Appearances are deceiving
Few months ago I wrote a long-form post for Lucient blog about the Data Visualization and Storytelling rules and Best Practices. If you want read it, you can find the two parts here and here.
In this post and the next ones I will come back to some concepts and features focusing on one at a time.
A lot of times we use in Power BI the map visual, assigning to bubble size a meaning. Other times we use the bubble size also in a scatterplot to give users an additional information.
However, these information could be deceiving! We have to go back to the famous catalogue of visualization categories created in 1984 by William Cleveland and Robert McGill.
As you can see in the picture above, Areas are not the best visuals for an accurate estimate. This is because of our brain is not designed to detect angles and areas with the same precision than lengths.
We can use Power BI to verify that.
I will use a simple data model with some data about English Football League, loaded from the web.
For our demonstration, I’m going to compare the titles won by two of most ancient and famous English clubs: Arsenal and Aston Villa. I create an empty scatterplot and drag the number of titles as bubble size without other measures on the x and y axis. Filtering appropriately the club dimension, now we have the only two bubbles of Arsenal and Aston Villa on the same point in the middle of the visual, as in the picture below.
If I tell you that Aston Villa has won 7 titles, how many titles have Arsenal won in your opinion?
I asked this simple (?) question to a lot of people and most of them answered 9 or 10, some brave tried with 11, no more. Well, the correct answer, for the happiness of Arsenal fans, is 13! Almost the double! Unbelievable, right?
We can verify it simply changing the visual type from scatterplot to clustered bar chart.
Summing up, we can conclude that bubble size could be used only for a qualitative purpose, not quantitative. For the letter, a bar or column chart is always the best choice.
