80 types of charts for data visualization. Part 2: Correlation Charts
by Ivan Kilin — Data Visualization Specialist & Dieuwertje van Dijk — Data Visualization Designer @
Recently we’ve rewritten our essential article about 80 types of charts. Our main goal was to make the information structured and helpful right away. We talk about every chart starting from the look and feel and then going into practical advice on when you should and when you shouldn’t use this particular chart type. We will start publishing parts of this article here. One chart group per each article.
In the second part, we’ll talk about correlation graphs. Correlation graphs are used to explore relationships between two sets of data. They allow to see if changes in one variable are linked to changes in another variable. Correlation graphs are great to identify potential connections in the data. It’s also important to remember that correlation graphs show trends but not causation. The go to option for visualizing correlation are scatter plot and bubble chart.
Let’s explore correlation graphs one by one. Keep on reading!
Heatmap
Alternative names: Heat map, Heat table, Density table
What is a heatmap?
A heatmap shows values as colors in a grid or as a colored layer on a map. The palette makes patterns, trends, correlations, and density differences instantly visible. Adding a date or time scale to the x-axis lets you track how values change over time, helping you spot anomalies. In a matrix layout, it reveals correlations. On a geographic base, it highlights where a value is more or less concentrated.
When should you use a heatmap?
Heatmaps are most helpful when you need an initial view of the data and want to let the colors guide you. Using color is a very natural way of spotting patterns, trends, and correlations. So a heatmap quickly tells you where to look next: for outliers, other viewpoints, or specific angles. Typical situations include:
Exploring wide‑ranging data to see overall structure before deeper analysis
Monitoring changes over time by adding a temporal x‑axis
Showing spatial density on a geographic base layer
Revealing correlations in a matrix of two categorical variables
Focusing on the evolution of a single variable with a one‑dimensional heatmap (e.g., temperature variations)
When you shouldn’t use a heatmap
A heatmap is primarily an exploratory technique and less suited for precise comparison. If your audience needs to measure exactly how much one value differs from another, color alone can be ambiguous. Subtle differences may fade, and extreme values can dominate the palette. In such cases, choose a chart that expresses magnitude more explicitly with position or length. For example, with parallel coordinates, a multi‑series line or bump chart, or a bubble chart.
Check our heatmap resource page and discover pro tips on how to design the best heatmap chart yourself.
Bubble chart
Alternative name: Bubble plot
What is a bubble chart?
Deriving from a scatter plot, a bubble chart is a chart that looks at the relationship between three or more (numeric) variables. Two of those variables are represented by the y-axis and x-axis. The third value is represented by the size of a bubble.
With some expansions, a bubble chart can represent up to seven variables at once. You can add variables to color, stroke, icon, or label, for example. However, keeping the total to three or four variables keeps the chart readable. Analysts value bubble charts for spotting correlations, clusters, and outliers, and for their strong data‑to‑space ratio.
When should you use a bubble chart?
Use a bubble chart to explore relationships among several numeric variables, highlight clusters or extreme points, and keep the visual compact. Its ability to show multiple dimensions makes it popular for research dashboards and infographics. It also works in any setting where space is limited, but a richer view than a simple scatter plot is required.
When you shouldn’t use a bubble chart
Bubble charts lose effectiveness if you overload them with too many bound variables, or if the range of the size variable is so narrow that bubbles look alike. Dense overlaps can also hide information unless you tweak the opacity. In these cases, consider a simpler alternative such as a scatter plot, heatmap, or dot plot that focuses on fewer dimensions.
Refer to our bubble chart resource page and discover pro tips on how to design the best bubble chart yourself.
Scatter plot
Alternative names: Scatterplot, Scatter chart, Scattergram, Scatter diagram, Scatter graph
What is a scatter plot?
A scatter plot shows two numerical variables by placing dots at their X and Y coordinates. With every mark the same size, you get a clear view of how each observation sits on the two‑variable plane.
When should you use a scatter plot?
Choose a scatter plot when you need a quick look at possible correlations, patterns, clusters, or outliers. Its strong data‑to‑space ratio and simple structure make it the usual first step in data exploration. Because the layout is so flexible, a scatter plot can evolve into bubble charts, categorical scatter plots, quadrant charts, hexagon binning, or other hybrids.
