Participant network analysis

Participant network analysis is the practice of mapping the interactions between people in meetings as a network, where each person is a node and each direct exchange is a connection. It answers a question that attendance lists cannot: not who was in the room, but who actually engaged with whom. A meeting of eight people where two dominate and six stay silent looks very different on paper to one where conversation flows across the whole group.

The analysis draws on signals like who responded to whom, who was addressed by name, who asked and who answered, and who shared follow-up work. From these signals it builds a graph. Some people sit at the centre with many connections. Others sit at the edge with one or two. A few act as bridges, linking otherwise separate clusters of the team.

The value is in the pattern. A healthy collaboration network spreads connections widely and has few isolated nodes. A fragile one concentrates connections on one or two people, which creates a single point of failure when those people are away or overloaded.

The headline measure is network density: the share of all possible connections that actually exist. With eight participants there are 28 possible pairs. If only 10 pairs interacted directly, density is 10 divided by 28, or roughly 0.36. A density near 1 means everyone engaged with everyone. A density near 0 means interaction ran through a tiny number of people.

Density alone is not enough. The same density can describe a balanced web or a star with one person at the centre. To tell them apart you need node-level measures, which describe each person rather than the whole network.

1. 1

   Participants (n)

   The count of distinct people in the network for the period under review. Decide early whether observers who never speak are included, as this changes density.

2. 2

   Actual connections

   Each unique pair of people who directly interacted, counted once. A direct interaction is a response, a question and answer, a hand-off, or being addressed by name.

3. 3

   Degree per node

   The number of distinct people each participant connected to. High-degree nodes are central. Zero-degree nodes are isolated and worth following up on.

4. 4

   Betweenness per node

   How often a person sits on the shortest path between two others. High betweenness marks the bridges who hold separate groups together and become bottlenecks when absent.

A single density number tells you the network is thin or rich, but not why or what to do. A metric tree breaks the network into the drivers that move it, so a flat or concentrated network points to a specific action rather than a vague concern about collaboration.

There is no universal target, because the right shape depends on the meeting. A daily stand-up is meant to be lean and will look sparse. A cross-functional planning session should look dense and mixed. The ranges below are a starting reference for recurring working meetings of five to fifteen people, not a rule for every format.

Improving the network is not about forcing everyone to speak equally. It is about removing the structural reasons that connections fail to form: meetings that are too large, formats that reward the loudest voice, and silos that keep teams apart. The actions below target the drivers in the tree.

1. 1

   Treating connections as a scoreboard

   Ranking people by connection count rewards talking over listening. It pushes people to interject for visibility and degrades the meeting. Use the network to fix structure, not to grade individuals.

2. 2

   Counting attendance as interaction

   Being in the room is not a connection. If the analysis treats every attendee as connected to every other, density is meaningless. Only count direct exchanges.

3. 3

   Ignoring meeting type

   Comparing a stand-up to a planning workshop on the same density target produces false alarms. Benchmark each meeting format against its own history, not against a single universal number.

4. 4

   Reading one meeting in isolation

   Networks are noisy in any single session. A quiet member one week may carry the next. Track the trend over several meetings before drawing conclusions about isolation or dominance.