When a radar chart is the right choice, when another chart reads better, and how to draw one that tells the truth. The same rules hold in every chart library. For picking between chart types in general, see Choosing a chart.
What it is for
A radar chart, or spider chart, shows a profile: several measures of one item, each on its own spoke, joined into a closed shape. It answers whether the item is balanced or strong in a few places, and how two profiles differ, such as two products scored on the same criteria.
When to use
Three to about eight measures that share one scale, or that you rescale to one.
One to three profiles, compared by their overall shape.
The reader asks where an item is strong and where it is weak, more than by how much.
Use something else when
The reader needs exact values or close comparisons. Use a grouped bar chart or a table.
There are more than three profiles. Use small multiples, one radar per item, or a heatmap with items as rows.
The measures have different units and no fair way to share one scale. Use a table, or one bar chart per measure.
The spokes are points in time and the trend matters. Use a line chart.
You rank items on one measure. Use a sorted bar chart.
There are more than about ten measures. The labels collide and the shape turns into noise. Use a table or a heatmap.
How it reads
Each value is a distance from the center along its spoke. The reader does not compare distances, though: they see the shape and its area. The area depends on how neighboring values combine, so the same data in a different spoke order draws a different shape and a different area.
Area also grows faster than the values. A profile that is 20% higher on every measure covers about 44% more area, so a larger shape overstates the gap.
Do it well
Put every spoke on the same scale, and say which one, such as 0 to 100 or the share of the best result.
Make outward mean better on every spoke. Invert measures where lower is better, such as a race time, and say so.
Start every spoke at zero in the center.
Choose the spoke order once, with related measures next to each other, and keep it the same in every radar you compare.
Draw at most two or three profiles. Use light, translucent fills or outlines only, so one shape does not hide another.
Label the spokes outside the circle and give a few rings their values.
Pitfalls
Judging profiles by area. Area changes with spoke order and grows faster than the values. Compare spoke by spoke.
Spokes in different units on one scale. A measure in thousands flattens one in tens. Rescale every measure first.
Reordering spokes between charts. The same data looks like a different profile.
An opaque fill over another series. The series underneath disappears. Lower the fill opacity or draw outlines.
Many overlapping profiles. Past three, the shapes blur into one. Split them into small multiples.
Variants
Filled: one profile as a translucent shape.
Overlaid: two or three profiles on the same spokes.
Outlines only: no fill, for overlapping profiles that must stay readable.
With dots: a marker at each value, so single values stand out.
Polygon or circle grid: straight rings that match the shape, or round rings that do not suggest straight lines between spokes.
Compared with similar charts
Against a grouped bar chart: the same values, read accurately on a shared baseline. The radar shows the overall shape at a glance.
Against a table: a table gives exact values for any number of items. The radar gives a quick profile for a few.
Against parallel coordinates: the same idea with straight, parallel axes, which holds many more items.
Against a polar line: a polar line follows one measure around a cycle. A radar puts a different measure on each spoke.