Digital survey tools have given marketers a cost-effective means of learning about their customers and prospects, but the information they gather is meaningless without solid survey design.
Designing a good survey is both art and science. It depends on a shrewd understanding of the language a target audience uses when speaking about a specific product or service category. What’s also necessary is a consideration for how easily survey questions can be misleading by their ambiguity or biased presentation.
Contrary to popular opinion, creating a survey that measures what it is actually supposed to measure is not easy or intuitive, but the results are worth it. Here are some ground rules to keep in mind when designing yours.
Talk to Your Audience
Before drafting a survey, have first-hand conversations with members of the target audience. This will help you understand the language they use to articulate perceptions, attitudes, beliefs and behavior about the category and the designated product or service. Doing so may also uncover some revelations about the category or your product that you can measure in your survey.
Question Order Matters
The sequence of questions should move from general questions about the category (related perceptions and behaviors) to more specific ones about your product or service. Measures of overall opinion about an issue change depending on where they fall in relation to more specific questions about that issue. Be sure to randomize and rotate questions to eliminate order bias.Respondents become fatigued and pay less attention to questions that are lower on a list or that come later in a survey. To combat this, ask your most important questions in the first half of the survey.
Pay Close Attention to Question Wording
When wording survey questions, ensure that it means the same thing for everyone who reads it. This is easier said than done, especially when there is complex jargon associated with a category.
Ask Clear Questions
If a question doesn’t have the same meaning for everyone, the conclusions drawn from responses will be wrong. Consider this question from a survey about the insurance industry:“How strongly do you agree or disagree that, based on your situation, XYZ Insurance is being flexible?”Does the term “flexibility” refer to the insurance company’s policies or actions relating to an insurance claim, a premium or a payment schedule? The wording is not precise enough in defining the context.
Ask One Question at a Time
Another common mistake made is asking two questions in one. “How strongly do you agree or disagree that your insurance claim was handled fairly and accurately?”Should the answer to this question be interpreted as an assessment of fairness, an assessment of accuracy or some meaningless average evaluation of the two?
Ask Questions that Respondents are Able to Answer
Certain questions framed too far in the future or past will force respondents to make estimates, which may be far from a true reflection of reality. For example, “How many vacation days will you take in the next two years?” will receive unclear responses because people don’t have their vacations planned out multiple years in advance. It would be much simpler to ask “How many vacation days did you take in the last 12 months?”
It’s also difficult to answer “How many times did you eat out in the last 12 months?” because people generally don’t track their restaurant spending habits on a yearly basis. Instead, try a shorter time frame, such as “How many times did you eat out in April?” Respondents will be better able to recall and answer those questions accurately.
Pre-test, Pre-test, Pre-test
The best way to avoid these survey hazards is to pre-test your survey face-to-face with some target audience respondents. Have them talk to you out loud as they fill out the survey—this will help you understand where there are issues with comprehension, question sequence, etc.
On the surface, designing a survey may seem easy. In reality, it requires a circumspect process to maximize the value that these cost-effective survey tools offer.