By Emily Goyne

At Precision.BI, we believe that the data you display should speak for itself and not require your audience to search for meaning in the display. We encourage our users to select graphs and pictorial representations of their data that the end user will easily grasp and understand. Line graphs effectively display data points over time and histograms or bar charts display comparative summary data by category. On the other hand, while pie charts may seem like an obvious choice for displaying percent of total or summary data by category, they are not the most clear visual display. As Stephen Few writes in his column “Save the Pie Charts for Dessert”, there are more effective ways to present percent of total information. Based on the angle of display, the label clarity and overall “busy-ness” of the pie chart, data can be misinterpreted or lead to more questions rather than answers. We agree with Few and offer alternative graphing options in Precision.BI.

Consider the following set of data: total charges and a count of charges by CPT code for 2 years. The analysis is sorted by CPT code so the top values are not immediately evident, but can be determined pretty quickly.

If we were to put this set of data into a pie chart:

It is easy to see which CPTs are the very highest volume, but after the top two or three CPTs, comparing the slice angles becomes difficult without a lot of labeling.

By presenting the same set of data, with similar labeling in a percent of total bar chart in a PresentationCenter, the relationship between the data points become much clearer:

If we wanted to compare the two years of data using pie charts we would have to resort to two pie charts. Start with the set of graphs on the top, top and bottom. Comparing the angles and data points for these two pie charts becomes nearly impossible. We can add labeling (graphs on the bottom) to give us the total counts on each graph, but our end user still needs to look back and forth between two year graphs to draw any conclusions.

If we were to move back to our bar chart percent of total view, we can now see two years as two separate series in the same chart where we can easily draw conclusions based on the data. The result is an uncluttered display with high amount of meaningful data.

As Few argues, all too often pie charts are given more responsibility than they can present. Also with modern graphing tools in Microsoft Excel and other software packages, we have the ability to make the pie charts more dynamic and visually pleasing. But, with all those enhancements, the limitations on comparing slices remain just as difficult as ever, if not more so. If you are comparing summary figures and are tempted to use a pie chart, consider using a percent of total bar chart as we reviewed or a simple grid such as the one below. In addition, incorporating a variance adds quite a bit of value and is something you couldn’t easily include in a pie chart or other graphic display.

We recommend keeping data displays clean and uncluttered. Free from unnecessary graphics lines and gadgets, the data you present can speak for itself. Sometimes that means electing to avoid graphs and gauges altogether and sometimes it means trying a different presentation method.