Pie charts or parts-to-whole analysis

Those who have read Learning QlikView Data Visualization may wonder why I didn’t talk about pie charts in the book.  Some may even assume that I am a data visualization idealist that tries to deny people’s temptations to create pie charts, but while I do support those idealists, I have created more than a few pie charts in my eight years working with QlikView.

I have strived to create alternate charts that more effectively show a parts-to-whole relationship, but some users with not accept any other chart that is not a pie chart.  So, inevitably pie charts sometimes end up in my QlikView applications.

Sadly, they take up a relatively large amount of space in return for communicating very little information.   I compare a pie chart and its ability to communicate to a sentence written by first-grader.  It’s not hard to imagine the sentence, “My favorite food is spaghetti,” written in gigantic, red print on dark blue construction paper in the halls of an elementary school.  Years later after becoming more sophisticated writers and communicators, why do we regress and use flashy, circle-shaped objects to convey so little information?

This may surprise you, but if you are new to data visualization, I recommend you go ahead and use pie charts for demos and your first QlikView applications.  Experimentation is a necessary step for some of us to learn that pie charts are a poor use of space when we want to create insightful data visualization.

For an example of poor data visualization techniques, look no further than a QlikView application I made in 2008.  I was still in data visualization elementary school creating pie charts on top of black backgrounds.

QlikView_VeryOldUglyDashboard

Incredibly, I’ve also witnessed my customers becoming more aware of great data visualization and analysis.  Fewer customers ask for pie charts, and more ask for other types of charts that have greater data density and communicate more clearly.

Take courage and stop creating pie charts because you think the other person likes them, and start proposing more insightful methods to discover data.  I preferred to show these other more insightful charts in my book, so I eliminated the section about pie charts when I went over the book’s page limit.

If you are interested in creating pie charts or an alternate method to show a parts-to-whole relationship, please follow along with the exercises below.  Before you begin the exercises download the QlikView application Sales_Project_Analysis_Sandbox.qvw.  If you have the QlikView Personal Edition follow the instructions found in this blog post.

Pie Charts

Pie charts are created in the same way as bar charts.  In the first step to create a chart, select the following option.

QlikView_PieChartIcon

Pie charts work best when the dimension has few unique values.  In our first pie chart, we use Division as the dimension.  Lastly, we define the metric to be the sum of net sales

sum([Net Sales])

Those three previous step will give you a basic pie chart.

QlikView_FirstQlikViewVersion

Along with the tips included in the data visualization style guide found in my book, we also make the following changes to a pie chart’s properties.

Dimension tab

If necessary, reduce the number of slices shown in the pie chart by enabling the Restrict which values are displayed using the first expression checkbox.

Expressions tab

We want to add both percentage and numeric labels within the chart, so we perform the following trick

  1. First, we add labels showing each slice’s percentage share of the pie by copying and pasting the expression so that the same two metrics are listed in the Expressions tab.
  2. Select the second instance of the metric and enable both the Relative checkbox and the Values on Data Points checkbox.
  3. We keep the first metric enabled so that we can show number values in the legend.

Presentation tab

We show each division’s numeric values by enabling the Show Numbers in Legend checkbox

Sort tab

Finally, it’s always easier to read a pie chart if the slices are ordered from greatest to least, so we enable the Y-value checkbox.

After adjusting the size of the chart’s subcomponents using Ctrl+Shift, we have the following chart.

QlikView_FinalPieChart

Parts-to-Whole Bar Chart

An alternate method is a bar chart that makes the part-to-whole relationship stand out more than in normal bar charts.

QlikView_PartToWholeBarChart

We can easily identify the approximate percentage of net sales each division possesses at the same time that we compare more accurately the difference between each division.  Additionally, since the chart is a bar chart, we can use various techniques mentioned in my book to increase its data density.  For example, we can make embed a heat map.

After creating a bar chart to analyze the ranking of each division according to its net sales, we make the following changes to the chart’s properties to convert it into the previous chart.

Expressions tab

Add the following expression with the label Others’ Net Sales.

sum(total [Net Sales])-sum([Net Sales])

Style tab

Enable the Stacked radio button.

Important Functions

Total

Remember that we use the Total keyword to calculate the total net sales regardless of the dimension value.  The result of an expression that contains Total will be the same for every row.

As a final note, if we were true minimalists, we could eliminate the light blue bar stacked above the darker blue bar and fix the x-axis to 100% of net sales.  The parts-to-whole analysis would be done comparing the bar’s length with length of the x-axis.

Hope to see you around,

Karl

Author: Karl Pover

Owner of Evolution Consulting, which provides Qlik consulting services throughout Mexico. Author of "Learning QlikView Data Visualization" and "Mastering QlikView Data Visualization." Qlik Luminary since 2014.

4 thoughts on “Pie charts or parts-to-whole analysis”

  1. Karl,

    In my blog post about linear fractions, I did not actually suggest an alternative form of display for part-to-whole analysis. My post is about a better way to teach children the concept of fractions. I have written about alternatives to pie charts for part-to-whole analysis many times in the past, however. For almost everything that you must do when examining and comparing parts of a whole, a regular bar graph works best. Bars, which are linear objects that share a common baseline, are easy to compare, but slices of a pie are not.

    1. Stephen,
      Thank you for your feedback and correction. I’ve deleted the section that referenced linear fractions as an alternative to pie charts. Readers can view your original post here. Sorry.

      I agree that normal bar charts are usually enough to perform a part-to-whole analysis. I find it difficult to estimate a bar’s percentage participation of the whole by comparing it with other bars. I depend on comparing a bar’s size to an x-axis that starts at 0 and ends at the total amount of all bars.

      My proposal goes a little further and risks adding redundant data to each bar so as to emphasize that type of comparison. My thought is that if I can clearly define the whole within proximity of the part, I can more easily estimate the percentage of the whole occupied by that part.

      Maybe, just as easily as I understand the linear fractions on your post.

    1. Hi Ulrich, thanks for the comment. Real pies are also hard to resist. It’s funny to be pie charts shaped like donuts, candy and pills. We love circles for some reason.

Leave a Reply