Finding a way to write globally

The Problem

How do you improve strings in a language you don’t even speak? Even still, how do you design content for countries you’ve never been to and cultures you’re not a part of? It sounds limiting. And to be fair, it is. But there are ways to work within those limitations and find paths to measurable improvement and high-quality experiences.

Constraints

• Write for languages other than English, which is all I speak

• Learn key differences in tone

• Stay aligned with brand voice and tone

• Adhere to UX best practices

• Understand space limitations – the expansion rate for many languages when translated from English is 15–30%

• Consider UX research from targeted areas, including literacy rates and general preferences

• Research regional knowledge if needed

Area

Google Search

Users

Global audiences

The work

I’ve written for localization and internationalization in a few different capacities. To kick off my UX writing career, I focused on writing clear descriptions for translators. Later, the strings I wrote for text experiments were tested in more than 50 languages and their performance was analyzed for each. As a content designer for Search’s AI, I wrote ideal responses in English to guide the final content design in other languages, which were used to inform instructions for writing that trained the large language model.

Decision-making

In writing for global audiences, I start by following certain rules that set the work up to be optimally translatable and most likely to be well received by users. These rules tend to align with UX best practices such as being clear and concise and avoiding jargon, but may also include things like making sure not to use gerunds (nouns formed from verbs with an -ing ending that are commonly difficult to translate) and being hyper-aware of colloquialisms.

Designing content in one language for another language requires an extra heaping of empathy and a healthy curiosity. Asking the question “Is the general population highly literate, or do most not read?” will lead to an answer about how scannable it should be, how short the sentences should be, how many sentences each paragraph should have, how simple the words should be, and so on.

Metrics

There are different ways to measure success for all of these examples. For a good message description, it’s translation quality. For a successful content design, we’d consider how much positive user feedback it receives in UXR and at launch. For string improvement in experimentation, we can look at some cold, hard numbers.

Testing candidates I wrote for the “People also ask” Search feature resulted in an increase of 583,000 daily user tasks completed in single sessions and affected 42 languages all from testing strings like these:

Carefully considering the subtle and not-so-subtle differences among my candidates led to quantifiable reasons to launch אנשים אחרים רוצים לדעת (Other people want to know) in Hebrew, Outras perguntas (Other questions) in Brazilian Portuguese, and 관련 질문 (Related questions) in Korean.