Skip the Scroll: How to Use Branched Questions for Mobile Surveys
Visual real estate on your survey is valuable and limited, and especially so for the 20 percent of your respondents who will take your survey on a mobile device. Since screen space is so limited, it’s best to design your survey to avoid scrolling as much as possible.
However, generally, it’s best practice to ask questions based on bipolar constructs (questions that range from negative values to positive) using seven-point response scales with each point labeled, which may require scrolling on some devices. So how do we reconcile these potentially competing recommendations?
The solution is to use branched question types, which involves dividing bipolar questions into two distinct questions. The first question will determine what end of the scale the respondent falls on and the follow-up question determines the position within that end of the scale.
For example, if you are asking your customers whether they are satisfied with a particular product, you can first ask whether the respondent falls in the satisfied or dissatisfied portion of the scale and then you can dig deeper and figure out if they are slightly, moderately, or extremely satisfied or dissatisfied with a follow-up question.
But is making your survey longer by asking two questions instead of one even worse than forcing your respondents to scroll? Research indicates that respondents are usually able to answer both items of a branched question as quickly as they would answer a single question with a seven-point scale. In addition, many respondents find it easier to answer branched questions, meaning your data may be more reliable.
Using this method does, however, add an extra step for you during the analysis phase. After you’re finished collecting data, you must merge both questions into a single continuum to re-create the response scale. In the satisfaction example above, you would simply take the first question (or the “stem”) and the follow-up question (or the “branch”) and create a new variable that ranges from ‘Extremely satisfied’ to ‘Extremely dissatisfied.’