When you shouldn’t use a scatter plot
Skip a scatter plot if dense overlaps make either the dots or their data labels hard to read. The same applies when readers need precise value‑by‑value comparisons rather than overall trends. They are also not ideal when variables are categorical or already grouped into ranges. In those cases, a categorical scatter plot, heatmap, or dot plot will convey the information more clearly.
For more design tips, visit our scatter plot resource page.
Connected scatter plot
What is a connected scatter plot?
A connected scatter plot begins exactly like a regular scatter plot: two numerical variables go on the x‑axis and y‑axis, and each dot represents one observation. After plotting the dots, you draw a line that links them in sequence. The line links the points in order. However, each point stays visible, so you can read both the individual observations and the overall trend.
When should you use a connected scatter plot?
Choose a connected scatter plot only when you have two numerical variables that change in a clear sequence, such as year‑by‑year data. The line shows how one variable moves as the other shifts, letting you highlight dramatic moments. Keep the point count low and the order unmistakable. If a standard line chart or a pair of separate lines can show the story just as well, use those simpler options instead.
When you shouldn’t use a connected scatter plot
Most readers find connected scatter plots hard to interpret unless you add plenty of explanatory annotations. If the same story can be told with a regular line chart, multiple lines, or a plain scatter plot, choose those simpler options instead. Also, avoid the connected version when the data has no clear order or when a large number of points would make the path cluttered and misleading.
Hexagonal binning
Alternative names: hexagonal plot, hexagonal bin plot
What is hexagonal binning?
Hexagonal binning is a heatmap of point density made with hexagons instead of squares. Lay a honeycomb over the x–y plane. Each cell counts how many observations fall inside, and the cell’s color shows that count (or a normalized density). Darker means more, lighter means fewer. Hexagons have six equal neighbors, which helps reduce grid artifacts and produces a smoother pattern. This makes clusters and gaps easier to see in very dense data.
When should you use hexagonal binning?
Use a hexagonal binning chart as a replacement for a crowded scatter plot. When thousands of points overlap, hexbinning reveals where data concentrate and where gaps exist without the noise of over‑plotting.
When you shouldn’t use hexagonal binning
Think twice if your audience is not used to reading hex‑bin charts. Without a clear legend that links color to counts, many readers will be unsure how to interpret the picture. Because binning replaces points with aggregated counts, it hides exact values and can mask rare outliers. So this makes it a poor choice for small samples or analyses that depend on precise numbers. Also, the result is very sensitive to bin size. If you set the hexagons too small, the display looks noisy. If you set them too large, meaningful detail disappears.
Creating a good hex‑bin plot also takes more skill and software support than drawing a standard scatter plot. You need a tool that supports hexagonal tessellation, and you need experience to choose an appropriate bin width and color scale. If a simpler chart, such as a plain scatter plot, can tell the story clearly, it is usually the better option.
Contour plot
What is a contour plot?
A contour plot shows three numerical variables on a flat page. Two variables form the x‑axis and y‑axis, just like a scatter plot. The third variable appears as curved lines that link every point with the same value. Picture a topographic map: one line marks 200 m above sea level, the next marks 300 m, and so on. Wherever the lines crowd together, the value changes quickly, and where they spread apart the surface is flatter. If you fill the spaces between the lines with colour, the plot reads like a smooth relief map of the data.
When should you use a contour plot?
Contour plots shine when you need to show a continuous surface. For example, elevation on a landscape or temperature across a weather map. Their flowing lines help readers see gradients quickly. This is why they are common in cartography, meteorology, oceanography, and other scientific fields that map smooth phenomena such as density, pressure, or magnetic strength. However, they can also suit business or retail data if you want to highlight dense clusters without clutter.
When you shouldn’t use a contour plot
Avoid a contour plot if the audience is not used to reading contour lines. It takes practice to translate the spacing of the lines into a sense of magnitude. The method also assumes that the data vary smoothly. If the surface is patchy or if there are sharp jumps, the interpolated lines can be misleading. Contour plots require enough data to create reliable interpolation, so they are not suitable for small or widely spaced samples.
Finally, if a simpler chart such as a heatmap can communicate the story clearly, prefer the simpler option for ease of interpretation.
That’s it for correlation charts. In the next part we’ll explore part-to-whole & hierarchical charts. See you soon!
You can also check all other parts here:
Part 1: Comparison charts
Part 3: Part-to-whole charts
Part 4: Data-over-time charts
Part 5: Distribution charts
Part 6: Geospatial and other charts
