About two weeks ago, I asked my LinkedIn community:
“Do you use pie charts/donut charts to visualize data?”
67% of respondents said that they use pie charts and/or donut charts (often or sometimes). Below is a screenshot of the answers received on LinkedIn.
Of course, this is not scientific research, so take it with a grain of salt. Now, writing an article about data visualization without trying to visualize the data would be like planning to cook a great meal without salt and pepper (and, no, I am not a great cook but I've learned this lesson early on!). So, let's visualize these results in a few different ways: a regular bar chart, a stacked bar chart, a donut chart, and a pie chart.
Which one do you think is the best?
Which one is the worst?
Answer both questions in terms of how fast and accurately you can decode the information.
Remember your answer. We’ll come back to it later in this article.
Now that you answered these questions, I won’t jump in to tell you what the right answer is. Neither will I tell you what you should do. Instead, I invite you on a journey. A journey through the research published in this area. A journey that will enable you, my reader, to decide whether pie charts are the worst…or the best.
Let’s begin!
1914: Willard Brinton, a pioneer in the field of information visualization, published the book Graphical Methods. The book is remarkable and helped establish the foundation of data visualization as a discipline that is still discussed and researched today. Regarding pie charts, Brinton wrote:
“The circle with sectors is not a desirable form of presentation... if the horizontal bar method were used as frequently as the sector method, it would be found in every way more desirable.”
Conclusion #1: avoid pie charts.
----
1926: Over a decade later, Walter Crosby Eells, a prolific writer who frequently published on mathematics and education, was the first one to defend pie charts, in a study entitled The Relative Merits of Circles and Bars for Representing Component Parts. His research evaluated both the speed and accuracy of judgment. The result?
Pie charts can be read as fast as the bar charts, and more accurately than the bar charts.
Conclusion #2: pie charts are great.
----
1994: This was a critical year in the field of data visualization. This was the year when William S. Cleveland, Professor at Purdue University, published The Elements of Graphing Data. In his book, Cleveland concluded that
“Data that can be shown by pie charts always can be shown by a dot chart. This means that judgements of position along a common scale can be made instead of the less accurate angle judgements.”
Conclusion #3: avoid pie charts.
----
2001: Almost a century after Brinton published his highly influential book, Edward Tufte, professor at Yale University and one of the most prominent figures in the field of data visualization, published The Visual Display of Quantitative Information. In this book, Tufte shared his perspective on pie charts:
“A table is nearly always better than a dumb pie chart; the only thing worse than a pie chart is several of them [….] Given their low data-density and failure to order numbers along a visual dimension, pie charts should never be used.”
Conclusion #4: pie charts are the worst.
----
2007: As pie charts were growing in popularity in the business world, they were becoming increasingly criticized in the data visualization world. Stephen Few, another well-known voice in the field in data visualization, published an article entitled Save the Pies for Dessert. Although the title of the article was provocative, Few acknowledged that pie charts have a unique strength – “the message part-to-whole relationship is built right into it in an obvious way.” However, he went on to say that:
"...a bar chart, while it is not immediately designed for part-to-whole comparisons, it could be used for this purpose by using a percentage scale."
Conclusion #5: avoid pie charts.
----
2011: More recently, Cole Nussbaumer Knaflic, the author of Storytelling with Data, was vocal against the use of pie charts. Her blog post, death to pie charts, highlighted her dislike for pie charts and strongly advised her readers against using them. It’s worth noting that, in another article published six years later, Cole softened her view on the topic:
“Pie charts are not inherently evil. Like pretty much any tool, they can be used well and they can be used not-so-well.”
Conclusion #6: avoid pie charts.
---
2016: Drew Skau and Robert Kosara published two papers: Arcs, Angles, or Areas: Individual Data Encodings in Pie and Donut Charts and Judgment Error in Pie Chart Variations. In these two papers, the authors question the assumption that pie charts are read primarily by central angle (as some of the previous research had assumed). They concluded that “angle is not likely the main, and certainly not the only, way we read pie charts.” They also showed that:
"Both studies point to angle being the least important visual cue for both charts, and the donut chart being as accurate as the traditional pie chart."
Conclusion #7: pie charts and donut charts are great.
---
2019: A paper entitled Dissecting Pie Charts was published in the Human-Computer Interaction journal. The authors, Harri Siirtola, Kari-Jouko Räihä, Howell Istance and Oleg Špakov, carried out an eye tracking study which revealed that:
“doughnut charts with a medium size hole have a slight edge over the other chart types (e.g. pie charts).”
They also showed that, “contrary to common claims, for information extraction also the area and length of sector arc are used in addition to the angles of the sectors.”
Conclusion #8: donut charts are better than pie charts.
----
2019: A second key study this year, The Cost of Pie Charts, was presented at the 23rd International Conference Information Visualization. Harri Siirtola, university researcher in the TAUCHI Research Center and Adjunct Professor of Interactive Technology, showed that:
“ a pie chart is slower and less accurate than the stacked bar chart, especially when the difference between the elements is small.”
Conclusion #9: avoid pie charts.
Here, it is worth pausing and noting that, while many highly influential ideas in the field of data visualization have had some degree of originality and have been based on research, they are also, as the Latin idiom goes: “nanos gigantum humeris insidentes” (or “standing on the shoulders of giants”). In other words, many contemporary data visualization authors have been discovering truth by building on previous discoveries, not necessary by making new discovering. Stephen Few, for example, acknowledges that his ideas were influenced by William Cleveland’s writings.
To summarize, most data visualization experts continue to dislike pie charts and donut charts. While practitioners that have authored books are very vocal about their dislike for pie charts, research shows mixed results. Some studies concluded that pie charts can be read as fast and as accurately as bar charts, while other papers showed that bar charts are both more efficient and more effective than pie charts.
Now, do you remember my questions about the pie, donut and bar charts showing the LinkedIn poll results?
Which one did you think was the best?
Which one did you think was the worst?
After reading this article, has your answer changed? Whatever your answer is, my hope is that this article provided you with the tools to make your own informed decision. Although there should be some science behind your answer, your decision remains, at least to a certain extent, subjective.
If you’re curious about my perspective, here you go: IT DEPENDS.
When I offer data visualization trainings, I typically recommend the following: Consider your audience, the number of pie slices and the story that you are trying to tell.
If you don’t have more than 2-4 slices, are looking for a part-to-whole comparison and are convinced that a pie chart or a donut chart is the best way to convey the story to your audience, then I’d say go ahead and use it.
Do pause, think twice and put yourself in the shoes of your audience before making the final decision.
Comments