Force-Directed Edge Bundling
A technique that uses simulated physical forces to attract nearby edges into smooth bundles — reducing clutter in dense network graphs without requiring any hierarchical structure.
// 01 — The chart
What it looks like
A force-directed edge bundling diagram showing airline routes between US cities. Edges naturally bundle along major corridors (e.g., the west-to-east coastal route via Denver), revealing the network’s main traffic arteries.
// 02 — Definition
What is force-directed edge bundling?
Force-directed edge bundling (FDEB) is a technique for reducing visual clutter in network graphs by bending nearby edges toward each other so they merge into smooth bundles. Unlike hierarchical edge bundling, which requires a tree structure to route edges through, FDEB works on any graph — it only needs nodes and edges.
The algorithm treats each edge as a flexible spring divided into multiple control points. Neighboring edges exert attraction forces on each other’s control points: if two edges run roughly in parallel, their midpoints are pulled together. After several iterations of this force simulation, edges that share a similar trajectory converge into visible bundles, while dissimilar edges remain unbundled.
The result is a visualization where the “highways” of the network become immediately apparent. Dense traffic corridors, shared pathways, and major arteries stand out as thick bundles, while isolated or unusual connections remain as thin individual lines. This makes it ideal for geographic networks (flight routes, trade flows) and large-scale communication graphs.
Origin: Force-directed edge bundling was introduced by Danny Holten and Jarke J. van Wijk in 2009 in their paper “Force-Directed Edge Bundling for Graph Visualization.” It extended the edge bundling concept to work without hierarchies, making it applicable to a much broader class of networks.
// 03 — Anatomy
Parts of a force-directed edge bundling diagram
// 04 — Usage
When to use it — and when not to
- Your network has no hierarchy but many edges create a tangled hairball
- You want to reveal traffic corridors, migration routes, or communication pathways
- Node positions are fixed (e.g., geographic coordinates on a map)
- The network is too dense for a standard node-link diagram to be readable
- You need to identify major flow arteries across a geographic or abstract space
- Presenting an overview where individual edge precision is less important than overall pattern
- You need to trace individual connections precisely — bundling obscures exact edge paths
- The network is small (under 30 edges) — unbundled edges will be clearer and simpler
- Real-time performance matters — the force simulation can be computationally expensive for very large graphs
- Edge weights or directions must be precisely communicated — bundling merges visual cues
- Your audience needs to count exact connections between specific node pairs
- The graph is very sparse — there are too few nearby edges to create meaningful bundles
// 05 — Reading guide
How to read a force-directed edge bundling diagram
Follow these steps to interpret a force-directed edge bundling visualization.
Identify the thickest bundles first
The thickest visual bundles represent corridors where many edges run in parallel. These are the dominant pathways in your network — the main highways. In a flight network, these might be transcontinental corridors; in a communication graph, they are the busiest channels.
Trace bundles from source to destination
Follow a thick bundle to see which groups of nodes it connects at each end. The endpoints tell you which regions or clusters have the strongest mutual connections. Unlike hierarchical bundling, these groups are defined by spatial proximity, not organizational structure.
Look for unbundled outlier edges
Individual edges that don’t merge into any bundle represent unusual or unique connections. These might be the most interesting findings — anomalous routes, unexpected communications, or rare pathways that deserve investigation.
Compare bundle density across regions
Some areas of the visualization will have dense bundles while others will be sparse. Dense regions indicate high connectivity; sparse regions indicate isolation. This geographic or spatial distribution of connectivity is one of the key insights FDEB provides.
Use interaction to explore details
Because bundling inherently trades detail for overview, most useful FDEB visualizations support hovering over nodes to highlight their specific connections. Use this interaction to drill down from the overall pattern to specific node-level connections.
// 06 — Common mistakes
Mistakes to watch out for
Over-bundling that creates false corridors
If the attraction forces are too strong, edges that are only vaguely parallel will be pulled together into artificial bundles. This creates the visual impression of a major corridor that doesn’t actually exist in the data. Calibrate the force parameters carefully and compare the bundled result against the raw edge layout.
