Charts
Use a chart to show a comparison, trend, composition, distribution or relationship. Pick the simplest chart type that supports your point, and lay it out with the chart template.
Choosing a chart type
Find what you want to show, then answer the questions about your data.
Showing change over time? Usually a line chart
- Is the data continuous (daily, weekly)? → Line chart
- Do the totals matter? → Area chart
- Is the composition changing? → Stacked area chart
- Is the data discrete (years, quarters)? → Column chart
- Is the composition changing? → Stacked column chart
- Only two points in time, several categories? → Slope chart
Comparing categories? Usually a horizontal bar chart
- Are the categories unordered? → Horizontal bar chart (the default)
- Do they have a natural order (age groups, months)? → Column chart
- Are you comparing across two dimensions? → Clustered bar or column chart
- Are the values close together? → Lollipop or dot chart
- Are there many series that would overlap? → Small multiples
Showing parts of a whole? A pie for a few parts, otherwise a stacked bar
- Are there only 2–5 parts? → Pie or donut chart
- Are there more parts, or are you comparing across categories? → Stacked bar chart
- Are the parts hierarchical? → Treemap
Showing a distribution? Depends on what matters most
- Is it the frequency of values? → Histogram
- Is it the spread, median and outliers? → Box plot
- Is it the overall shape? → Density or violin plot
Showing a relationship? Usually a scatter plot
- Two numeric variables? → Scatter plot
- Three numeric variables? → Bubble chart
- A categorical and a numeric variable? → Heatmap
When in doubt, use a horizontal bar chart. It handles long labels, reads naturally from top to bottom, and shows values as position along a common baseline, which people read most accurately.
Rules for every chart
Remove the clutter. Charts have no outline, no borders around bars or slices, no 3D effects, no shadows and no background fill.
Label directly. Put labels on or next to the data, and use a legend only when direct labels don’t fit. Once every value is labelled, drop the value axis and gridlines.
Order categories meaningfully: by value, in their natural order, or with the key category first. Never sort alphabetically unless readers need to look items up.
Highlight with grey. See the highlight pattern.
One value axis per chart. Dual y-axes let the maker set the apparent relationship by choosing the scales. Use two panels side by side, or index both series to a common start.
Show uncertainty where it changes the message: error bars or ranges on estimates, and a dashed or shaded line for forecasts. Explain the measure in a note, such as “95% confidence interval”.
Group categories into “Other”, use small multiples, or switch to another chart type if you visual is too crowded to be legible.
Chart chooser sources: Atlassian, Datawrapper (2025), Highcharts Chartchooser and the Data Visualisation Catalogue. See references.