simple automated processes that are natural, fully integrated parts of your product releases and quality management systems.

We’ve already touched on some of the tools you’ll need, but the key features of a TMS that will help you build and implement a linguistic quality program are: 

• Visual context: Rich visual context through Sketch, Figma, and Adobe XD integrations, plus the possibility to filter translation content by screenshot is another powerful way to support precise, meaningful, and consistent translations.
• Glossary: To allow you to keep terminology consistent across different content buckets.
• Translation memories: To support a coherent tone of voice that is important for strong brand messaging, but also relevant for keeping your translation costs down.
• Automated QA: Built-in quality assurance checks significantly improve the speed of late-stage software testing as they reduce redundant technical errors such as any in placeholders, numbers, and HTML tags, and help translators to focus on consistent translations.
• Over-the-air SDK: Allows you to push product updates post-launch and make improvements on the fly. 

In the past, global brands lacked the tools to launch updates quickly and make continuous improvements. Today, companies use these tools with systems like continuous localization and design-stage localization to continuously launch new languages and improve existing ones. 

For example, Revolut ships localized products in 30+ languages with the ability to quickly make changes when errors occur:

“In the case that some translation is not right, or any other update is necessary, the team can quickly jump on, make the necessary changes, and release the update with the over-the-air SDK.”

Note: This post doesn’t cover design-stage localization. If you want to build a unified workflow that starts at the design stage, get our complete guide.

Build a linguistic quality program for your end users

Ultimately, language quality shouldn’t be treated as another checkbox to complete at the end of a project. More than a review step for translations, you need a “quality mindset” that includes language as a key component of the end-product experience.

Further reading

• How to choose the best translation management system for your team and company
• Define, review, consolidate: translation review best practices
• Translation services: 6 tips for effective collaboration
• How to choose the right translation company (+ a list of 12 reliable companies)
• The complete guide to design-stage localization
• Continuous localization 101: what it is and when it makes sense