Most of the surveys I’ve designed, analyzed, and even taken have included a check-all-that-apply question. Today’s post includes three ideas for visualizing those survey questions and a bunch of transparency about my own mistakes along the way.


The Survey Questions

Last month I spoke with Harvard University graduate students about visualizing survey results. This is one small portion of their survey.

Here are the check-all-that-apply survey questions that I wanted to graph.


My First Attempt

I started by visualizing the first question.

  • I’m pretending that 30 people completed the survey. I focus on raw numbers of people when I’m dealing with smaller numbers (less than 100) and I convert numbers into percentages when I’m dealing with larger numbers (greater than 100). You’ll notice that the scale along the bottom intentionally goes from 0 to 30 people rather than from 0 to 100 percent.
  • Once in a while, when I sense that the audience is really hungry for details, I’ll also include percentages alongside the raw numbers. In Excel, I simply concatenated the values from a few cells together.
  • I built stacked bar charts instead of regular bar charts because I wanted to remind viewers that while some people checked the survey boxes (dark blue) others did not (light blue). I only labeled the dark blue bars because that’s the segment that really matters.
  • Rather than following the order of questions from the survey (super boring), I listed the items from greatest to least (more useful).
  • I outlined the dark blue and light blue segments in white to provide just the right amount of distinction between categories. White outlines also make printing look a hair crisper.
  • I deleted all the garbage (borders, grid lines, and so on).

I flew through the visualization of the first question and bolted just as quickly into the second question. Within minutes, I had built this one-pager that the researchers could pass around at a meeting.


Oops! My Second Attempt

Then, at the very end, I took a moment to actually read the survey questions. You know, that thing you should probably do first. The categories looked familiar. Too familiar. Eerily familiar. I wondered, Didn’t I already read about ‘industrial and Naval history?’ I think ‘studio art and graphic design’ sounds familiar, too… 

In my zest for building the graphs, I had forgotten to think strategically about the page layout. Would viewers want to read about the survey respondents’ prior academic experience first, and then read about respondents’ prior work experience second? Probably not. Viewers could care less about the order that you asked questions on the survey. There’s always a better way to order your results than by the order in which you asked the questions on the survey. Viewers were probably trying to design museum exhibits by topic–were visitors already familiar with American history? How? Through their work experience or academic experience? In this second draft, I rearranged the page by topic instead of by survey question.


Oops! My Third Attempt

Then, I took another moment to reflect, and was disgusted by the stacked bar charts I had designed. Clustered bar charts are my least favorite chart of all time because 1) there’s (almost) always a more effective alternative but 2) despite these alternatives they’re used over and over and over and over and over.

In this scenario, dot plots are the better choice because they align the academic experience and work experience numbers on the same plane. If I want viewers to compare academic experience and work experience, then it’s my responsibility to place those numbers as close as physically possible to each other. They also take up less ink and space on the page. It’s cheaper to print little dots than big rectangles. You save viewers’ energy and trees; everybody wins.

If you want to explore how I designed the stacked bar charts, the clustered bar charts, the dot plots, or the handouts, you can purchase the templates here:

Purchase the templates