The Slippery Slope of Slider Scales
Every once in a while we opt for slider scales on surveys we have designed – usually when we want a numeric rating from 0 to 10. It seems easier for respondents than shifting over to the number pad, and it offers a more fun, interactive experience than simply clicking radio buttons. Sliders also allow for finer gradations of numbers; for example, we can design sliders that allow hundreds or even thousands of different points along a continuum.
But every time we use them, I am reminded of the many difficult problems they present. They can be cool and powerful, but they must be used with caution. Here are three big difficulties with slider scales that we and other researchers have experienced and documented:
1. In our most recent effort, for which we were using a new mobile programming platform, we discovered that the slider functionality conflicted with the swipe functionality of an android phone. So back to radio buttons we went. The technology behind sliders is especially important when programming surveys for multiple devices (smartphone, tablets, and desktops), and it requires thorough testing on each device, each potential operating system, and so on.
2. Studies show that it takes respondents twice as long to answer with slider scales than with radio buttons or by entering numbers. Some consider this good, as it may indicate deeper engagement and more thought. But even if true, the evidence also suggests that data from sliders are no more valid than other options.
3. Anchoring the slider has profound effects on the data. The slider has to “start” somewhere, and data may distribute around that starting point in odd ways. Here is a graph offered by Jon Puleson from GMI, a panel sample provider, comparing three methods of eliciting ratings on a 1 to 9 scale:
The lesson from this chart? Anchor your slider at zero; never anchor it in the middle.
Personally, I like sliders and other modes of interactivity that get respondents into less routine states of mind: Card sorts, big buttons, pretty colors – all of these make surveys more interesting. And it matters because, unfortunately, most surveys are shamefully tedious and boring, which almost certainly causes shamefully low-quality data as well.
But all of these elements need to fit within the context of what you are trying to accomplish with a survey. It is important that the folks designing surveys have a good deal of experience working with multiple modes and techniques, and detailed knowledge of the research literature as well. Questions? Feel free to give us a call at 312-348-6089. We are always eager to help.
