Data Analysis › Charts & Visualization
Choosing a Chart
Matching the comparison you want to make to a chart form that shows it.
Also known as: choosing a chart, which chart to use, chart type selection
Choosing a chart is matching the comparison the reader is making to a form that shows it. The form is not the hard part; knowing what you were asked is.
Start from the question
| You want to show | Reach for |
|---|---|
| Compare amounts across categories | Bar chart |
| Change over time | Line chart |
| Relationship between two measures | Scatter plot |
| Distribution of one measure | Histogram, or box plot |
| Part-to-whole | Stacked bar; treemap when there are many parts |
| Composition changing over time | Stacked bar, or area |
| Ranking with a long tail | Sorted bar chart; Pareto chart to show the concentration |
| One number against a target | The number, with a sparkline beside it |
| A single status over time | Sparkline |
Part-to-whole is worth a caveat: pies are weak for it. Angle and area are harder to judge than length, and a pie with eight slices cannot be read at all. A stacked bar does the same job with a comparison people can actually make.
Three questions before you pick
- What comparison is the reader making?
- How many categories are there?
- Is time one of the axes?
Most mis-chosen charts fail because the answer to the first question was never settled — the analyst charted the column they had rather than the comparison they were asked for. A scatter plot does not answer “which region is biggest” any more than a bar chart answers “do these two numbers move together”.
The classic mistake
Reaching for the form you know. If your default is a bar chart, every question becomes a ranking; if it is a line chart, every question becomes a trend. Learn what each of the main forms is good at — histogram, box plot, scatter plot, stacked bar chart, treemap and Pareto chart each have one — and the question picks the chart for you.