Parallel Sets
A multi-dimensional categorical visualization that uses ribbons between parallel axes to show how categories co-occur and combine across dimensions.
// 01 — The chart
What it looks like
A parallel sets diagram showing Titanic passengers classified by ticket class, gender, and survival. Ribbon width encodes passenger count.
// 02 — Definition
What are parallel sets?
Parallel sets (sometimes called “parallel categories”) are a visualization technique for exploring relationships between multiple categorical variables. Each horizontal axis represents one dimension, and ribbons flow between axes showing how categories combine.
Think of a contingency table made visual. Instead of reading rows and columns of numbers, you see colored ribbons whose width encodes the frequency of each category combination. This makes it easy to spot which combinations are common and which are rare.
The technique was introduced by Robert Kosara, Fabian Bendix, and Helwig Hauser in 2006 as a way to make categorical data exploration more intuitive than tables alone. Unlike parallel coordinates (which handle continuous data), parallel sets are purpose-built for categorical dimensions.
Interactive origins: Parallel sets were originally designed as an interactive tool — users could brush and filter categories to explore subsets. Static versions lose some power but still excel at showing the overall distribution of category combinations.
// 03 — Anatomy
Parts of a parallel sets diagram
// 04 — Usage
When to use it — and when not to
- Exploring how 2-5 categorical variables relate to each other
- You want to see which category combinations dominate the dataset
- Replacing large contingency tables with a visual overview
- Investigating survey data where every question has categorical answers
- Showing how a population is segmented across multiple dimensions
- Interactive exploration of categorical data subsets
- Your variables are continuous — use parallel coordinates instead
- You have more than 6 dimensions — the ribbons become unreadable
- Each dimension has too many categories (more than ~8)
- You need to show temporal change — use an alluvial diagram instead
- Exact percentages are critical — tables or bar charts are more precise
- The static version alone must tell the story — the format works best interactively
// 05 — Reading guide
How to read a parallel sets diagram
Follow these steps whenever you encounter parallel sets in the wild.
Identify the dimensions
Each horizontal axis represents a categorical variable. Read the labels to understand what dimensions are being compared (e.g. class, gender, survival).
Read the category segments
On each axis, the bar is divided into colored segments. Each segment represents one category, and its width is proportional to the number of observations in that category.
Trace the ribbons
The ribbons flowing between axes show how observations are distributed across category combinations. A wide ribbon means many observations share that combination.
Compare ribbon widths
Look for the thickest ribbons — these represent the most common category combinations. Thin ribbons indicate rare combinations.
Look for patterns of association
If ribbons from one category mostly flow to a single category on the next axis, those variables are strongly associated. If ribbons spread evenly, they are independent.
// 06 — Pitfalls
Common mistakes
Too many dimensions
Limit to 3-5 dimensions. Beyond that, the overlapping ribbons create visual chaos. Prioritize the most important categorical variables.
Unclear dimension ordering
Place the most important or explanatory dimension first. The axis order dramatically affects readability — experiment with different orderings.
Ignoring ribbon color consistency
Use consistent coloring across all ribbons that originate from the same category. This lets readers trace a single group's path across all dimensions.
Missing category labels
Every segment on every axis needs a label. Without them, the diagram is a pretty but meaningless pattern of ribbons.
Confusing with parallel coordinates
Parallel coordinates use continuous axes with polylines. Parallel sets use categorical axes with ribbons. Don't mix the two concepts.
// 07 — In the wild
Real-world examples
Titanic survival analysis
The classic dataset — parallel sets reveal how class, gender, and age interacted to determine survival, making patterns visible that tables hide.
Survey response patterns
Market researchers use parallel sets to visualize how demographics (age, income, region) relate to product preferences, brand loyalty, and purchase frequency.
Clinical trial outcomes
Medical researchers display how patient characteristics (treatment group, disease severity, biomarker status) relate to treatment outcomes across multiple dimensions.
// 08 — Quick reference
Key facts
Also known as
Parallel categories, categorical parallel coordinates
Data type
Multiple categorical variables
Best for
Cross-tabulation, category co-occurrence, survey analysis
Audience level
Intermediate — works best with interaction
Dimension limit
3–5 dimensions for clarity
Related to
Alluvial diagram, Sankey diagram, parallel coordinates
// 09 — Variations
Variations and extensions
Interactive parallel sets
The original design — users can click, brush, and reorder dimensions to explore subsets of the data dynamically.
Vertical parallel sets
Axes run vertically instead of horizontally, with ribbons flowing left to right. Useful when dimension labels are long.
Weighted parallel sets
Ribbon width encodes an additional numeric measure (e.g. revenue) rather than simple frequency counts.
// 10 — FAQs
Frequently asked questions
What is a parallel sets?+
Parallel sets (sometimes called "parallel categories") are a visualization technique for exploring relationships between multiple categorical variables. Each horizontal axis represents one dimension, and ribbons flow between axes showing how categories combine.
When should you use a parallel sets?+
Use a parallel sets when exploring how 2-5 categorical variables relate to each other. It also works well when you want to see which category combinations dominate the dataset, and when replacing large contingency tables with a visual overview.
When should you avoid a parallel sets?+
Avoid a parallel sets when your variables are continuous — use parallel coordinates instead. It is also a poor fit when you have more than 6 dimensions — the ribbons become unreadable, or when each dimension has too many categories (more than ~8).
What is another name for a parallel sets?+
Parallel Sets is also known as Parallel categories, categorical parallel coordinates. The name varies between fields, but the visualisation technique is the same.
What size of dataset works best for a parallel sets?+
Parallel Sets works best for Cross-tabulation, category co-occurrence, survey analysis. Outside that range the chart either looks empty or becomes too cluttered to read clearly.
Is a parallel sets suitable for dashboards?+
Yes — a parallel sets can work well in dashboards as long as the panel is large enough for readers to perceive the encoded values, has a clear title, and includes the legend or axis labels needed to interpret it.