A scatter plot is a visualization that places each data point at the intersection of two numeric values, revealing relationships, clusters, and outliers between the two variables. Unlike most charts, it doesn’t aggregate first and plot second — every dot is an individual record (or group), which is exactly what makes it good at showing the shape of a relationship rather than a single summary of it.
When to use a scatter plot
Reach for a scatter plot when the question is “how do these two numbers relate?”
- Correlation. Discount offered vs. order size, ad spend vs. conversions, deal size vs. days to close. If the cloud of points slopes upward, the variables move together; if it’s a shapeless blob, they don’t. (A visible relationship still isn’t proof of causation — the scatter shows association, not mechanism.)
- Clusters and segments. Customers plotted by order frequency and average order value often separate into visible groups worth treating differently.
- Outliers. A point far from the cloud — the rep with huge pipeline and no closes, the page with massive traffic and zero conversions — is often the most interesting row in the dataset, and a scatter plot is the fastest way to find it.
Adding a third variable as dot size turns a scatter plot into a bubble chart — deal size vs. close time, with bubbles sized by contract value.
When not to use one
The classic misuse is plotting a time series as a scatter. If the x-axis is time and there’s one value per period, connect the dots — a line chart shows the trend you actually care about. Scatter plots earn their keep when both axes are measures, not when one of them is a sequence.
They also need enough points to show a shape: with a handful of dots there’s no pattern to read, and with one categorical variable you want a bar chart instead. At the other extreme, thousands of overlapping points turn into an unreadable smear — consider aggregating, sampling, or switching to a density view. And if you only care about the distribution of a single numeric variable, a histogram is the right tool.
Scatter plots in Metabase
Metabase has a native scatter chart type. Build a question with two numeric columns — either raw fields or aggregations grouped by some dimension — and pick the scatter visualization; you choose which column goes on each axis in the settings. Add a third numeric column and Metabase can use it to size the dots, giving you a bubble chart without changing chart types.
Like other Metabase charts, scatters support drill-through: click an interesting outlier to see the records behind it. On a dashboard, a scatter responds to the same filters as its neighbors, so you can scan the relationship for one segment at a time. See the scatterplot and bubble chart docs for details.
Related terms
Further reading
Put it to work
- Conversion rate — Metric
- Sales analytics — Overview
- Sales pipeline dashboard — Dashboard
