Measuring Team Disruption.

Using Social Network Analysis To Measure and Minimise Disruption To Your Best Performing Teams.

This article was written by Mohammed Alagha from LSE, with the support of Riccardo Volpato, Laura Weis and Ted Lappas from Satalia.

Teams are the human capital that constitutes the social structure of organizations. People are allocated to teams based on the need for skills and requirements of projects, thus teams continuously change in size and composition. Consequently, disruption is an inherent property of team formation. For organizations operating in dynamic and fast-paced environments, understanding and quantifying disruption is both an objective and a challenge.

At Satalia, we use tools and resources from Social Network Analysis and Data Science to provide customized support for those navigating today’s organizational complexity. Together with the London School of Economics and Political Science, we decided to further develop and evaluate our tools. Here, we describe our joint work about quantifying team disruption.

Social Network Analysis

Social Networks Analysis (SNA) investigates social structures and their patterns through the use of networks and graph theory. People, projects and organizations are represented by nodes, while the interactions and relations between them are represented by links. Nodes have attributes that capture properties such as age, social status and location. Similarly, links carry attributes describing the underlying interactions such as communication between the nodes, authority, power and kinship. SNA spans a wide variety of applications, but here we focus on one interesting application, team formation, where members and their interactions constitute a social network.

Team Formation & Disruption

Bruce Tuckman describes team formation as going through four distinct stages: forming, storming, norming and performing. In each stage, disruption is inevitable and could be internal or external. For instance, removing members results from retirement or downsizing, while adding new members occurs by promotions or expansions. In any case, disruptions happen because people challenge the existing norms and dynamics of the team, old links break and new ones form. Disruption forces the team to stay in the storming stage which affects team functionality, harmony, efficiency and thereby performance.

Given that disruption takes different forms and arises from numerous sources, we only considered scenarios when members exit a team. More specifically, we looked at how to tackle two specific questions that arise when dealing with team disruption:

1. Who are the critical members that may disrupt the team when removed?
2. What is the effect on the team when certain members are removed?

Identifying Disruption

To unpack the first question, we need to define what a critical member is. As disruptions affect both the structure and the functions of a team network; a member can be critical to the connectivity (structural disruption) and/or the purpose (functional disruption) of a team.

Structural disruptions affect the topological connectivity of a team, which we determined by finding the minimum number of members that — when removed — break the network in more than one component. We refer to the members that this step identifies as the critical set. This set contains team members that hold the team as a single connected component. Robust teams have a critical set with a large size, so we can measure the robustness of a team as the ratio between the size of the critical set and the total number of members in the network.

Functional disruptions refer to the removal of members that have strategic positions and influential roles. To identify these members, we use the concepts of degree and betweenness centrality. Members with high degree centrality (also called hubs) are central members with many direct connections. In directed networks, hubs have high access to other members in terms of influence (out-degree) and support (in-degree). Analogously, members with high betweenness centrality (also called bridges) connect different parts of the team and influence the communication flow across the team.

The practical insight of identifying key-players in teams is that it allows us to get an x-ray of the teams and their respective social dynamics. This allows us to leverage the power of such insights to influence team resilience, robustness and sustainability.

The Effects of Disruption

When considering approaches to measure network characteristics, it is often very hard to come up with a single metric. The alternative is to consider several characteristics and compare their values before and after certain events, such as team disruption. In this way, we can observe the disruption happening and capture its effect on multiple dimensions.

Accordingly, we looked at the effect of disruptions on two levels: member and team. Member-level disruptions are those that do not affect the connectivity of the team and do not cut the flow of information, decision making or informal links. On the contrary, team level disruptions interrupt the flows and disturb the functionality of the network. When the removal of critical members results in the team breaking into more than one connected component, it is considered a team-level disruption. In the opposite case — when such disconnection does not happen we are looking at a member-level disruption.

Figure 1 — Primary team network of 8 members (top), member level disruption after removing member A (middle) and team level disruption after removing member C (bottom)

When introducing a disruption into the network, we measured it on two dimensions: scope and intensity. We measured scope by comparing the level of several metrics that describe a different aspect of the team network (density, diameter, radius, average shortest path length and global efficiency) pre- and post- disruption. We then defined intensity as the ratio between the number of removed links and the total number of links between the members in the network. The higher the ratio the larger the intensity of the disruption on the team.

One implication of this is that it shows how losing people who act as catalysts will result in adverse and negative consequences. The hidden costs cannot be expressed only in terms of hiring/firing costs. By providing a lens into the inner workings of the organization, leaders can analyze and understand value-creating networks that are vital for knowledge sharing, team collaboration, individual productivity, customer engagement, and innovation.

Scalable Experiments

To test our approach, we developed a large scale organizational network generator. Specifically for this exercise, we generated a small-scale set containing 20 team members allocated to 4 different teams and two large-scale networks of 200 members allocated to 40 different teams. Members could be part of more than one team simultaneously, with their contributions to each team expressed as percentages of their time. For each dataset, we created three different graphs describing different mechanics of an organization, namely the formal, decision making and information sharing organisational networks. A formal network describes who works with who, while an information-sharing network describes who people seek or provide advice to. Decision-making networks describe who someone consults before making a decision.

Figure 2 — Formal (top), information sharing (middle) and decision making (bottom) network of 20 members. We can see that the formal network is the densest while the decision making one is the least dense.

Using a synthetic network generator allowed us to establish in advance ground-truth values and rigorously test our approach. To test our disruption metrics, we artificially created in advance one key-player for each team. Thus, we applied our approach to each team and confronted the critical members and the members with the highest disruption effects with the ground truth values.

The table shows the results of our evaluation, where the fourth column indicates the percentage of teams for which our approach successfully identified their key-player. Our approach identifies all key players in all datasets except for a single member in the second case.

Going Beyond

Our experiments confirmed the usefulness of a wide variety of network metrics in measuring and preventing disruption. However, disruption is a complex phenomenon that touches more dimensions than the ones we looked at. For instance, one can look at the effect of disruption over time by collecting longitudinal data describing team networks at different times. When doing so we can explore how disruption affects the rate of change of informational flow, functional disabilities or decision-making failures. Beyond the disruption caused by removing members, we could also look at the disruption caused by introducing new people.