Laurel Hyatt gave us a quick tour of the system she uses to diagnose and treat ailing charts. Taking the medical metaphor further, she said that at the substantive editing stage, the goal is prevention; at the copy-editing stage, the goal is successful treatment; and at the proofreading stage, sometimes all we can do is try to keep the chart alive. The earlier you can intervene when you spot a poor chart, the better.
Charts are the trifecta of communication: numbers, words, and pictures. When they come together in harmony, said Hyatt, they can be a beautiful thing. If one or more of those elements goes wrong, the chart can be a dog’s breakfast.
Charts should tell a story. Hyatt showed us examples functional and dysfunctional charts in each of four of the most popular types of charts. Here is just a sample of her advice:
Bar charts
Do
- make the y-axis begin at 0. Doing otherwise could exaggerate the difference between two bars and be misleading.
- show scales (such as years) in even increments.
Don’t
- use more than about ten bars per chart.
- use more precision in number labels than necessary.
If you have too little data, consider using text instead. If there’s too much data, try a table. Even-year time series may work better as a line chart, and parts of a whole that add up to 100% may work better as a pie chart.
Line charts
Do
- use a scale that clearly shows changes over time (or whatever you’re measuring on the x-axis).
- use even increments of time (or whatever you’re measuring on the x-axis)
Don’t
- use too many lines. Even with a legend, crowded lines will be confusing and hard to interpret.
When you have uneven increments of time, a bar chart might be a better choice; if you have too many lines, a bar chart or table might be more appropriate. If the data don’t change enough over time, consider using text.
Pie charts
Data visualization specialists like Edward Tufte dislike pie charts, but Hyatt believes that they can serve a function when the aim isn’t to do any precision comparisons.
Do
- use pie charts to show parts of a whole.
Don’t
- use fewer than three or more than about six slices.
- use more than one pie to compare apples and oranges.
- use slices that represent 0%.
If you have too many slices, or the slices are too thin, a bar chart or table might work better. If there are only two slices, summarize the data as text. Changes in time are better compared using a line chart rather than separate pie charts.
Pictographs
We don’t come across too many pictographs in our work, but they can be very effective when done creatively. You can suggest using them at the developmental and substantive editing stages if you think they work well to get the message across.
Do
- use proportionate size to indicate data.
Don’t
- use a pictograph just because it looks cool.
- use three-dimensional objects to represent anything other than volume.
Too often, said Hyatt, when there’s geographically sensitive data, people default to using some kind of map, but maps are not always the most effective choices, especially if you’re expected to compare data between locations. Opt for other types of charts, tables, or text if the data you’re representing is very technical or if it has to be shown precisely.
Further resources
- “Save the Pies for Dessert” by Stephen Few
- How to Lie with Statistics by Darrell Huff
- The Visual Display of Quantitative Information by Edward R. Tufte
- Making Data Meaningful, United Nations Economic Commission for Europe