Quick tips to improve your data visualizations
It’s budget season again, which means everyone you'll soon be burning the midnight oil trying to tell the story of your city and what you want to accomplish next year. While a well-reasoned narrative is important, quality data visualizations are critical to telling your story.
There’s a reason they say a picture is worth a thousand words! A well-crafted data visualization can help quickly explain broad trends, distill large amounts of information into a consumable format, and help to explain confusing data.
A bad visualization, on the other hand, can actually detract from your message by confusing, cluttering, and distracting your readers. Here are three things to keep in mind when you need to reach for a chart or graph.
1. Avoid visual clutter
Extraneous noise only makes it harder for your readers to see the important stuff. Thing like borders, gradients, lots of colors (although be mindful of the need for good contrast ratios), and especially 3D rendered charts.
None of these flourishes help you tell the story of your data. Rather, they add superfluous noise which can take precedence over the important bits. Just because Excel has an option for it, that doesn’t mean you should use it!
Consider the two charts below, which present the same data. On the left, the story is drowned out by the colors and shadows and unnecessary 3D perspective. The chart on the right is plain, sure, but the data are front and center.
- Remove any visual clutter that doesn’t explicitly tell your story (borders, gradients, etc).
- Consider using shades of the same color rather for a more cohesive looking design, just be mindful of low contrast ratios which can reduce accessibility.
- Soften gridlines by reducing the opacity or using a lighter color.
- Always ask yourself, does this design flourish add to or take away from the story I’m trying to tell?
2. Pick the right chart for your data
Not every chart type is right for every data set. Each chart type encodes the data it represents in a different way, so knowing which type to use for your data is very important. Most of the data you’ll be presenting can be shown on either bar/column chart or a line chart.
Bar and Column Charts
Bar and column charts are functionally the same, so I’ll just refer to them as “bar” charts from now on.
Bar charts present differences in items, particularly discrete (non-connected) items. Some examples of discrete items you might be charting are expenditures broken down by department, revenues broken down by type, or patrons broken down by recreation program.
Bar charts encode the value of the data based on the length of the bar. Since the brain is quite good at comparing the lengths of things, bar charts are an easy way to compare differences (much better than pie charts, which you should never use).
Bar charts are perfectly fine for time series and connected data as well (things like revenue over time).
Since the length of the bar represents the value, you should always set the origin at zero. This also means being wary of stacked bar charts, since only changes in the bottom value can be easily deciphered.
Line charts present change in a data point over time. Since a line is a series of connected points, it should only be used for connected data (a time series, for example).
If you need to compare multiple time-series variables, you can use multiple lines. If the values are significantly different, it might be tough to see subtle changes.
The value is encoded based on the point, so you don’t necessarily have to have the origin set at zero. However, be mindful that small changes can appear large when the axis range is narrow, so make sure not to skew the interpretation of the data.
(Please don’t ever use) Pie charts
Pie charts encode data based on the angle of each wedge, and present value relative to a whole.
For some reason, pie charts are quite popular despite being very difficult to read at a glance. Humans don’t easily distinguish angles like we can with lengths, which is why bar charts are almost always a better option.
Additionally, we present pie charts with raw numbers, even though the data is encoded as a percentage of the whole:
By doing this, we’re fighting against the purpose of the pie chart while also making it harder for the reader to understand:
There’s rarely a good reason to use a pie chart, and we highly recommend avoiding them. (We don’t use any in ZacTax!) But if you must, here are some recommendations:
- Limit the number of wedges as much as you can;
- Order the slices from big to small, starting at 12 o’clock and moving clockwise;
- Present the data as a percentage of the whole (and make sure your percentages add up to 100);
- If you have more than 3-4 wedges, you should probably add labels;
- If you’re going to use a pie chart, please follow the first tip and don’t make it a 3D pie chart;
But really, you should just avoid pie charts altogether!
3. Don’t be afraid to get creative
You can go a long way with bar/column and line charts. Although it’s important to reduce visual clutter, they don’t necessarily have to be as boring as the examples shown above. Here’s a couple of parting tips:
- Since you’re removing borders from your charts, you can better incorporate them into the design of your documents. Be creative with it! Make your graphics the centerpiece rather than an afterthought.
- You can export charts from Excel to PDF, and then really spruce them up with a vector graphics tool like Adobe Illustrator or Affinity Designer.
- Other chart types can be extremely useful too, so don’t be afraid to use them when they convey your data in a clear and understandable way.
- Sparklines are a relatively new chart type that are meant to display data inline for supplemental understanding. They can be a great addition to long tables of historical data.
If you want to see some more examples of what to avoid, be sure to check out the Data is Ugly Reddit.