Scatter chart: definition, examples, and best practices.

What is a scatter plot?

A scatter plot places each observation as a single point, positioned by its value on two variables, one on each axis. With enough points, patterns emerge that a table of numbers cannot show as clearly: a cluster of points moving diagonally suggests a relationship, a loose cloud suggests little relationship, and a handful of far-flung points mark the outliers.

The type is also called a scatter chart, an XY chart, or a correlation chart, all pointing at the same core idea of plotting two variables point by point.

A worked example

A revenue team plotting customer acquisition cost against lifetime value across around forty accounts might see a sample of points like this.

When to use a scatter plot

• Checking whether customer acquisition cost correlates with lifetime value across your account base.
• Identifying outlier accounts, employees or products that behave differently from the rest of the group.
• Testing whether two operational metrics, like tenure and output, move together before building a model around that assumption.
• Spotting clusters of similar observations that might warrant a different segment or a different strategy.

When not to use a scatter plot

• If you need to show a third variable through point size, use a bubble chart instead.
• If your data is a single trend over time rather than two independent variables, a line chart is clearer.
• If you have very few observations, fewer than about fifteen, a scatter plot will look sparse. A clustered bar chart often communicates a small dataset better.

How to read a scatter plot

Look at the overall shape of the point cloud before focusing on any single point. A tight diagonal band suggests a strong relationship between the two variables, while a wide, shapeless cloud suggests a weak one. A trend line, when added, summarises that relationship, but a relationship on a scatter plot is a correlation, not proof that one variable causes the other, and sample size matters: a handful of points can suggest a pattern that more data would erase.

Best practices

• State the sample size directly on or near the chart, since a scatter plot's credibility depends heavily on how many points it contains.
• Add a trend line when a relationship is genuinely present, but label it clearly as a trend, not a guarantee.
• Never claim causation from a scatter plot alone. Correlation shown on the chart does not establish which variable, if either, causes the other.
• Use consistent axis scales when comparing multiple scatter plots side by side, so the reader is not misled by different ranges.
• Highlight or label true outliers directly, since they often carry more of the story than the main cluster does.