My Data Visualization Mission Statement

1. Reduce the time to insight
2. Increase the accuracy of insights
3. Improve engagement

If these three principles look familiar, it’s because I also call these the benefits of data visualization. If I ever have a decision to make while I’m creating a data visualization, I ask if the change will accomplish all three of these things. If it fails at any one of them, I simply don’t do it.

My favorite example, and the way I often get an audience to evolve from their spreadsheet mentality, is to show the difference between a raw table of data and a highlight table. When adding preattentive attributes to improve processing time, you always (1) reduce the time to insight and (2) increase the accuracy of insights. I also argue that a highlight table is much more (3) engaging than the raw table, so all three criteria are met.

I wouldn’t be doing what I’m doing if I didn’t believe data visualization was the key to understanding data, but just illustrating that these principles affirm my reasoning. Let’s look at another example: pie charts. When I argue against the use of pie charts, it’s almost become a humorous debate. I’m often confronted with people who are just positive they have the one perfect use case where they are forced to choose a pie.

I want to go on the record to say that when I recommend avoiding pie charts, it’s not because I’m a data visualization elitist living up on the hill in my castle. It’s because of my data visualization mission statement. While it can be argued that pie charts are more engaging than a bar chart, they fail relatively dramatically at reducing the time to insight and improving the accuracy of insights.

Let’s try to answer a basic business question using Tableau’s Sample – Superstore dataset: What are the three worst-performing product sub-categories in sum of sales?

Pie Chart

Sorted Bar Chart

To keep it a somewhat fair fight, I sorted the dimension members in both charts. With the pie chart, the size of the slices prevents us from showing all of the labels. You can provide this information in a tooltip if using Tableau, but it doesn’t help non-interactive versions of the view. This means you are forced to provide the sub-category names through a color legend, increasing the cognitive load on your end user (i.e. they have to look back and forth at the chart and the color legend). This increases the time to insight.

New to Playfair Data?

Watch our 90-second intro video and receive tips on making your data usable.
 
Watch Video

Further, we are much better at comparing the lengths of bars compared to areas of a pie. If I wasn’t such a nice guy and sorted the charts, it would be much harder to determine the bottom three sub-categories when looking at the pie chart compared to the bar chart. Using a bar chart would provide more confidence that you are correctly picking the bottom three. In other words, it would increase the accuracy of your insight.

I know I’m not going to convince everyone, but sharing this because the reasoning behind my opinion on pie charts is objective. If you think you have the one case where a pie chart is easier to process and improves the accuracy of insights – go for it! I just have never come across one.

Pie charts are engaging, which is one of the reasons they’re still so pervasive in reporting. I’m a huge believer in engagement, but it has to be done tastefully with a balance between the first two principles of the mission statement. To help make some more effective chart types more engaging, I documented 3 Ways to Make Beautiful Bar Charts and 3 Ways to Make Lovely Line Graphs.

For some general ideas on making your data visualization more engaging, check out the Design section of the Triple Crown Framework.

Thanks for reading,
– Ryan