Let’s look at an example where a clear visual hierarchy has been established and discuss the specific design choices that were made to create it. Imagine you are a car manufacturer. Two important dimensions by which you judge the success of a particular make and model are (1) customer satisfaction and (2) frequency of car issues. A scatterplot could be useful to visualize how the current year’s models compare with the previous year’s average along these two dimensions, as shown in Figure 5.6.
Figure 5.6 Clear visual hierarchy of information
Figure 5.6 lets us quickly see how this year’s various models compare to last year’s average on the basis of both satisfaction and issues. The size and color of font and data points alert us where to pay attention and in what general order. Let’s consider the visual hierarchy of components and how they help us process the information presented. If I articulate the order in which I take in the information, it looks something like the following:
First, I read the graph title: “Issues vs. Satisfaction by Model.” The bolding of Issues and Satisfaction signals that those words are important, so I have that context in mind as I process the rest of the visual.
Next, I see the y-axis primary label: “Things Gone Wrong.” I note that these fall along a scale from few (at the top) to many (at the bottom). After that, I note the details across the horizontal x-axis: Satisfaction, ranging from low (left) to high (right).
I am then drawn to the dark grey point and corresponding words “Prior Year Average.” The lines drawing this point to the axes allow me quickly to see that the prior year’s average was around 900 issues per 1,000 and 72% satisfied or highly satisfied. This provides a useful construct for interpreting this year’s models.
Finally, I am drawn to all of the red in the bottom right quadrant. The words tell me satisfaction is high, but there are many issues. It’s clear because of how the visual is constructed that these are cases where the level of issues is greater than it was for last year’s average. The red color reinforces that this is a problem.
We previously discussed super-categories for easing interpretation. Here, the quadrant labels “High Satisfaction, Few Issues” and “High Satisfaction, Many Issues” function in this manner. In absence of these, I could spend time processing the axis titles and labels and eventually figure out that’s what these quadrants represent, but it’s a much easier process when the pithy titles are present, eliminating the need for this processing altogether. Note that the left quadrants aren’t labeled; labels are unnecessary since no values fall there.
Additional data points and details are there for context, but they are pushed to the background to reduce the cognitive burden and simplify the visual.
Upon sharing this visual with my husband, his reaction was “that’s not the order I paid attention—I went straight to the red.” That got me to thinking. First, I was surprised he started there, given that he’s red-green colorblind, but he said that the red was different enough from everything else in the visual that it still grabbed his attention. Second, I look at so many graphs that it’s ingrained in me to start with the details: the titles and axis titles to understand what I’m looking at before I get to the data. Others may look more quickly for the “so what.” If we approach that way, we’re drawn first to the bottom right quadrant since the red signals importance and that attention should be paid. After taking that in, perhaps we back up and read some of the other detail of the graph.
In either case, the thoughtful and clear visual hierarchy establishes order for the audience to use to process the information in a complex visual without it feeling, well, complicated. For our audience, by highlighting the important stuff, eliminating distractions, and establishing a visual hierarchy, the data visualizations we create afford understanding.
Accessibility
The concept of accessibility says that designs should be usable by people of diverse abilities. Originally, this consideration was for those with disabilities, but over time the concept has grown more general, which is the way in which I’ll discuss it here. Applied to data visualization, I think of it as design that is usable by people of widely varying technical skills. You might be an engineer, but it shouldn’t take someone with an engineering degree to understand your graph. As the designer, the onus is on you to make your graph accessible.
Poor design: who is at fault?
Well-designed data visualization—like a well-designed object—is easy to interpret and understand. When people have trouble understanding something, such as interpreting a graph, they tend to blame themselves. In most cases, however, this lack of understanding is not the user’s fault; rather, it points to fault in the design. Good design takes planning and thought. Above all else, good design takes into account the needs of the user. This is another reminder to keep your user—your audience—top-of-mind when designing your communications with data.
For an example of accessibility in design, let’s consider the iconic London underground tube map. Harry Beck produced a beautifully simple design in 1933, recognizing that the above-ground geography is unimportant when navigating the lines and removing the constraints it imposed. Compared to previous tube maps, Beck’s accessible design rendered an easy-to-follow visual that became an essential guide to London and a template for transport maps around the world. It is that same map, with some minor modifications, that still serves London today.
We’ll discuss two specific strategies related to accessibility in communicating with data: (1) don’t overcomplicate and (2) text is your friend.
