Kumu Blog

Tools and practices for tackling complex systems.

Weighted SNA metrics

Social network analysis metrics can be incredibly powerful for spotting hidden influencers, bottlenecks, and leverage points. But when your connections aren't binary and instead include attributes like strength or frequency, unweighted calculations just don't cut it.

With weighted metrics, you can now include attributes like strength and frequency in the calculations. Certain metrics (betweenness, closeness and degree) use connection attributes for weighting while others (size and reach) use element attributes for weighting.

Here's a quick example of the value of weighted metrics:

Not all connections are created equal

It's often said, it's not what you know but who you know. Even more so, it's who you know and how strong or weak your relationships are with them, and the strength/weakness of the connections they have that you can leverage.

We'll use an anonymized version of a real network where connections are either weak, medium or strong (we've assigned numerical values to each of the three categories: weak = 1, medium = 2 and strong = 3). If we run betweenness without weighting, we artificially inflate the betweenness of some individuals by treating the weak ties the same as the strong ties. Here's the network scaled by betweeness, with generic labels assigned (A, B, C, etc.) based on the six highest scoring elements:

We can re-run the betweenness metric but instead choose to weight based on the strength of connections. Here's the result:

You'll see that when we account for the strength of connections, C scores the highest for betweenness. Remember, betweenness is a measure of the number of times an element lies on the shortest path between any two elements. In the weighted case, we're giving more credit to the shortest paths that are made of stronger connections.

Run your first weighted metrics

For the metrics that allow weighting, you'll see an advanced options link once you select the metric:

You'll be able to choose any numerical attribute for the weighting, so if you don't see the value you want to use show up, make sure you've changed the attribute type to numeric.

Give weighted metrics a spin and let us know what you think!

Subscribe for updates

Loading comments...