Network Analysis for Decision-Making
An Applied Introduction to Social Network Analysis for Innovation Intelligence
Network Analysis for Decision-Making
A practical, data-driven guide to Social Network Analysis — from theory to patent intelligence, collaboration mapping, and knowledge-flow analytics.
📡 Social Network Analysis
🔬 Innovation Intelligence
📄 Patent Analytics
🔗 Knowledge Flows
📊 Applied Methods
Why This Book?
We are comfortable with the statistics of things: averages, distributions, histograms, regression models. But some of the most valuable data we encounter is not about things at all — it is about relationships.
A co-authorship, a patent citation, a financial stake, a research collaboration, a supply chain link: these are relational data, and they carry a kind of informational richness that a bar chart simply cannot reach. To unlock it, we need a different toolkit.
The core insight: When you represent data as a network, you gain access to a new analytical dimension — not what entities are, but where they sit in a web of interactions. Structure reveals what attributes alone cannot.
The field that provides this toolkit has a slightly misleading name: Social Network Analysis (SNA). Its methods were born in the study of human social systems, but they have long since spread across every scientific discipline. Whether you are mapping co-inventor relationships, tracing knowledge flows through patent citations, or identifying strategic alliances between firms, you are using the same underlying mathematics. The interpretation changes with the data — but the theory is universal.
This book has a dual ambition. It presents the theory of network analysis clearly and rigorously. And it presents a curated set of applied case studies that ground every concept in real data from real decisions.
What Will You Learn?
Each case study in this book follows the same three-part structure:
- The data — where it comes from, what it actually records, and what biases to watch for.
- The network construction — what choices were made, what cleaning was required, and why.
- The analysis and interpretation — reading the structure in context, with honest acknowledgement of what can and cannot be concluded.
The topics unfold progressively, from foundations to complex analytical challenges:
Part I · Theory
Graph Foundations
Nodes, edges, direction, weight, loops. The vocabulary of network science, built carefully so that later concepts click into place.
Part I · Theory
Network Types
Undirected, directed, bipartite, multiplex, multilayer, interconnected — each type calls for different indicators and different interpretive caution.
Part I · Theory
Micro-Level Indicators
Degree, betweenness centrality, closeness centrality, eigenvector centrality, radiality, topology coefficient, clustering coefficient. What each captures, and what each misses.
Part I · Theory
Meso-Level: Communities
The modularity framework, the Louvain algorithm, how to evaluate a partition — and what it really means when a group of nodes clusters together.
Part I · Theory
Macro-Level Descriptors
Degree distributions, density, diameter, average path length — the structural fingerprint of a network.
Part II · Applications
Textual & Semantic Networks
Six co-occurrence indicators — raw counts, Cramér, χ², mutual information, cosine, distributional — with worked calculations so you know exactly what you are choosing.
Part II · Applications
Patent Citations
Inventor vs. examiner citations, citation categories (X, Y, E, P, O, T), strategic biases, knowledge-flow interpretation — a rigorous field guide to the most complex data source in the book.
Part II · Applications
Collaboration Networks
Patents, publications, EU projects, ANR grants: how each source constructs a different network, carries different biases, and supports different conclusions.
Part II · Applications
Technological Proximity
The Breschi, Jaffe, and Nesta–Saviotti proximity measures, applied to IPC co-classifications. From pairwise proximity matrices to complementarity graphs for partner identification.
Part II · Applications
Patent Thickets
How to detect dense IP overlap structures that threaten commercialisation, using citation networks filtered to blocking citations and triangular dependencies.
Part II · Applications
Part III · Advanced Topics
Inventive Trajectories
Main path analysis and emerging methods for tracing how a technology evolves through time, from foundational patents to recombinant frontier.
A Data-First Book
Every chapter in the applied sections is anchored to real sources. The data touched in this book includes:
Lens.org (JSONL exports) EPO / Espacenet Questel Orbit PatStat Scopus Web of Science CORDIS (EU Projects) ANR (data.gouv.fr) Google Patents Cortext Manager
Sources are used critically — their biases are documented, not swept under the rug.
The running example threading through the theoretical chapters is a collaboration network extracted from ANR-funded 5G research projects — a compact, real-world network that makes every indicator tangible. The applied chapters then scale to industrial case studies: tyre manufacturers, chemicals firms, aerospace actors, biotech clusters.
A Note on Interpretation
Networks are seductive. Visualisations are striking, centrality scores feel authoritative, communities look satisfying once coloured. This book tries hard to counteract that seductiveness with rigour.
The methods and indicators are universal. The interpretation depends entirely on the data. A co-classification link, a citation, a collaboration in a patent, and a collaboration in a scientific publication do not mean the same thing — even if the mathematics treats them identically.
Every chapter therefore spends time on what a given indicator cannot tell you: the limits of density as a comparator, the national biases embedded in examiner citations, the ambiguities of co-deposited patents, the difference between textual proximity and semantic proximity. Sound analysis depends on knowing these limits before drawing conclusions.
Who Should Read This?
Masters students in innovation management, science & technology studies, strategic intelligence, data analytics for sustainability, and related fields — especially those encountering relational data for the first time and needing both theory and worked examples.
Professionals in competitive and technology intelligence who already work with patent and publication data but want to add the structural dimension — mapping ecosystems, tracking knowledge flows, identifying gatekeepers and dependencies.
Applied researchers in economics, management, and bibliometrics who need a precise reference for the indicators and their underlying assumptions, with enough mathematical detail to make informed methodological choices.
R practitioners who will combine this book with the companion NetworkIsLifeR package for end-to-end patent parsing, network construction, and indicator calculation — without leaving the R ecosystem.
How to Read This Book
The theoretical chapters (Part I) are designed to be read sequentially — each concept builds on the previous. The applied chapters (Part II and III) are written to be more modular: once you are comfortable with the foundations, you can move to whichever data source or analytical challenge is most relevant to your work.
Throughout, you will find:
- Worked numerical examples — small enough to follow by hand, precise enough to verify a piece of software.
- Definition boxes — formal statements of key concepts, clearly separated from the prose.
- Interpretive remarks — honest assessments of what an indicator does and does not reveal in context.
- Visual illustrations — network figures drawn from real analyses, annotated to support reading rather than just display.
The companion R package NetworkIsLifeR implements many of the workflows described here and is available at github.com/JPvdP/NetworkIsLifeR.
A Word on Nomenclature
The field is called Social Network Analysis because that is where the foundational work was done — mapping human social systems. The name has never changed, even though the methods are now applied just as readily to protein interactions, financial exposures, citation flows, and technology co-classifications.
Do not let the name mislead you. When we analyse a firm’s patent citations or map co-classifications across a technology domain, we are doing SNA. The objects are different; the mathematics is the same.
Network Analysis for Decision-Making · Johannes van der Pol · Utrecht University
Rendered with Quarto · English translation in progress