While I understand how exciting it is the first time you create a bubble chart or how cool you think a tree map looks when you are first learning data visualization there is ALMOST always a better option than the three charts listed in this blog.  Data visualization is not always about making the data look as exciting as possible, no matter how fun that is.  It is about making the data as easy to digest as possible so you better get used to bar charts!  Below I have listed three charts that you should always avoid using if you can, but ofcourse everything has it’s place… At least almost everything.

 

Bubble Chart

My initial application had 2 bubble chart and my follow up dashboard for the second stage of my interview also included a bubble chart.  I thought they added something new and made the dashboard more visually appealing however I have since learned just how much harder it makes data visualization.  Throughout this dataset I am using the Superstore dataset to make these visualizations so you can also play along at home.  In the first bubble chart I have plotted the amount of Sales by each Region with the size of the bubble displaying their total sales.

While I agree it is a ‘cool’ visualization to look at, it is extremely hard to determine which region has more sales between West and East and even if we added labels it requires way too much time and effort to really figure out what is being displayed in this plot.  An alternative to this is the trusty bar chart as we can see below.  The differences between regions is made a lot more clear, along with this we have a scale on the left provided by the axis ticks and each region is easily determined by the labels on the X axis.

 

Tree Maps

Perhaps with a very small amount of variables a Tree Map may work as a more visually appealing way to display groups within a data set however in almost every case it is a very confusing, wasteful way of displaying data.  I have fallen trap to the appeal of a Tree Map before however when asked what was shown in the smaller boxes I had no answer, outside of the big boxes of focus the Tree Map didn’t show anything of importance.  Below I have used a Tree Map to show the sum of Sales by Sub Category.

It is clear that Chairs and Phones are the two largest but which one is 1st? Also, how do the others compare? The color shading helps separate it however it is a wasted opportunity to use color to represent something else.  Below is yet another bar chart, however it is already clear why it is a better way to display the data.  We can see that Chairs has the most overall sales and then follow the Sub Categories down to see how all others compare to each other.  I have also coloured the bars by the amount of orders for each Sub Category, adding another dimension which is now available as the differences are clear due to the easily comparable bar chart.

 

Tables

A controversial one but if we are trying to visualize data we should try to opt towards visualizations not just numbers in a table.  There are times where tables are appropriate however a good visualization will save the user time and show the important insights straight away.  When using a table it will take time to identify where there are big changes and even if we use colours it takes time to read and compare numbers against each other so charts combined with colours should be considered as an alternative.  Below I have used a table to show the Profit for each Ship Mode, Category and Sub Category.

It is easy to see what each combination of variables makes in Profit, however it takes time to identify and comparing each value is hard to do too.  For example, how long does it take you to find all cells that do not make profit or which combination of variables made the most money?   Below I have used a bar plot to show the same thing and coloured the losses in red.

It is very easy now to see which rows did not make a profit and even compare them to the other rows.  We can see where the most profit was made and make any other comparisons needed with a quick look.   If we needed to investigate further we could use our mouse to hover or manually label each row with the profit amount for each cell.

 

Conclusion

There is no perfect chart or plot for every situation however the 3 listed in this blog are usually not the ones.  When choosing your visualization it might be best to see if you can figure out insights within a few seconds, or test it on some one less data literate.  Our job is to make data easy to read, not exciting.  If we can do both it’s even better but the first job is to make it digestible.  It can be fun to play around with creative data displays but don’t get too carried away or it won’t be useful for anyone.

Mikael Nuutinen
Author: Mikael Nuutinen