Social Network Analysis (SNA) is among the most trending topics in People Analytics. HR analytics professionals are now starting to conduct SNAs to unearth collaboration, detect silos, and identify potential leavers.
Even though SNA is only now starting to become a standard tool in People Analytics, advanced network and graph theory can already provide various theoretical insight that can inform SNA and demonstrate how important it is to understand organizations as what they really are: an interconnected accumulation of social groups. In this article, I will argue that People Analytics teams should pay more attention to the dynamics and quality of social networks. Real-world networks are never static and so our analyses of them shouldn’t be either. I will address three streams of theory that provide a new approach to analyze organizations through the lens of network theory.
Dynamic Network Theory
Only very few networks in nature are static, however most conducted SNA create networks that only visualize the current state of interconnectedness within the organization. This single snapshot of the organization can already provide lots of insight about numerous metrics, such as the network’s density or its number of distinct clusters. However, these analyses cannot tell the larger and perhaps more important story of how they evolved.
Actively comparing two states of the same network brings about a whole new dimension of network metrics and parameters. The insight which nodes (employees) of a social network produce more connections than others and how new nodes become integrated within the network can provide valuable insight for any people analytics team. Helpful for practitioners is the knowledge about critical thresholds in the development of (social)-networks. When networks grow (or shrink), a certain amount of connections and members is critical for the establishment of its effectivity.
The underlying idea, the so-called network effect states that the relationship of nodes (employees) to edges (connections) is not linear, but exponential. As the number of connections in a network grows by the function:
A network with 5 members can only have 10 possible connections:
The addition of just 4 more members allows the organization to build up to a network of already 36 connections. For a company, this means that despite a linear headcount growth its interconnectedness can grow exponentially!
The critical threshold that enables efficient collaboration and robustness with n members of a network is the average number of connections of log(n). At this threshold of average connections, the members typically start to be at least indirectly connected to each other.
How networks grow
The likelihood of new edges within a social network can also be seen as a statistical probability, given the characteristics of members that are already connected in the network. Knowledge about these characteristics can be used to identify patterns and derive rules of preferential attachments in your organization.
Typically social networks form under the influence of homophily, which describes the tendency of connections to establish between similar members of a network (McPherson, Smith-Lovin & Cook, 2001). If a SNA e.g. reveals that employees of the same age, or the same educational background tend to stick together, dynamic network analysis allows to calculate the probability of how likely it is that any two members will form a new link, given their educational and age attributes. Homophily can actually reveal itself as a driver of silo constitution. When trying to enhance diversity within an organization, such analysis could be an insightful starting point.
A good theory to explain how (social) networks grow is the percolation theory, which describes how connections in any network develop through the paths of least resistance: the network then becomes the product of the resource constraints the environment places on it (Newman, & Watts, 1999).
Put differently: The network is shaped and limited by the environment’s characteristics. This can be observed in companies that occupy buildings with multiple floors. Social connections are usually more likely to develop between coworkers on the same floor. The social network is then showing numerous clusters of coworkers on identical floors. In People Analytics, understanding which resource constraints are influencing the development and topology of the organizational network can be tremendously important and, for example, inform organizational development projects.
Robustness of Social Networks
Networks cannot only differ in their size or density but also on a dimension that is not as easy to spot on a first glimpse: its structural robustness. The robustness of a network describes its ability to remain functional after it loses members, or connections within the network dissolve. For People Analytics teams the level of robustness can be a good indicator how vulnerable the organization is to potential turnover.
A key figure that correlates with robustness is
- In a
high density networkthe reduction of one edge is not damaging its overall structural integrity: the network is able to compensate through two other direct connections
- In a low density network the reduction of one edge can disconnect two parts of the network that can only be bridged by a longer chain of indirect connections
The measure of network closure describes the extent to which two connected people have one or more mutual connections – in the given example above, this closed relational triad on the left side compensates for the loss of the connection – the network on the right side fails to do that.
When thinking about the reduction of nodes or edges, the degree of betweenness of a social network is crucial to understand its robustness. It describes the amount of critical bridges in a network that connect two clusters and which removal can thus cause vast damage to the average distance between nodes within the network. Understanding and fostering the betweenness, density and connectivity of a network should thus be the aim of every People Analytics team, conducting SNA.
The ambivalence of high density networks
Strong ties and bonds between members are preferential in social networks. In People Analytics they often correlate with employee satisfaction and innovation (Hulbert, 1991). Therefore, highly interconnected organizations are often seen as desirable.
But what makes high density networks so powerful can also reveal itself as a weakness: high density does not only empower the collective intelligence of an organization, but can also accelerate the spread of negativity or low levels of engagement. Toxicity has in fact been found to be contagious within organizations (Dimmock, & Gerken, 2018). This is aligned with findings in biology, where high density networks are more prone to be disrupted through virus infections.
Furthermore, the density of organizational networks does only partly correlate with performance: every manager knows – too many stakeholders and feedback loops can also hinder effective work streams. In People Analytics the density of organizational networks should therefore be seen as both, a chance for higher collaboration and diversity but also as a risk for the collective spread of low engagement and toxicity. As per usual, the truth lies somewhere in between, which brings me to my last argument.
Why weak ties matter
What organizational design (and its analysis) should aim for, is a network with links of different quality: Teams, tribes
In these cases, weak connections to members of a different cluster that are only held by one or two members of a clique are extremely important as they can contribute different perspectives and constitute bridges between clusters. In short, weak ties can contribute fundamentally to the diversity of a network, while not putting it at risk of inefficiency or turning it into a highly contagious environment.
People Analytics or SNA can help us to better understand this diversity and the quality of collaboration and connectedness within an organization.
Many concepts, ideas