Networks
In a globalized world, everything is interlinked - politics, people, companies, finance, social trends, and economies. Strategies fail because firms attempt to place complex systems into a black-and-white category. Removing context in favor of synthetic classification creates a massive risk to any strategy or investment decision. Network analysis excels at exploring complex data sets - surfacing how people, regions, events, and trends are connected, allowing organizations to find risks, blind spots, and opportunities not possible with traditional analytics or dashboards. The Insights extracted can be used to shape business decisions or summarize influential and emerging narratives. Business leaders are often surprised by the complexity within their initial area of focus. Combining network analytics with advanced natural language processing allows firms to quickly find connections between unstructured data to see how narratives and thematics intersect at a level of granularity and context not possible with traditional research.
How to Read a Network? Below are 1500 news articles (OSINT) mentioning AstraZeneca during the suspension of their COVID-19 vaccine by EU member states.
Nodes (circles in the network ) with similar language are connected with a link (degree) and are nearer to one another in the network.
Nodes that bridge different clusters throughout the network are often insightful to investigate, given they share connections across different topics, entities, or themes.
Nodes without links have a unique language that does not substantially match any other documents. This feature can be inconsequential or novel, so they should also be explored. Sometimes, these topics can represent an outside disruptor or issues that are emerging but not well known.
Network Metrics
Centrality: Nodes with high betweenness centrality have many thematics which extend across the network. Nodes with low betweenness centrality do not connect to other nodes/clusters beyond their immediate vicinity.
Degree: Degree measures the number of shared connections for a node or cluster (clusters are made of many similar nodes).
Flow: Flow represents the combined strength of a node’s connections. Nodes with a high flow have a more shared language and more connections with other nodes in the network. Nodes with a low flow contain thematics that are peripheral to the network.
Inter-Cluster Connectivity: Nodes with high inter-cluster connectivity contain language shared by nodes in multiple different clusters. A node with low Inter-Cluster Connectivity is not connected to many other clusters. The metric helps us identify thematics and entities that bridge multiple topics so that one strategy can encapsulate multiple issues or opportunities.
Entity Extraction
In addition to understanding the overall trends, algorithms can extract the entities associated with the clusters and nodes (circles). The graph below is the SAME data from the network above, only the visual shows how people and institutions are connected.
Identify Latent Risks Years In Advance
For example, in 2015, OSINT data, network analysis, and machine learning (left tweet) uncovered that EU GDPR was a potential risk to medical research. At the time, most firms thought of policy in terms of physical regions, not that ideas and narratives, and influence were borderless. So it came as no surprise that few, if any, people were making the connection between Health and digital privacy. It wasn’t until 2019 that top researchers made the connection in the US and identified the risk GDPR had to the research global community.
