Scatter plot

When to use

• Exploring the relationship between two numerical variables (e.g., engagement rate vs. response volume by region)
• Identifying clusters or segments within a dataset (e.g., grouping consultation participants by sentiment and engagement level)
• Spotting outliers — individual data points that sit far from the main distribution
• Showing distribution across two dimensions when the dataset has enough points (20+) to form meaningful patterns

When not to use

• Don’t use for non-quantitative data — both axes must represent numerical values; categorical axes don’t work in a scatter plot
• Don’t use with very few data points — fewer than about 10 data points are better presented as a table
• Don’t use when the relationship between variables isn’t the question — if the goal is to compare category totals, use a bar chart
• Avoid for general audiences without context — scatter plots require more data literacy to interpret correctly; always include a clear explanatory caption

Anatomy

• Data points — Each dot represents one entity (a region, respondent group, time period). Dot size can encode a third variable (bubble chart variant).
• X-axis — The horizontal variable. Label clearly with units. Include a zero line where meaningful.
• Y-axis — The vertical variable. Label clearly with units.
• Trend line (optional) — A regression line showing the overall direction of correlation. Include only when the statistical relationship is meaningful to the audience.
• Tooltip — On hover or keyboard focus, show the data point’s label and both axis values.

Variants

Scatter plot

Standard form. Two quantitative axes, each point represents one record. Best for exploring relationships in a data-literate context.

Bubble chart

A scatter plot where dot size encodes a third quantitative variable. Use with caution — bubble area is difficult to compare precisely. Always label bubbles or provide a size legend.

Connected scatter plot

Data points are connected in sequence (usually time order). Shows how the relationship between two variables evolves. Rarely appropriate in consumer-facing contexts; use for analyst tools.

Accessibility

• Don’t rely on position alone — Individual data points in a dense scatter plot cannot be distinguished without interaction. Provide a data table that lists each entity and its x/y values.
• Colour for grouping — If colour encodes a categorical grouping, use the Orbit colour-blind-safe palette and also use shape to distinguish groups (circle, square, triangle). See Colour in charts.
• Keyboard access — Each data point should be focusable. In dense scatter plots with hundreds of points, provide a filterable data table as the primary accessible alternative rather than making every point individually focusable.
• Caption and summary — Always include a descriptive caption that states what the chart shows and what the main pattern is (e.g., “This scatter plot shows a positive correlation between participation rate and response volume across 32 regions”).