Don’t overcomplicate
“If it’s hard to read, it’s hard to do.” This was the finding of research undertaken by Song and Schwarz at the University of Michigan in 2008. First, they presented two groups of students with instructions for an exercise regimen. Half the students received the instructions written in easy-to-read Arial font; the other half were given instructions in a cursive-like font called Brushstroke. Students were asked how long the exercise routine would take and how likely they were to try it. The finding: the fussier the font, the more difficult the students judged the routine and the less likely they were to undertake it. A second study using a sushi recipe had similar findings.
Translation for data visualization: the more complicated it looks, the more time your audience perceives it will take to understand and the less likely they are to spend time to understand it.
As we’ve discussed, visual affordances can help in this area. Here are some additional tips to keep your visuals and communications from appearing overly complicated:
Make it legible: use a consistent, easy-to-read font (consider both typeface and size).
Keep it clean: make your data visualization approachable by leveraging visual affordances.
Use straightforward language: choose simple language over complex, choose fewer words over more words, define any specialized language with which your audience may not be familiar, and spell out acronyms (at minimum, the first time you use them or in a footnote).
Remove unnecessary complexity: when making a choice between simple and complicated, favor simple.
This is not about oversimplifying, but rather not making things more complicated than they need to be. I once sat through a presentation given by a well-respected PhD. The guy was obviously smart. When he said his first five-syllable word, I found myself impressed with his vocabulary. But as his academic language continued, I started to lose patience. His explanations were unnecessarily complicated. His words were unnecessarily long. It took a lot of energy to pay attention. I found it hard to listen to what he was saying as my annoyance grew.
Beyond annoying our audience by trying to sound smart, we run the risk of making our audience feel dumb. In either case, this is not a good user experience for our audience. Avoid this. If you find it hard to determine whether you are overcomplicating things, seek input or feedback from a friend or colleague.
Text is your friend
Thoughtful use of text helps ensure that your
data visualization is accessible. Text plays a number of roles in communicating with data: use it to label, introduce, explain, reinforce, highlight, recommend, and tell a story.
There are a few types of text that absolutely must be present. Assume that every chart needs a title and every axis needs a title (exceptions to this rule will be extremely rare). The absence of these titles—no matter how clear you think it may be from context—causes your audience to stop and question what they are looking at. Instead, label explicitly so they can use their brainpower to understand the information, rather than spend it trying to figure out how to read the visual.
Don’t assume that two different people looking at the same data visualization will draw the same conclusion. If there is a conclusion you want your audience to reach, state it in words. Leverage preattentive attributes to make those important words stand out.
Action titles on slides
The title bar at the top of your PowerPoint slide is precious real estate: use it wisely! This is the first thing your audience encounters on the page or screen and yet so often it gets used for redundant descriptive titles (for example, “2015 Budget”). Instead use this space for an action title. If you have a recommendation or something you want your audience to know or do, put it here (for example, “Estimated 2015 spending is above budget”). It means your audience won’t miss it and also works to set expectations for what will follow on the rest of the page or screen.
When it comes to words in data visualization, it can sometimes be useful to annotate important or interesting points directly on a graph. You can use annotation to explain nuances in the data, highlight something to pay attention to, or describe relevant external factors. One of my favorite examples of annotation in data visualization is Figure 5.7 by David McCandless, “Peak Break-up Times According to Facebook Status Updates.”
Figure 5.7 Words used wisely
As we follow the annotations from left to right in Figure 5.7, we see a small increase on Valentine’s Day, then large peaks in the weeks of Spring Break (cleverly subtitled “Spring clean?”). There’s a spike on April Fool’s Day. The trend of break-ups on Mondays is highlighted. A gentle rise and fall in break-ups is observed over summer holiday. Then we see a massive increase leading up to the holidays, but a sharp drop-off at Christmas, because clearly breaking up with someone then would simply be “Too Cruel.”
Note how a few choice words and phrases make this data so much more quickly accessible than it otherwise would be.
As a side note, in Figure 5.7, the guidance I previously put forth about always titling the axes has not been followed. In this case, this is by design. Of more interest than the specific metric being plotted are the relative peaks and valleys. By not labeling the vertical axis (with either title or labels), you simply can’t get caught up in a debate about it (What is being plotted? How is it being calculated? Do I agree with it?). This was a conscious design choice and won’t be appropriate in most situations but, as we see in this case, can—in rare instances—work well.
In the context of accessibility via text, let’s revisit the ticket example we examined in Chapters 3 and 4. Figure 5.8 shows where we left off after eliminating clutter and drawing attention to where we want our audience to focus via data markers and labels.
Figure 5.8 Let’s revisit the ticket example
Figure 5.8 is a pretty picture, but it doesn’t mean much without words to help us make sense of it. Figure 5.9 resolves this issue, adding the requisite text.
Figure 5.9 Use words to make the graph accessible
In Figure 5.9, we’ve added the words that have to be there: graph title, axis titles, and a footnote with the data source. In Figure 5.10, we take it a step further by adding a call to action and annotation.
Figure 5.10 Add action title and annotation
In Figure 5.10, thoughtful use of text makes the design accessible. It’s clear to the audience what they are looking at as well as what they should pay attention to and why.
Aesthetics
When it comes to communicating with data, is it really necessary to “make it pretty?” The answer is a resounding Yes. People perceive more aesthetic designs as easier to use than less aesthetic designs—whether they actually are or not. Studies have shown that more aesthetic designs are not only perceived as easier to use, but also more readily accepted and used over time, promote creative thinking and problem solving, and foster positive relationships, making people more tolerant of problems with designs.
A great example of the tolerance with problems that good aesthetics can foster is a former bottle design of Method liquid dishwashing soap, pictured in Figure 5.11. The anthropomorphic form rendered the soap an art piece—something to be displayed, not hidden away under the counter. This bottle design was wildly effective in spite of leakage issues. People were willing to overlook the inconvenience of the leaking bottle due to its appealing aesthetics.
Figure 5.11 Method liquid dishwashing soap
In data visualization—and communicating with data in general—spending time to make our designs aesthetically pleasing can mean our audience will have more patience with our visuals, increasing our chance of success for getting our message across.
If you aren’t confident in your ability to create aesthetic design, look for examples of effective data visualization to follow. When you see a graph that looks nice, pause to consider what you like about it. Perhaps save it and build a collection of inspiring visuals. Mimic aspects from effective designs to create your own.
More specifically, let’s discuss a few things to consider when it comes to aesthetic designs of data visualization. We’ve previously covered the main lessons relevant to aesthetics, so I’ll touch on them here only briefly and then we’ll discuss a specific example to see how being mindful of aesthetics can improve our data visualization.
Be smart with color. The use of color should always be an intentional decision; use color sparingly and strategically to highlight the important parts of your visual.
Pay attention to alignment. Organize elements on the page to create clean vertical and horizontal lines to establish a sense of unity and cohesion.
Leverage white space. Preserve margins; don’t stretch your graphics to fill the space, or add things simply because you have extra space.
Thoughtful use of color, alignment, and white space are components of the design that you don’t even notice when they are done well. But you notice when they aren’t: rainbow colors, and lacking alignment and white space, make for a visual that’s simply uncomfortable to look at. It feels disorganized and like no attention was paid to detail. This shows a lack of respect for your data and your audience.
Let’s look at an example: see Figure 5.12. Imagine you work for a prominent U.S. retailer. The graph depicts the breakdown of the U.S. Population and Our Customers by seven customer segments (for example, age ranges).
Figure 5.12 Unaesthetic design
We can leverage the lessons covered to make smarter design choices. Specifically, let’s discuss how we can improve Figure 5.12 when it comes to the use of color, alignment, and white space.
Color is overused. There are too many colors, and they compete for our attention, making it difficult to focus on one at a time. Going back to the lesson on affordances, we should think about what we want to highlight to our audience and only use color there. In this case, the red box around segments 3 through 5 on the right signals that those segments are important, but there are so many things competing for our attention that it takes some time to even see that. We can make this a more obvious and easier process by using color strategically.
Elements are not properly aligned. The center alignment of the graph title makes it so it isn’t aligned with anything else in the visual. The segment titles at the left aren’t aligned to create a clean line either on the left or right. This looks sloppy.
Finally, white space is misused. There is too much of it between the segment titles and data, which makes it challen
ging to draw your eye from the segment title to the data (I have an urge to use my index finger to trace across: we can reduce white space between the titles and data, so this work is unnecessary). The white space between the columns of data is too narrow to optimally emphasize the data and cluttered with unneeded dotted lines.
Figure 5.13 shows how the same information could look if we remedy these design issues.
Figure 5.13 Aesthetic design
Aren’t you more likely to spend a little more time with Figure 5.13? It is clear that attention to detail was paid to the design: it took the designer time to get this result. This creates a sort of onus on the part of the audience to spend time to understand it (this sort of contract doesn’t exist with poor design). Being smart with color, aligning objects, and leveraging white space brings a sense of visual organization to your design. This attention to aesthetics shows a general respect for your work and your audience.
Acceptance
For a design to be effective, it must be accepted by its intended audience. This adage is true whether the design in question is that of a physical object or a data visualization. But what should you do when your audience isn’t accepting of your design?
Storytelling with Data Page 11