Survey Translation from 3,000 Feet
Posted by Jeffrey Henning on Fri, Feb 19, 2010
In "Survey Translation from 30,000 Feet", I recapped, at a high level, Nancy Porte's presentation to the Technology Services Europe conference in Barcelona. Nancy discussed how the impact of translation must be considered at each stage of the survey process: Study Design, Survey Design, Fielding and Analysis. Now I want to get closer to ground level and focus on translation as it affects survey design.
The biggest mistake you can make is to assume that you simply have to translate the English questionnaire into other languages. This actually combines two mistakes: first, assuming the English questionnaire is a good model; second, assuming the destination translations should be by language. When you focus on the word translation, as we all tend to do, we are making this a language-specific problem. Kathleen Bostick, VP of Global Marketing for Lionbridge, really made this distinction clear to me: you want to approach this as a problem of localization, not of translation.
When you think about the problem as one of localization, you realize that you need different editions of the questionnaire for different markets. Some questions you need alternates to, not translations of. No sense translating the choices of the question "Which of the following hotels did you stay at?" Choice lists often must be tailored for major brands of different markets. Entire sections of the questionnaire may differ depending on structure of the industry in that country.
Some location-specific questionnaires will then need to be translated: you will want a French and English translation of the Canadian edition of the questionnaire; you might want a Spanish translation of the U.S. questionnaire, a translation that may look very different from the Spanish-language Mexican questionnaire.
Write the master questionnaire with translation in mind. Avoid jargon, slang, technical terms, idioms and clichés: rewrite it for readability. Use as few open-ended questions as possible, to save translation costs after fielding. Use location-specific skip patterns. Keep an eye out for use of weights and measures, currency and measures of time. Some rating scales don't translate well to other languages; they become too wordy or too subjective. Some scales exaggerate cultural issues. Use images without text so that they will work in multiple cultures; graphics and icons may be inappropriate or confusing for specific cultures. You may need to reduce questionnaire length to fit a limited budget for translations.
A survey translation is not just a translation of the survey itself. You also need to translate the invitation and reminder emails, introductory text, button captions, validation messages (e.g., "Please select no more than three of the following choices"), end pages and thank-you pages and emails.
Last but not least, don't submit the master questionnaire for translation until it is "final final". Translation is the last part of the survey cycle. Surveys must be finalized before having them translated to avoid confusion, re-work and added cost. I'll never forget the client who sent me changes to a questionnaire that we had submitted for translation. "But you said it was final!" I said. "Yes," she replied, "but I didn't say it was final final."
For other fly-bys of the survey-translation process, see these articles: