Survey Design Tips
Simple steps for better surveys and better outcomes
February 2, 2018
Here are some practical tips to help you improve your survey design. Their purpose is to raise awareness that the choices you make when designing your survey have consequences on the quality of the answers you get as well as on the quality of reporting.
Taken on their own they seem like details, but they all contribute to making your survey design more effective in gaining insights and supporting decision making.
- Use page names to clarify context
Page names or topics will help you group similar questions together, which makes it easier for respondents to focus while also providing a structure for reporting. It's a Win-Win-Win.
If needed, add a page description or an intro text for detail. For example: "The questions on this page are about [insert topic]. When answering, think back to your experience in the past 3 months/latest visit/etc."
- Make sure each question text stands on its own
Make it clear what the questions is about. Don't ask: "Are you..." (commonly used to avoid having to write "Sex" when asking about gender ...) or "Do you ...". Think how it might look in a Table of Contents.
- Only ask about one thing per question
- Make sure question wording and labels work well together
- Read your questions out loud to yourself (or have your computer read them back to you)
- Avoid long questions
If instructions are needed, use a preamble separate to the question itself (or put the instructions in parentheses)
- Avoid starting several questions with the same text
For example: "On a scale of 1 to 5, where 1 is very poor and 5 is very good, please tell us...". Again, consider if it would be helpful in a Table of Contents.
- Use consistent labels for scales where possible
- Use a consistent number of scale points
- Consider having all scale questions "point the same way"
i.e., negative left, positive right. (unless you are specifically using it as a validation device)
- Use balanced or neutral scales
- Don't ask for for more precision or detail than respondents are able to distinguish
See the section below on "The Magical Number"
- Don't ask respondents questions that they don't have the experience to answer
You can avoid this with question routing logic, or provide N/A options.
- Have an open text question for each topic
If your survey has several pages or topics, consider a "Do you have any additional comments about [Topic Name]?" question at the end of each page or topic. This pre-classifies the answers by topic rather than having just one catch-all question at the end. It also captures sentiments when they are fresh in the respondent's mind. Practical survey design.
- If you ask for comments, spell out what the comments are about
Don't word the question just as "Comments". A text question might not be reported next to the question that precedes it in the survey. (Imagine a Table of Contents of a report of text questions where every entry is "Comments")
- Include demographics to help understand differences between categories
You can put them on their own page with a name like "About You" or "Background Information".
- Selection questions: Have a good reason for allowing more than one choice to be selected
("Tick the main reason why you decide to ..." vs "Tick all the reasons you decided to...")
- If there is a set list of answers, provide a list to choose from
Don't rely on free text.
TEST TEST TEST
- There is no substitute for testing your questions on others
Make sure you get feedback on your questions and wording. It hard to proof read your own copy but easy to spot faults in what others have written.
- Test your reporting while you are designing you survey
With ReportGorilla, you can test your reporting with realistic previews BEFORE you collect any answers. This a great way to ensure that your survey delivers the outputs that you expect. You can change, tweak and re-test your design as many time as you like, FREE, to make sure that your survey effort delivers the best possible result. Read more about using report previews during survey design
- Remember to remove any test responses when you go live
The Magical Number
Based on research published by Miller in 1956 about short term memory, the magical number 7±2 indicates how may items a human can keep in mind or consider at the same time. Later research about "chunking" is also relevant. The basic message is keep the number of items to consider at each level low. Make it easy for the respondent. Read more about the Magical Number
For surveys and reporting, this has implications for the number items in a list, questions per topic, question per page and so on. For example, you may have 48 Counties in England, but they can be divided into a smaller number of Regions. Or, you may intersted in ages of respondents, but each age from 16 to 75 makes for a long list. Age-ranges will be easier to manage and report on.