Conducting customer surveys is a great way to get feedback on your products and services and to better understand your customers. However, we’ve all taken surveys that were confusing, frustrating, or obnoxious. This blog post has some practical advice for writing survey questions, and here are some related thoughts of mine.
Think about the layout. A few “warm-up” questions can ease participants into the survey, but no survey should include fluff. If you don’t have a good reason to ask a question, don’t ask it. Fortunately, warm-up questions can help you segment responses. For example, if your first question is how familiar participants are with a topic, you can then analyze subsequent answers with respect to familiarity – i.e., Do users with low familiarity find your product less useful? (The author of aforementioned blog post says you have 20-30 questions worth of attention span – I think that’s an overestimate.)
Provide unambiguous answers. There are a few considerations here. Don’t have overlapping answers. If you ask about the number of children in the household, and the answer choices are 1-2, 2-4, and 4-6, what does someone with two children choose? Similarly, don’t leave anyone out; what does someone with 0 or 7 children choose? Cramming more than one question together creates ambiguity. If you ask participants if their house has children and pets, and yes/no are the answer options, what does someone who has no kids but does have a dog choose? (These examples are deliberately simplistic, but surveys have ambiguous answers so often that it’s worth mentioning.)
Avoid leading questions and biased language. When participants read your questions and answer options, they should not be able to tell how you expect them to respond or how you feel about the subject matter. This one is especially difficult for me and, I imagine, for other survey designers. We have hunches and theories about our customers and we have feelings about our products, competitors, or whatever else we’re asking about. Always have a few other people read your survey before you send it – bonus points if they disagree with your theory. Randomize answer options wherever it makes sense (it does not make sense for sequences – how many children: 3-4, 1-2, 5 or more, 0?). Play devil’s advocate with yourself. Biased questions will give you biased (worthless) data.