One of the early data visualization best practices I was interested in learning more about was how and where to place my vertical axis labels. It seems like a simple, basic aspect of data visualization design. Most default software programs rotate the vertical axis label 90 degrees, and I never really thought that was too big of a deal. For the most part, axis labels tend to be short, so it’s not like you have to turn your head to read them.
One of the articles that informed my original views was this 2013 post by Naomi Robbins. In it, she explores four options for the vertical axis label (I’m using the same chart style and layout as Naomi, but with different data).
The label in the top-left chart—what Excel calls a “Vertical Title”—is, well, ridiculous. Even with a short title like “Percent” you can’t really read it. Naomi called this “unacceptable” and I agree.
The label in the top-right chart—what Excel calls a “Horizontal Title”—is easier to read, but it takes up a lot of unnecessary space. Here again, Naomi and I are in agreement.
The bottom-left chart is the usual default axis label—Excel calls this the “Rotated Title” (though, honestly, if it’s the default, shouldn’t it just be called the “Title”?). This label is relatively easy to read and is compact within the entire graph space. With longer axis labels, however, this can be more difficult to read—more on this in a moment.
Finally, the bottom-right chart takes the axis label and inserts it as a subtitle to the chart—let’s call this the “Subtitle Title” (terrible name, I know). (You’ll note I’ve also left-aligned both the title and axis label). In this case, Naomi wrote that, “The placement on the left figure [the Rotated Title] more clearly associates the label with the axis while the placement on the right [the Subtitle Title] is easier to read….It’s six of one and half dozen of the other.”
For a long time, I adopted Naomi’s position: The “Rotated Title” works fine, is close to the axis, and is the default, so why mess with it? But lately I’ve been thinking that the Subtitle Title approach is superior for three primary reasons:
- It’s directly linked with the title of the chart, and people tend to read chart titles, as a 2015 paper by Michelle Borkin and her co-authors showed. In that paper, participants reviewed a series of graphs and were most likely to mention and recall the titles of the graphs. As the authors note, “Across all textual elements [titles, labels, paragraphs, legends, axes, source, and other text] the title is among the most important.” Thus, even though the label is further from the axis—though a bit closer when I left-align it with the vertical axis—its location closer to the title makes it more likely people will read it. If the horizontal axis is labeled (or is obvious, such as country names or years), the reader isn’t going to think twice about which axis the Subtitle Title applies. (By the way, hat tip to Jen Christiansen and Robert Kosara for finding the Borkin et al paper for me).
- It’s not that much harder to add this label to the chart. In fact, in some cases (e.g., Excel) it’s one less “object” to add to your chart because you can include it in the same text box as the chart title.
- Finally, axis labels are not always short. Here’s an example from CityLab where you really do have to turn your head to read this axis. Moving it into the subtitle (and getting rid of the script text) certainly makes it easier to read, and perhaps also faster (you’ll note that I also deleted the 18 percentage signs because the axis titles tell you its percentages). (Oh, and I just eyeballed the data here).
I’ll also note that many places use the Subtitle Title approach, or some variation therein, as their default style. The Urban Institute, for example, places the axis label next to the vertical axis (see the style guide) and the Congressional Budget Office places it just above the vertical axis.
By comparison, the Sunlight Foundation rotates the vertical axis (though, I would argue in the wrong direction) to leave room for an explanatory subtitle.
Of course, if the axis title is short, maybe you just need a callout instead, as the Dallas Morning News describes in their style guide (you can find a selection of other data visualization guides in the Style Guides on my website).
In this case, maybe simply putting a percentage sign next to the top y-axis label would suffice.
In sum then, my view on what to do with the vertical axis label has changed. I now prefer to make it a subtitle—a little smaller, maybe even a slightly different color—which better links the axis label and the chart title.
Thoughts? Where do you prefer to put your axis labels? Leave your comments in below or let me know on Twitter.
I always put the axis title to the top, in horizontal text, but I had never thought of using the title box. I usually go for the “Axis title” (your top right example), rotate it to horizontal and move it manually to the top. Thanks for the tip!
Great question, Jon! Let’s break it down into 3 empirical questions:
1. When someone first encounters a graph, how quickly can they determine what the y-axis means?
2. If someone is already looking at a graph (e.g., they’re looking at an outlier), how quickly can they remind themselves what the y-axis means?
3. If someone isn’t highly motivated to read the graph, how likely are they to review the label? In other words, could having a label next to a graph be just enough of a nuisance to dissuade someone from reading the graph at all?
Jon introduced a fifth position to add to my four: associating the axis label with the chart title. I was only considering a horizontal axis label placed above the axis. I’ve seen many cases where chart titles appear below the figure; e.g., with the caption. Jon – where would you place the axis label in this case?
I have no trouble reading the longer label in the CityLab example without turning my head. In that example, I’m delighted to see that Jon removed unnecessary percent signs. I’ve been preaching for years that too many percent signs clutter a figure and make numbers more difficult to read, but very few follow this advice.
Good canvassing of the options.
When identifying y-axis label units, I often do it by putting the unit label (eg %, $, kg) in place of the 0 value. By implication, the other tick-labels on the y-axis show that it’s 0. Less useful if your y axis doesn’t start at 0 of course (or perhaps ‘thousands’).
But, stylistically, that approach never seems to work well on the x-axis – probably because it’s not zero bound.
cheers
Thanks for the post, I thought your analysis was spot on. I like to avoid making my users pivot their heads as much as possible, so their chiropractors won’t be mad at me. How would you recommend placing labels when you have 2 Y axes, left and right?
Vlad
Some people say the vertical alignment of title words leads the eye from bottom to top. Then why are the numbers horizontal, not vertical.
But most important is naming the graph. The title should instantly tell you the most salient message the author hopes the viewer obtains. Dont make the viewer work hard unless that is your intention.
Les Posen
Presentationmagic.com
All good and well, but how do I position the label to “Subtitle Title” position without simply dragging the label there? I want something more precise.
You could plot it as a scatterplot point and position it that way, but that might require additional work to modify the y-axis and gridlines. I’m not sure there is a more precise solution.
Thanks.