Re-working an Old Viz: US Government Spending

by Vivien Ho

After our first week of learning about data visualization best practices, DS22 received their first ever project! Each of us were given one of Andy’s old vizzes and tasked with deciding what was done well, what could be done better and afterwards, to implement some improvements of our own. I was given a viz on U.S. Government Spending in 2013, which shows how the annual budget was split across different federal agencies.



  • FORMATTING: the text colour and font are easy to read with sufficient contrast.
  • TITLE: I liked how Andy asked a clear, concise question which could be reasonably answered by the data. The short subtitle is also great at adding some extra context.
  • INSTRUCTIONS: for most people, packed bubbles are quite self-explanatory, with the total spending correlating to size. Similarly, the filters are clearly labelled and user-friendly, allowing the user to select their desired agency.
  • LABELLING: the white text on each bubble provides a nice contrast, however, as some agencies received a relatively small amount of funding, they were not labelled at all.


  • COLOURS: Andy allows the audience to choose the bubble colour, I’ll take each option in turn:
  • Colour by total spend: this is a diverging blue/red colour palette. Although this highlights the top 3 agencies quite effectively, diverging palettes should be used where the measure has a specific mid-point (think positive and negative growth) - this isn’t the case here.
  • Colour by agency: this becomes overwhelming due to the amount of agencies present...
  • INSTRUCTIONS: despite the instructions above the filters being clear, the option to choose ‘discretionary vs. mandatory’ as an option is confusing. Firstly, mandatory/discretionary spending should be defined, and how this metric is calculated should be noted somewhere too.
  • ANNOTATIONS: there are no insights to guide the reader to interesting parts of the viz.
  • CHART TYPE - both charts essentially show the same metric of total spending, but both struggle to show magnitude in an effective way:
  • The bubbles for the agriculture and veterans affairs agencies are almost equal in size, despite one being $20B larger than the other.
  • For tables, it is difficult to gauge the difference between agencies without looking at each cell individually.


  • MORE SPACE: I enlarged the viz to a standard 1200x800 size, allowing me to select a font size that was readable, and to increase the whitespace between elements.
  • CONTEXT: I made the title bigger but kept the original wording. However, I added some key insights, using bolding and colour to draw attention to certain statements. The information icon also provides a definition of mandatory/discretionary spending.
  • FILTERS: the ‘bubble colour’ filter was removed and the ability to select a bubble by clicking on the table was replaced with a simple Highlighter filter.
  • Some people could possibly find the amount of agencies to be overwhelming to look at, so I added a top N filter. This way, the audience can choose to only see e.g. the top 10 agencies in terms of spending.
  • COLOUR: I opted for my favourite sequential palette, the teal provides sufficient contrast but isn’t as intense as the bright red/blue.
  • CHART TYPE: I decided that both charts could be converted into a bar chart, sorted by total spending.
  • LABELLING: I right-aligned all text to create clean lines, and converted the agency names to Upper case using the UPPER() function. I liked the ‘Grand Total’ at the end of the original table, so to take this one step further I added a % of total indicator at the end of each bar. The grand total was moved to a more logical location and is much bigger.
  • CLUTTER: All gridlines, borders and zero-lines were removed as they added little value to the chart. I also got rid of the VIZWIZ logo (sorry Andy!) and replaced it with something more minimalist.


  • Due to time pressure, I overlooked the fact that by adding a Top N filter, this would subsequently recalculate my % of total every time the filter was changed.
  • I am still unsure whether the sequential colour palette is necessary given that the magnitude of spending is already shown in the length of the bar. I would possibly colour the majority of the bars light grey, with the highest spenders in teal.
  • The agencies themselves have very long names and take up quite a bit of space, I would probably explore ways of changing the aliases of some of these.

Overall, I found our first task really fun, and it definitely taught me to be more careful with my design choices in the future!

Fri 30 Apr 2021

Fri 26 Mar 2021

Thu 25 Mar 2021