• Ben Watt

Pie Charts: a pleasant slice.

In the Business Intelligence world, there is consensus among some that Pie Charts have no place on a report. There's even a Twitter hashtag: #SayNoToPieCharts. I'm here to offer my defence of this visual, and by the end of this article, if you haven't reconsidered your stance I will gladly eat humble pie chart.

Without going too deeply on this subject, data visualisations are intended to invoke visual queues for us to spot a trend, comparison, outlier to direct our eyes to something that needs our attention. Excluding the basic table visual, if I were to take a guess, Line and Bar/Column Charts take up 60-80 % of visuals we use as they clearly convey a trend/comparison/outlier. Scatter charts (my personal favourite) are not as popular but tell a great story of three metrics without us having to mentally process the numbers and their insight, something referred to as cognitive load.


Pie Chart Discord


There is a reason pie charts have been pushed out of the cool kids gang. Unfortunately, they are almost always used in a way that does not clearly show a trend/comparison/outlier. When several categories are used in a pie chart, suddenly you are forced to engage your brain to scan numbers, compare them and deduce any insight. This is totally counter-intuitive to what a visualisation is supposed to do.


For example, the below two charts are displaying the same data. The pie chart is a terrible choice in this scenario and as mentioned above this is where its very frequently used, or more accurately, misused. It's hard to compare the values and determine the ranking order & comparative difference without looking at the numbers. Furthermore, multiple colours were needed to separate the slices, which is not a data viz best-practice in this case.



Pie Chart Ubiquity


This data viz has been around a long time and is a frequent go-to for people building dashboards. I did a search for "dashboard" on Google and Bing images. In the top 18 Google responses (that fit onto my monitor) 15 of them had a pie chart. Bing had 12 out of 21. Given their popularity, how can we just blanket refuse them?


They are commonly used for a reason:


Pleasing to the Eye


Over the years I came to realise that in building a report, function-lead-design satisfies the insight task, but form-lead-design satisfies the human being. As such, I suggest incorporating a mix of these.


Pie Charts are among the top of the pile when it comes to adding a small dose of pleasantness to a report. Feel free to forgo a couple of pixels on your report for something pleasing to the eye, you will always get a positive response from your users. It doesn't have to be every page, but that first time you deliver a report, make it look good! You never get a second chance to make a first impression!


Have a look at the Power BI home page. The below is shown as an example dashboard. It's designed purely to make you say "wow" and not for you to find insight, and rightly so. A fully best-practice data viz example here would not look as good.



The Pie Chart Use Case


There are two scenarios where I use a Pie Chart.


First is when I'm simply looking to make a dashboard look nice for demos. It's purely a sales technique to make a report pleasing, and it works EVERY TIME.


Second is an actual data scenario that pie charts work for. The use case is when there are only two categories and one of those categories is at least 3x bigger that the other. In other words if the ratio is 75/25 or more then a pie chart looks fine.


Take a look at this example. In my opinion, the pie chart is a bit easier to understand the ratio, and it has the pleasant-factor.


Hopefully this has given some food for thought, or at least made you hungry for some pie!

Copyright © 2019 Datalineo Ltd. All Rights Reserved