Top Ten Survey Design Tips

  1. Define your population.  In a perfect world, who would you like to respond to your survey?  Does the group you're sampling (sending the survey to) really represent them?  Remember - your sample affects the conclusions you can draw.

  2. Prepare written objectives for the research.  State clearly what you want to measure.  Have your objectives reviewed by others.  Review the literature related to the objectives. Can you test your objectives with the results from your survey?

  3. Give the respondents options.  If it's plausible some respondents may not know an answer or have an opinion, give them those options.  Otherwise, the legitimacy of your results and analysis may be called into question.

  4. Ask personal information at the end.  Some people are uncomfortable being asked for their name, age, income range, etc.  Wait until they've invested time in filling out the survey and know your survey is legit before asking these questions.

  5. Keep it short.  Ever given up on completing a survey, or started checking off responses just to be done with it?  That's "survey fatigue".  Respect your subject’s time.  Don’t ask extraneous questions or for information available from other sources.

  6. Define all terms.  In order to be a precise measurement, a survey question must mean exactly the same thing to every respondent.  Therefore any words that are not absolutely precise should be defined.  See #7.

  7. Use the appropriate question type(s).  Do you knowwhen it's best to use a multiple choice vs. a scale, when you should require respondents to select only one or all that apply, and how these choices will affect your data analysis?

  8. Proofread.  Get at least two other people to proofread your questionnaire for errors, and go over it with them in person so they can share with you any phrases, terms, or questions they don't understand.  Before you pre-test it (#9), get a sense of whether respondents might just be checking off answers to items they don't get vs. intelligently answering your questions.

  9. Pre-test.  After it's been proofed, administer the questionnaire to a small, convenient subset of your population.  Just because a question makes sense to you doesn't mean that someone in your population will understand it the same way.  Did you get contradictory information?  Most importantly, do you see any flaws in the responses of your subset that will prove problematic for your much larger, "real" analysis?

  10. Cover email.  Write a simple, professional cover email or letter.  State who you are, the purpose of your survey (in general terms), how long it is or will take to complete (e.g., 8 questions, 5 minutes) provide contact information (name, title, email address, and perhaps phone number), and assure confidentiality.

If your survey requires personally-identifiable information, contact Bentley’s Institutional Review Board (IRB) to determine whether it needs IRB approval.  If you are asking for personally-identifiable contact information simply to provide an incentive for taking the survey, consider linking to a second, unrelated "survey" at the end that will store this information separately - and anonymously - from your survey responses database.

Excellent references for survey design: