How to Shorten a Survey

Actress demonstrating initial reactions of fea...

Image via Wikipedia

I know, I know, you NEED all those questions. You can’t gather the learnings you require without every last one of those questions. Well, here are some tips that I hope will help you shorten your surveys and increase your data quality.
1) Remove or reduce all the questions that are there because they are interesting but never actually do anything about. You know what these are. The 40 item grids that people straightline through anyways. The questions that have so much jargon and marketing speak that you barely understand them. The questions that go into such minute detail that no sane person would be able to answer them. (No peeking, what is the exact fabric content of the shirt you are making right now.)
2) Remove all but two concepts. Seriously, do you think that by the time responders are on the third concept that they’re actually reading it? I doubt it. It’s pretty boring to read through the same series of questions over and over. I probably wouldn’t.
3) Chop the survey in half and deliver each half to a different set of people. Chances are, the survey is written as modules anyways. You can probably separate it out into two or three separate surveys and still achieve your overall goals without the loss of analysis. A good researcher will be able to do this.
4) Take advantage of the demographics gathered through profiling surveys. Most survey panels gather hoards of data from panelists when they join. This information is then updated on a regular basis, including automatic aging of people. If you can save ten minutes of survey time by piping in variables, just imagine what you can put in there instead. Even better, don’t replace those ten minutes.
5) Forget the 7 or 10 point scale and stick to a 5 point scale. Sure, you’ve given responders more options and yourself more data points. BUT, remember that you are going to average those scores across hundreds of people anyways. A 8.4 out of 10 is the same as 4.2 out of 5. You really really do have exactly the same number of gradients in both cases. Save some time, avoid annoyance, and just use 5 point scales.
6) Using branching and hiding skip patterns does NOT shorten a survey. It simply hides length. That is not a solution.
Hm… Now that I think about it, the topic of this chatter is not survey length but data quality. Fancy that.

Related Articles


  • You are too old to understand social media so don’t even try
  • #MRA_FOC #MRX Effective Data Visualization by Naomi Robbins, Part #1
  • CAUTION: The Scariest Thing EVER
  • 1topic5blogs: The only thing cell phone surveys are good fer
  • I don’t give a rat’s ass about probability sampling
  • Building a bad reputation before we even start: Privacy in social media research
  • Advertisements
    %d bloggers like this: