Back in November, Ryan Sleeper, the 2013 Tableau Iron Viz Champion, published a blog post here on how to make donut charts in Tableau. The following month, Andy Kriebel, data visualization expert and Tableau Zen Master, posted additional instructions here. Both Ryan and Andy would be the first of many data visualization guys to tell you, "Don't use pie charts and donut charts" in almost any situation. In fact, Andy does exactly that in another blog post from 2012 here. However, in this one particular case, they both argue that the donut chart works well. Their example is a very common use case. It involves a Key Performance Indicator or KPI. More specifically, they are measuring the performance of something against its goal in particular regions (North, South, East, West).
Example:
Yesterday, Steve Wexler, also a Tableau Zen Master, published a great blog post here that cautions users against using these types of charts and explains why in great deal. In a nutshell humans are really good at estimating the length and height of things, but not very good at all estimating angles and arcs. Steve goes on to offer some other alternatives, including bar charts and lollipop charts, both of which play to the visual strength of measuring length.
Three paragraphs into this post you are probably already saying to yourself, "Here we go, another debate or rant about donut charts". Actually, that's not the purpose of this post. Instead I want to explore some other options and spark some dialog in the data visualization community around these points.
When talking with Ryan about his original KPI donuts, he used the term "visual impact". He agreed that bar charts are best practice, but he also pointed out that in certain cases they had less of an impact visually. That statement seems to steer down a familiar path, the compromising dilemma and subsequent balance of function vs. form. This topic is not the purpose of my post either, so I'll save further discussion of that for another day. In Steve's post, he offered a few alternatives that he described as "have your cake and eat it too". I think Steve sums it up very well:
"And I agree that the donut chart is eye catching, but I hate to sacrifice information for the sake of decoration."
Expanding on this idea I will walk through some additional examples, all of which will have a different level of "eye candy" incorporated into the visual, but each will utilize the primary visual strength of bar length.
My first approach was a "progress bar". These are very common and are used in places you probably didn't even realize. For example, Time Warner Cable uses a beautifully designed progress bar to show what time a TV show started, when it ends and how far into the show it is at that particular moment in time. It's also a nice blue and gray color scheme, very similar to the color scheme being used here. Here is the donut KPI using a progress bar.
My next approach doubles up the progress bar and frames it. The second bar is redundant, offering little additional value, but it does provide a frame for the text which presents an alternative design.
In the next example I tried to incorporate the donut and bar together. This design is very similar to the famous "Mind the Gap" logo. It provides a very different design and preserves some of the eye candy, but the progress bar becomes the functional part of this design because the donut is now fragmented.
If you consider the donut chart as secondary (after all we are bashing donuts), then it doesn't really matter that it's even less useful than before. That's the whole purpose of having the progress bar front and center. One argument would be that there's a lot of digital ink being used for limited function. In this next example, I remove some of the digital ink and use half of the donut. This solutions offers a bit of eye candy while at the same time minimizes the space and gives the reader the progress bar, our trusted source of bar length.
In the final example, I leave the entire donut chart intact just as it was in the original example from Ryan and Andy. I simply made the donut thinner and added the progress bar to the middle of the chart. I think this is the solution that is the "have your donut and eat it too" option. The bar is the primary focal point with the text label. The border around the bar and text is really nothing more than a circle that encodes the same data.
Maybe you are required to use a donut chart; a demand of the customer or an executive. Maybe you feel the bar chart, lollipop or other examples are simply not impactful enough. Whatever the reason is for using them, I would consider this, does adding a simple progress bar to the donut chart make it less impactful? Why not give the reader some help, supplementing with a visual tool that plays to the strength of the visual system?
My preference would be to use the first example with the simple progress bars. It's the easiest way to compare all of the regions next to each other, it's very simple and easy to understand, and it aligns with best practices. However, I hope these other examples help illustrate ways to incorporate useful design elements into visualizations that are otherwise deficient.
A few additional design notes:
1.) The donut chart and all of the examples above have an additional drawback. All of these examples and the original examples are capped at 100% so it will max out at 100% of goal. There is no way to compare 122% of goal and 106% of goal other than the data labels. As a KPI indicator, that may not be a big deal, especially with the label in the middle, but the lollipop and standard bar charts with a goal line that Steve suggested solve for this problem.
2.) In Steve's second lollipop example, he uses a graphic of a runner as the marker for the end of his lollipops. Be very careful putting graphics on the ends of bars and lollipops. Dots and circles have the advantage of having a clear focused point. In classical music, a conductor uses a baton to conduct a large music ensemble. The very tip of the baton indicates the beat, and it's called the ictus. Dots and circles provide a very clear ictus. Using images or other shapes such as asterisks, diamonds, squares, etc. spread the points to a larger region and often with multiple points, would be the equivalent of conducting with 5 batons or if Wolverine were conducting a symphony orchestra. In addition, these graphics can add additional height or length to the bars or lollipop that distort the comparison. I recommend dots and circles (and not too large) for the lollipop chart, dot plots, scatter plots and any markers, such as dots on line charts.
And with that, I'm going to call it a night. I've spent enough hours this evening making donut charts in Tableau and Excel to last me for an entire year. I hope you find these examples interesting. My Tableau workbook with these example is located here and an Excel workbook of similar examples is located here. I would love to see others in the Data Visualization and/or Tableau community offer additional alternatives and post their own solutions or suggestions.
