Bubble Charts

A bubble chart is a scatter plot where each dot's size encodes a third quantitative variable. It's not really a different chart type — it's a scatter plot with an extra trick — but it's so useful that it gets its own page.

The pitch

If a scatter plot uses x and y for two variables, a bubble chart gives you three variables in one picture: x, y, and bubble size. Add color and you have four. Add facet_col and you have five. This information density is the bubble chart's superpower.

The most famous bubble chart is Hans Rosling's Gapminder animation — life expectancy vs GDP per capita, with population as bubble size and continent as color, animated over time.

Your first bubble chart

Four variables shown at once, and it's still readable. Hover over the big bubbles — China and India dominate the view because size="pop".

size_max controls the radius of the largest bubble in pixels. Without it, very-populous countries can swallow the chart.

Why size is a weaker encoding than position

The Cleveland-McGill ranking puts area near the bottom of encoding accuracy. The eye can perceive "bigger" and "smaller" but cannot reliably tell you "this is 2.3× that." So:

• Reserve bubble size for a variable where the reader only needs to notice "this is big" or "this is small." Population works perfectly — nobody is doing arithmetic with bubble sizes.
• Use x and y for the variables where precise comparison matters.

Linear vs square-root size scaling

By default, Plotly Express scales bubble area in proportion to the value — which is correct: when a value doubles, the area doubles, and that's roughly what the eye reads. Do not be tempted to scale by radius — it makes bubbles look 4× bigger when the value doubles, which lies.

If your population spans many orders of magnitude (e.g., 1,000 to 1,000,000,000), even area-proportional bubbles can become extreme. Solutions:

• Take np.log of the value before passing to size=.
• Filter out the extreme tails.
• Switch to a different encoding (color intensity).

Animated bubble charts: Hans Rosling's classic

Plotly Express makes the animated Gapminder essentially free:

Press the play button below the chart. You're watching 50 years of global development in 30 seconds. This is Hans Rosling's chart, in three Python lines.

Two arguments make the animation work:

• animation_frame="year" — each year is a frame.
• animation_group="country" — connect each country's dot across frames so it morphs in place instead of jumping.

A handful of range_x / range_y arguments keep the axes fixed across frames, so the eye reads change without re-anchoring.

When animation goes wrong

Animation is powerful but expensive — viewers can't pause and study a frame while it's moving. A few rules:

• Always provide a slider (Plotly Express does this by default).
• Use animation when time is the third dimension — never animate just because you can.
• For a printed report or a slide deck, make a static "small multiples" version with one panel per year as a backup.

A bubble chart that doesn't work

A bubble chart fails when:

• The size variable's range is too narrow to perceive (every bubble looks the same).
• There are too many bubbles and they overlap completely.
• The size variable is negative (negative areas don't exist).

In those cases, prefer color or a different encoding.

Check your understanding

QuestionSelect one

A bubble chart is essentially:

A pie chart with multiple values per slice.

A 3-D scatter plot.

A scatter plot where marker size encodes a third quantitative variable.

A type of histogram.

QuestionSelect one

Why should you usually pick less critical variables for the size encoding in a bubble chart?

Size is calculated last by the rendering engine.

Size cannot be larger than 50 pixels.

Area is a relatively weak encoding — the eye perceives "bigger / smaller" categorically but not precise multiples. Reserve size for variables where rough comparison is good enough.

Size is hidden from screen readers.

QuestionSelect one

Which Plotly Express argument turns a bubble chart into an animated Gapminder-style chart?

animation=True

play=True

animation_frame="..." (and usually animation_group="...")

time=True