Datasmith.org

Experiments in data visualization

Chord diagrams are quite familiar these days. There are several available implementations that use hierarchical edge bundling to show a network of dependencies, but this example tool creates bold, simple chord diagrams to show the amount of connection between entities (in the example data, it’s a breakdown of email traffic between different roles in a university).

To make the diagram visually interesting and to try and give a visual impression of the ‘footprint’ of a given entity in the overall picture, gradients are used so that the chords partially take on the color of their associated entities.

Options:

Symmetrical relationships: If unchecked, the size of each entity's arc will be determined only by its outward connections, not by its total connections. Useful depending on the data format.

Color Scheme:

Arc Padding:

Label Size:

Radial Labels:

Drag and pinch to adjust view.

Data:

The data format is simply 'A,B,amount' to create a connection from A to B of size 'amount'. Other lines will be ignored.

The chord diagram is recent enough that it has a specific inventor — Martin Krzywinski, who popularized this diagram type in the Circos software package. Chord diagrams have lots of visual impact but I feel they have practical value only in fairly specific cases, for instance:

When the need is to assess the degree of similarity, or degree of interrelatedness, of two systems. This is common in biotech.
When the need is to quickly see what major connections dominate a system.

It’s not useful for several common business use cases such as process flow, so as a result, it tends to be seen more in scientific than in business contexts.

Of course, the open secret is, that any dependency graph or hierarchy can me made to look like a chord diagram just by plotting it with rotational rather than cartesial co-ordinates…!