Ignoring computational cost
The force simulation involves iterating over every pair of edges multiple times. For graphs with thousands of edges, this can be extremely slow. Consider using GPU acceleration, multi-level approximation, or a simpler bundling method (like kernel density edge bundling) for very large networks.
Not enough control point subdivisions
Too few control points per edge produce angular, unrealistic bends instead of smooth curves. But too many control points increase computation time. A good starting point is 6–10 control points per edge, adjusting based on edge length and desired smoothness.
Using bundling on sparse graphs
If your graph has few edges, there simply aren’t enough nearby edges to create meaningful bundles. The result will look like a regular node-link diagram with unnecessarily curved edges. Save FDEB for networks where edge crossings are actually a problem.
Forgetting to show unbundled edges as reference
Without seeing the original unbundled layout, viewers may not understand what the bundles represent. Consider providing a toggle or side-by-side comparison so users can see the raw edges versus the bundled result, building trust in the transformation.
// 07 — Real-world examples
Where you’ll see force-directed edge bundling used
Transportation: Airline route maps
Airlines and aviation analysts use FDEB to visualize thousands of flight routes on a geographic map. Without bundling, routes between hundreds of airports create an unreadable tangle. With bundling, major transcontinental corridors and hub-spoke patterns emerge clearly, revealing the backbone of the air transport network.
TransportationTelecommunications: Internet backbone traffic
Network engineers visualize data flows between data centers and exchange points using FDEB. The bundled edges reveal the main traffic arteries of the internet, helping identify bottlenecks and plan capacity expansions. Geographic positioning of nodes makes the physical infrastructure immediately apparent.
TelecommunicationsUrban planning: Commuter flow analysis
City planners map commuter journeys between neighborhoods using FDEB overlaid on a city map. Dense bundles reveal the most-traveled corridors, directly informing decisions about where to add public transit, widen roads, or build bike infrastructure. The technique handles thousands of origin-destination pairs gracefully.
Urban Planning// 08 — At a glance
Quick reference
// 09 — Variations
Edge bundling approaches
Force-directed edge bundling is one of several approaches to reducing edge clutter. Each has different trade-offs.
Hierarchical edge bundling
Routes edges through a tree hierarchy. Produces beautiful radial layouts but requires hierarchical data structure.
Kernel density edge bundling
Uses kernel density estimation to identify edge corridors. Faster than force simulation and scales to very large graphs.
Winding roads bundling
An ink-minimization approach that creates smooth, road-like bundles. Produces aesthetically pleasing results for geographic data.
Confluent drawing
Uses junction nodes to merge edges through shared waypoints. Guarantees that each bundle encodes an actual graph property.
// 10 — FAQs
Frequently asked questions
What is a force-directed edge bundling?+
Force-directed edge bundling (FDEB) is a technique for reducing visual clutter in network graphs by bending nearby edges toward each other so they merge into smooth bundles. Unlike hierarchical edge bundling, which requires a tree structure to route edges through, FDEB works on any graph — it only needs nodes and edges.
When should you use a force-directed edge bundling?+
Use a force-directed edge bundling when your network has no hierarchy but many edges create a tangled hairball. It also works well when you want to reveal traffic corridors, migration routes, or communication pathways, and when node positions are fixed (e.g., geographic coordinates on a map).
When should you avoid a force-directed edge bundling?+
Avoid a force-directed edge bundling when you need to trace individual connections precisely — bundling obscures exact edge paths. It is also a poor fit when the network is small (under 30 edges) — unbundled edges will be clearer and simpler, or when real-time performance matters — the force simulation can be computationally expensive for very large graphs.
Is a force-directed edge bundling suitable for dashboards?+
Yes — a force-directed edge bundling 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.
What category of chart is a force-directed edge bundling?+
Force-Directed Edge Bundling belongs to the Network family of charts. Charts in that family are designed to answer the same kind of question, so they often work as alternatives when one doesn't quite fit your data.
How do you read a force-directed edge bundling?+
Start with the axis labels and legend, then look at the overall shape before zooming into individual marks. Compare prominent features against the rest of the data, and verify any conclusion against the underlying numbers when precision matters.