What Is Network Analysis?
Network analysis is a set of methods for studying relationships and connections between entities by representing them as nodes (the entities) and edges (the connections between them). The entities can be people, organizations, concepts, products, genes, web pages, or anything else that interacts with or relates to other things. Unlike traditional statistical methods that focus on attributes of individual cases, network analysis examines the structure of relationships themselves. The approach draws from graph theory in mathematics, sociometry from the social sciences, and computational methods from computer science. Researchers use network analysis to reveal hidden patterns of influence, identify central actors or bottleneck points, detect communities, and understand how information or behaviors spread through connected systems.
Why Network Analysis Matters
Most research treats observations as independent, but in reality, entities are embedded in webs of relationships that shape their behavior. A customer's purchase decisions are influenced by their social connections. An employee's engagement is affected by their team's communication patterns. A brand's market position depends on its competitive relationships. Network analysis makes these relational structures visible and measurable. Research published in Science found that network position predicted the adoption of new products more accurately than individual demographics in 74% of tested cases.
How Network Analysis Works
Building the Network
Every network starts with data about connections. For a brand association network, you might ask survey respondents which brands they consider when shopping a category, creating edges between brands that appear in the same consideration set. For an organizational network, you'd ask employees who they go to for advice, creating directed edges from advice-seekers to advice-givers. The resulting dataset is an edge list (pairs of connected nodes) or an adjacency matrix (a table showing which nodes connect to which).
Key Network Metrics
Several metrics characterize a network's structure. Degree centrality counts how many connections each node has; high-degree nodes are the most connected. Betweenness centrality identifies nodes that sit on the shortest paths between many other pairs of nodes, acting as bridges or brokers. Closeness centrality measures how quickly a node can reach all other nodes in the network. Density describes what proportion of all possible connections actually exist. A dense network (density approaching 1.0) means almost everyone is connected; a sparse network means connections are selective.
Community Detection
Networks often contain clusters of densely connected nodes with sparser connections between clusters. Community detection algorithms (like Louvain or Girvan-Newman) identify these subgroups automatically. In a brand network, communities might represent competitive sets that consumers view as interchangeable. In an organizational network, communities might reveal informal teams that don't match the official org chart. Identifying communities helps you understand the modular structure that isn't visible in aggregate statistics.
Visualization and Interpretation
Network visualizations use force-directed layouts where connected nodes pull toward each other and unconnected nodes push apart. The resulting map shows natural clusters, central hubs, peripheral isolates, and bridge nodes at a glance. Tools like Gephi, NetworkX (Python), and igraph (R) produce these visualizations. Interpretation requires moving between the visual map and the quantitative metrics: a node that looks central visually should have high centrality scores, and communities that appear as distinct visual clusters should be confirmed by the detection algorithm.
A Worked Example
A retail bank surveyed 1,200 customers about which financial products they used (checking, savings, credit card, mortgage, investment, insurance). Network analysis treated products as nodes and customers as edges (two products were linked when the same customer held both). The resulting network showed that checking and savings formed a tightly connected core, credit cards bridged to both the core and investment products, and insurance was the most peripheral product. Community detection identified two clear clusters: everyday banking (checking, savings, credit card) and wealth products (investment, insurance, mortgage). The bank redesigned its cross-selling strategy to follow the network's natural connection paths rather than pushing isolated products.
When to Use Network Analysis
- Brand mapping studies visualizing which brands compete directly, which occupy unique positions, and which serve as bridges between consumer consideration sets
- Stakeholder analysis identifying the most influential individuals or organizations in a decision-making ecosystem
- Product ecosystem research mapping how products, features, or services relate to each other in customers' minds or usage patterns
- Knowledge management understanding information flow within organizations to identify bottlenecks or isolated teams
- Co-occurrence analysis mapping which survey response themes, product attributes, or concepts appear together across respondents
Common Mistakes
- Confusing a highly connected node with the most important node because degree centrality and betweenness centrality measure different things: a bridge node with few connections may be more structurally important than a popular node with many connections
- Over-interpreting network visualizations without checking the underlying metrics, since force-directed layouts can produce misleading spatial relationships depending on algorithm parameters
- Collecting incomplete network data by surveying only a portion of the relevant population, which can make peripheral nodes appear isolated simply because their connections weren't captured
How Quali-Fi Supports Network Analysis
Quali-Fi's survey builder includes matrix and ranking question types that capture relational data like brand associations, product consideration sets, and interpersonal connections. The platform's data export preserves the paired structure needed for edge-list formatting, making it straightforward to move survey-derived relational data into network analysis tools.
Frequently Asked Questions
What's the difference between network analysis and social network analysis?
Network analysis is the broader methodological umbrella covering any study of relationships between entities. Social network analysis (SNA) is the specific application to social relationships between people or organizations. You can do network analysis on gene interactions, website links, or product associations without any social component.
How much data do I need for network analysis?
The minimum depends on the research question. For a small organizational study, 30-50 nodes can produce meaningful structure. For brand or product networks derived from survey data, you need enough respondents (typically 200+) to produce stable co-occurrence patterns. The number of edges matters more than the number of nodes for most analytical purposes.
Can I collect network data through surveys?
Yes. Name-generator questions ("Who do you go to for advice?"), roster methods (selecting from a predefined list), and consideration-set questions ("Which brands would you consider?") all produce network-ready data. Quali-Fi supports all of these question formats with appropriate response structures.
Related Topics
- Social Network Analysis
- Bibliometric Analysis
- Quantitative Content Analysis
- Factor Analysis Applied to Survey Data
- Data Collection Methods
Collect relational data with the right survey questions -- try Quali-Fi free for 14 days.