Multilingual surveys

Many surveys nowadays should include a multilingual component, if only to better engage with Hispanic audiences. MX8 Labs lets you configure multiple languages on a single survey and then serves each respondent the right version automatically based on their device preferences.

Configuring languages on the survey

Language setup starts on the survey itself. On the survey settings you pick three things: the Survey language (the language you wrote the survey in), the Markets where the survey can be fielded, and the Supported Languages that respondents can receive it in. MX8 Labs generates a translation for every supported language automatically, which you can then review and edit on the Translations tab.

Setting a default language on the respondent source

Each respondent source has its own Default language, configured when you create or edit the source. The dropdown only offers languages you picked as Supported Languages on the survey, so you can't accidentally route respondents to a source in a language the survey doesn't speak.

How a language is chosen for each respondent

When a respondent arrives at the survey, MX8 Labs picks a language for them using the following rules, in order:

1. Only one supported language on the survey. The respondent is served that language.
2. Multiple supported languages, and the respondent's device language matches one of them. The respondent is served their device language. MX8 Labs reads this from the browser's Accept-Language header, which reflects the language the respondent has set on their phone, tablet, or laptop.
3. No match. The respondent falls back to the Default language configured on the respondent source they came in through.

In practice this means that as soon as you add a language to a survey, every respondent whose device is set to that language is automatically shown the translated version, no matter which respondent source they came in through. If Spanish is a supported language, anyone with Spanish set as their device language gets the Spanish survey. The same applies to French in Canada, French in France, German in Europe, and so on.

The default language on the respondent source is therefore a fallback, not the primary signal. It matters most for respondents whose device language doesn't overlap with any of the survey's supported languages — for example, someone arriving on a Portuguese-set browser to a survey that only supports English and French.

Example: Hispanic and non-Hispanic sources in the US

A typical setup at MX8 Labs is to run two live respondent sources, one for Hispanic and the other non-Hispanic:

On the Hispanic source we set the default language to Spanish, and the survey is automatically translated into Spanish:

Because of the rules above, a Hispanic-sourced respondent whose device is set to Spanish gets Spanish, and a Hispanic-sourced respondent whose device is set to English gets English — the device preference wins in both cases, which is the best experience for the respondent. The Spanish default on the source only takes effect for respondents whose device language is neither English nor Spanish, steering them to Spanish rather than the English default on the non-Hispanic source.

Reviewing and editing translations

You can review the translations on the Translations tab and edit them if you're a language expert. If you're not, contact us and we can arrange a review of the survey by one of our partner translation houses. Each language is shown side-by-side with the original text, and the Edit, Download, and Upload controls in the top right of the tab let you make changes inline, export to a spreadsheet, or upload a translated spreadsheet back into the platform.

Working with external translation houses

If you want to hand the survey to an external translation house, you can export all of the translations to a spreadsheet and re-import the translated version when it comes back. Use Download to export the spreadsheet, then Upload to bring the translated file back into the survey.

The first column of the export is a unique identifier for each translatable string. Translation houses should treat it as read-only — do not reorder, rename, or delete the values in that column. MX8 Labs uses those identifiers to match each translated row back to the correct string on re-import, so changes happen in place rather than triggering a re-translation from scratch.

A few practical tips for round-tripping with a translation house:

• Send the export file as-is. The structure of the spreadsheet, including the identifier column, must be preserved.
• Translators should only edit the language columns, never the identifier column or the source-language column.
• Save the file in the same format before returning it.
• The upload only accepts files that came from the matching download — the columns and their order must be unchanged. The dialog will reject files that don't match.
• On re-import, any row whose identifier is missing or unrecognized will be skipped, so it is worth doing a quick diff between the file sent and the file returned before uploading.

This approach lets you work with larger translation vendors without giving them direct access to the platform, and it keeps the translation memory consistent between rounds.

Open-end coding across languages

When a survey is fielded in multiple languages, open-end responses come back in each respondent's own language. MX8 Labs codes them into a single, pooled code frame per question, so you don't end up with a separate set of codes per language.

How it works:

• The code frame is shared across all languages on the question. Adding, renaming, or merging a code applies to every language at once.
• The coder is aware of which language each response was written in, so it doesn't try to force English-language reasoning onto a Spanish or French verbatim.
• The coder is also aware of the reporting language. Code labels are shown in the language the report is being read in, while the underlying code identifiers stay constant across languages.
• Verbatims surfaced as examples in reports are shown in their original language, so you can see exactly what the respondent said. Reports themselves can be read in any of the survey's supported languages.

In practice this means that a tracker run across, for example, US English, Hispanic Spanish, and Canadian French will produce one consistent set of themes per open-end question, with one code frame to maintain and one set of cross-language results to read.