Organisation Network Analysis: Finding How Work Really Flows

Organisation network analysis examines patterns of work relationships—who seeks advice, coordinates decisions, shares knowledge or connects groups. It can reveal collaboration that an organisation chart misses, but it should not label individuals as valuable or expendable based on a network score.

Start with a decision

Define the question: improve cross-team hand-offs, understand change influence, protect knowledge, redesign collaboration or diagnose overloaded coordination. Avoid collecting relationship data merely because software can draw an impressive map.

Choose the network boundary

Specify population, time period, relationship type and work context. Advice, information flow, trust and formal approval are different networks. A broad question such as who do you work with produces ambiguous links.

Select an evidence method

Options include a carefully worded survey, workshops, interviews, workflow records or approved communication metadata. Each has limitations. Digital traces show contact, not meaning or quality; survey responses reflect recall and response coverage.

Design questions around behaviour

Ask whom employees rely on for a defined task, decision or knowledge and perhaps frequency or importance. Limit nominations only when analytically justified. Do not ask who is influential without explaining the work relationship.

Prepare privacy and employee communication

Explain purpose, data sources, access, use, retention and whether participation is voluntary under the applicable design. Obtain current privacy, employment and worker-representation review. Do not reuse network data for undisclosed performance or redundancy decisions.

Clean and contextualise the data

Resolve identities, teams, contractors, missing responses and recent role changes. Account for shifts, remote work, leave and formal coordination roles. A low connection count may be appropriate for focused specialist work.

Interpret common patterns

Central connectors may enable coordination or carry unsustainable demand. Brokers link groups but may become single points of failure. Isolated groups may need connection or may intentionally protect specialist work. Clusters can show communities, silos or ordinary team structure.

Combine maps with qualitative evidence

Interview or observe representative work to understand why links exist. Ask about delays, decision rights, access and workload. Never infer trust, engagement or performance from a line on a diagram alone.

Protect individuals in reporting

Use role or team views where possible and restrict identifiable maps. Small groups and unusual positions can reveal people even without names. Avoid publishing rankings of most connected employees.

Turn insight into system changes

Possible actions include clarifying decision rights, distributing expertise, documenting knowledge, changing meeting or hand-off design and supporting boundary roles. Do not simply tell a central person to connect less without redesigning the demand.

Address nonresponse and visibility bias

Employees with more survey access, longer tenure or office visibility may appear more connected. Report response coverage and test whether shifts, contractors, remote workers or new joiners are missing. Do not interpret absence from the map as absence of contribution.

Distinguish resilience from dependency

Ask whether critical work has alternative routes, documentation and backup—not simply how many connections one person has. A highly central expert may be effective today while representing continuity risk.

Use external analysts carefully

Define instructions, access, methodology, security, output ownership and deletion. Prevent vendors from retaining identifiable relationship data for unrelated benchmarking or product development.

Example

A product launch repeatedly waits for one HR partner. Network evidence and interviews show all policy questions flow to that person because no owner directory exists. HR creates specialist routes and shared guidance, then checks whether response time and dependency improve.

Evaluate the intervention

Recheck the specific work outcome, connector workload, cross-team access and unintended effects. A denser network is not automatically better; too many links can increase interruption and slow decisions.

Govern future use

Assign data and decision owners, document method, expire extracts and review whether repeated analysis remains necessary. The map should be a temporary evidence tool, not a permanent hidden profile of employee relationships.