Choosing the right chart for an infographic

Start with the comparison the reader needs to make, and the chart type mostly chooses itself. "Which is biggest?" wants bars. "How did it change?" wants a line. "What share?" wants a stacked bar, a unit grid or, occasionally, a pie. When there is only one number, a big number may beat any chart.

Infographics add a twist that dashboards and reports don't have: they are read quickly, often on a phone, often by people who didn't ask to see them. That favors simple, familiar chart types over clever ones.

Name the comparison first

Look at your message sentence and find the verb. It usually points to one of these:

If two comparisons are fighting for the same chart, make two charts.

Bars for amounts

The bar chart is the workhorse of infographics because everyone can read it. Length compares accurately, and labels fit easily beside horizontal bars.

• Start the axis at zero. A bar's length is its value, so a truncated axis exaggerates differences.
• Sort bars by value unless the categories have a natural order (age groups, months).
• Use horizontal bars when labels are long. They also fit phone screens better.
• Label values directly at the end of each bar when there are fewer than about fifteen.

In the inflation piece above, each bar is a separate yearly rate, so bars fit better than a line. One bar is highlighted because it carries the message.

Lines for change over time

Use a line when values are measured at many points and the slope is the story. Put time on the horizontal axis, running left to right.

• A line chart may start above zero, because it encodes position, not length. Say what the axis range is and label the gridlines.
• Space points by their real dates. If surveys happened at uneven intervals, an evenly spaced axis distorts the slope.
• Keep it to a few lines. Beyond four or five, consider small multiples: a row of small charts with the same axes, one per series.
• Label the lines directly at their ends instead of using a legend box.

When a value changes in jumps, such as a legal rate that stays fixed until it is changed, use a step line. Our minimum wage piece shows why a sloped line would mislead.

Shares of a whole

Parts-of-a-whole data invites pie charts, and pies are not always wrong. They work for two or three parts of clearly different sizes. They fail when there are many slices or the slices are close in size, because people compare angles poorly.

Better options in most cases:

• A single stacked bar for a handful of parts, labeled directly.
• A unit chart (a grid of 10 or 100 icons or squares) when the message is one percentage. Readers can count.
• A sorted bar chart when there are many parts and ranking matters more than the total.

The world online pictogram uses a 100-person grid for exactly this reason.

When a number beats a chart

If the message is one figure, a chart with one bar or one slice adds nothing. Set the number large, put a short label under it, and add one line of context: compared with what, when, and who was counted. Infographics have used big numbers since the start because they work.

Big numbers become a problem when there are six of them in a row with no hierarchy. Pick the one that answers the headline and give the others normal type.

Charts to use with care

Some chart types appear in infographics far more often than they help:

• Donut charts have the same problems as pies, and the hole removes the angle cue people use most.
• Bubble charts ask readers to compare areas, which people estimate badly. A bubble twice as wide has four times the area.
• Radar charts look technical but make comparisons hard, and their shape depends on the order of the axes.
• Word clouds show which words are frequent but not by how much, and long words look more important than short ones.
• Maps color large, empty regions as prominently as small, crowded ones. Use a map when location is the message, not just because the data has a place column.
• Icon-scaled pictures (a small house and a big house) mislead when both height and width grow with the value.

None of these is forbidden. Each needs a reason.

Test it like a reader

Show the chart to someone for five seconds and ask what it says. If they describe the chart ("there are some bars") rather than the finding ("2022 was the worst year"), the chart choice or the emphasis is wrong. Usually the fix is simpler than the first draft: fewer series, one highlight color, and a headline that states the finding.

When the chart is right, move on to color and type, which decide whether people can read it.

Questions