An obvious, but often missed, aspect of accessibility for surveys is translation. If the people you are asking don’t speak your survey’s language, or it is a second or third language for them, then you have put up a significant barrier to getting the answers you need.
Delivering surveys in multiple languages has a wide range of challenges, from the mechanics of managing multiple versions of the survey, to the mechanisms to get the right language in front of the right people. SurveyOptic was built as a multilingual platform from the very beginning. Bolting language support on to a platform that was designed for monolingual use always ends up delivering a substandard product, that was a lesson we had already observed when we were designing the platform. That meant the core functionality was built in from the start:
- Managing translation requests, with version control.
- Managing language reuse, to reduce translation costs.
- Auto-detecting language, with manual selection override.
- Dynamic layout, to adapt to content changes.
Those features, and many more that have been added since, have served us well. However, the world of survey translation is being transformed by AI. Mostly for the better, although not always so. Translating surveys has a range of additional challenges:
- Linguistic Equivalence – Idioms and colloquialisms often don’t translate directly. Even at the word level, a word that’s neutral in one language may carry stronger emotional weight in another. Questions that are clear in one language can become vague or open to multiple interpretations in others.
- Response Scale Issues – Likert scale labels (e.g., “somewhat agree,” “strongly agree”) don’t always have clean equivalents; some languages have fewer (or more) natural gradations.
- Response style bias: some cultures have tendencies toward acquiescence bias (agreeing more readily) or extreme response style, which can distort comparisons even when translation is technically accurate.
- Cultural and Conceptual Equivalence – some constructs simply don’t exist the same way across cultures (e.g., concepts of “privacy,” “happiness,” or “family” can carry different meaning). Questions about sensitive topics (income, health, politics) may need different framing depending on cultural norms.
- Technical/Layout Problems – translated text can be 30-50% longer or shorter than the original, breaking layouts, especially on mobile
- Character issues: non-Latin scripts, right-to-left languages (Arabic, Hebrew), or special characters can break rendering if the platform isn’t properly internationalised. Not all fonts render every language’s characters well.
There are also process issues. Managing translations at the same time as managing the survey development and revision process adds an extra layer of complexity. There are process pitfalls even within the translation steps, for example managing context and implementing the input from multiple translators – a single translator or a single review often misses nuance. Back-translation (translating back to the source language to check accuracy) can help, but that quality check can still miss cultural fit. The survey should be tested in each of the translated languages.
With an increased focus on inclusive design, survey creators are making more frequent, and more extensive, use of translation. To support that we have added a number of new features to SurveyOptic:
- The base language set has been extended to cover over 60 languages.
- Localisation support has been enhanced – automatically handling things like dates and numbers.
- Per question translation overrides – enabling both translation reuse and support for verb-echo languages (like Welsh, that don’t have direct equivalents of ‘yes’ and ‘no’, but instead words that mean ‘I am’ or ‘I did not’).
- In-app translation editor, as well as translation file export and import.
- Full translation for the registered respondent portal and registration flows.
Rather than focusing on geographic coverage (we already have all of the official EU languages), we focused on community reach and on legislated language responsibilities. We have added many community specific languages, to support our government and health care customers in including hard to reach communities. We have also wrapped in the lessons from over a decade of building and running surveys in Welsh (Cymraeg). Those features support running surveys across the Celtic languages, and also have benefits for survey design in general.
In the complex world of multilingual surveys there is always more to do. The passion our team and customers have for language, culture and inclusion is keeping us in a market leading position. We have more features on the way, so talk to the team if you would like an early preview or pointers to the documentation.