Revision of Packed Bubbles

I saw a packed bubbles data visualization highlighted on Tableau’s Viz of the Day blog the other day, and thought it could use a revision: http://www.tableausoftware.com/public/gallery/2013-us-federal-budget

Why this needs a revision:

- Unfortunately with Packed Bubbles currently in Tableau, the visual display may not be accurate. The smallest packed bubble Tableau draws is 1/25th of the largest bubble. So when the ratio of the value the bubbles are sized by is greater than 25, the display is not accurate. I believe Tableau is actively working to correct their drawing method.
- The use of color in this visualization is not helpful for several reasons. When agency is selected for coloring, colors are reused for multiple agencies because the palette has 20 colors, and there are more than 20 agencies. When changing the Color to different measures, blue can represent either large negative values or small positive values because of the dynamic range of blue to red. This also creates duplicate encoding when the color and size of the bubbles are the same measure.
- Sizing bubbles on both negative and positive values at the same time is not a good practice. Even if they were absolute values, bubble size is useful for one direction at a time, either positive or negative.
- The overall analysis this graphic enables is limited at best, only enabling a user to see the largest, and see only one measure at a time.

To address those issues and more, I created: http://public.tableausoftware.com/views/2013USBudget_3/BarsandDivergingBars

Benefits of this revision:

- We can read the full Agency name
- Displays multiple measures at once, instead of one measure at a time
- Configurable sorting with an aliased integer parameter
- There is a meaningful sort order, insted of a random clumping
- Visual size encodings match the value represented
- Color is used sparingly
- Overall, the user can make useful comparisons, and quickly see interesting outliers

I can see a number of interesting things to potentially investigate:

- Most large Outlays are mostly Mandatory, except for Defense and Interest.
- Other spending has a large negative % Discretionary, as represented by the over 100% Mandatory
- Most Agencies are mostly Discretionary
- Veterans Affairs is the only one close to 50/50 for Discretionary/Mandatory
- Can sort by Outlays, select the top 3, and see in tool-tip: combined they are 66% of Total Outlays

These points are more informative and useful than simply seeing which are the biggest.

Just because we can make a packed bubble visualization does not mean we should. I find it concerning that Tableau highlights the use of their product in ways that are not in alignment with the principles of visual understanding.