Communication network analysis

Communication network analysis is the practice of mapping the connections between people based on how they communicate, then measuring the structure of that map. Each person is a node and each communication link, such as a message, an email exchange, or a meeting, is an edge between two nodes. The result is a graph that shows how information actually moves through a team, a department, or a whole organisation.

The method matters because the real flow of information rarely matches the org chart. A formal reporting structure says who manages whom. A communication network shows who people genuinely turn to, which teams have gone quiet with each other, and which single individuals hold the organisation together by connecting groups that would otherwise be islands. Those connectors are often invisible on paper and irreplaceable in practice.

The analysis surfaces three patterns in particular. Connectors are the people whose links hold otherwise separate clusters together. Silos are dense groups that barely communicate with the rest of the network. Bottlenecks are nodes that sit on the only path between groups, so information has to flow through them. Reading these patterns helps explain why a decision stalled, why two teams keep duplicating work, or why one person leaving would sever a critical link.

There is no single number for a communication network. Instead, a small set of graph measures describe its structure, and density is the most common starting point. Density is the share of all possible connections that actually exist, which captures how tightly knit the network is. The inputs below define density and the other core measures you read alongside it.

1. 1

   Nodes and edges

   A node is a person and an edge is a communication link between two people. Defining what counts as an edge, such as a minimum number of exchanges, prevents a single stray message from creating a misleading connection.

2. 2

   Network density

   Density divides the actual connections by the possible connections. For a network of 10 people, the possible connections are 10 times 9 divided by 2, which is 45. If 18 pairs actually communicate, density is 18 divided by 45, or 40 percent. High density means information spreads easily; low density points to fragmentation.

3. 3

   Centrality

   Centrality measures how important each node is to the flow. Degree centrality counts a person's direct connections, while betweenness centrality counts how often they sit on the shortest path between others. High betweenness flags the connectors and the bottlenecks.

4. 4

   Clusters and bridges

   Clustering reveals the dense groups, and bridges are the edges that link them. A bridge held by one person is a structural risk, because removing that node would split the network into disconnected parts.

A worked example. A 12 person team has 66 possible connections. Mapping a month of messages shows 24 active communicating pairs, giving a density of roughly 36 percent. Two clusters appear, one in engineering and one in design, joined by a single bridge running through one product manager. That manager has very high betweenness centrality, which means information between the two clusters depends on one person. The numbers say the team is moderately connected; the structure says it has a single point of failure.

A network graph shows you that the organisation is fragmenting or that one connector carries too much load, but it does not assign the problem to anyone. A metric tree decomposes network health into the drivers that shape it, so a falling density or a rising bottleneck score becomes a route to a specific cause and a named owner. Network health sits below connectivity, resilience, cross-team flow, and information reach, and each branch can be measured and owned.

This is the gap between dashboards and decisions in a different form. A dashboard can render a striking network map and still leave leadership unsure whether to act, because the picture has no owner and no chain to a cause. The tree connects the structural signal to the team that can change it, whether that is reducing dependence on one connector or rebuilding a broken bridge between two functions.

Network benchmarks vary enormously with size, because density falls naturally as a network grows: a 5 person team can plausibly connect everyone, a 500 person one cannot. The ranges below are directional and assume small to mid-sized teams. The structural warning signs, such as a single bridge or a dominant connector, matter more than any single density figure.

Read density together with the structural measures. A network at 30 percent density that depends on one connector is more fragile than one at 20 percent with several redundant bridges. The single most important question is whether removing any one person would split the network, because that points to a risk no headcount figure will reveal.

Improving a communication network means strengthening the connections that carry critical information and reducing dependence on any single point. The aim is a resilient structure where information reaches the people who need it without flowing through one overloaded individual. The levers below target the most common structural weaknesses.

1. 1

   Comparing density across different sizes

   Density falls as a network grows, so a small team will always look denser than a large one. Comparing the raw number across networks of different sizes leads to false conclusions; compare structure and trend instead.

2. 2

   Reading content instead of structure

   The analysis is about who connects to whom, not what they say. Drifting into reading message content breaks trust and misses the point, which is the shape of the flow.

3. 3

   Ignoring edge definition

   If a single stray message creates an edge, the network fills with noise. Setting a sensible threshold for what counts as a real connection keeps the structure meaningful.

4. 4

   Treating connectors as expendable

   High-betweenness individuals look like just another node on a chart but hold the network together. Failing to flag them as a structural risk means their departure can quietly sever the organisation.