This data analysis technique, both an art and a science, produces insightful visualizations.
Imagine your delight when one of your social media accounts recommends a long-lost friend with whom you haven’t been in contact in years. You scratch your head for a moment wondering how on earth the social platform knew you and your friend were once connected, but happily click the “Add Friend” button in eager anticipation of reconnecting.
You’ve just experienced network analysis – the science (and art!) of representing traditional data as a network in order to find relationships and connections between objects. In the context of business, government, and other entities, network analysis is often used to understand the flow of information, identify influencers, and predict future trends.
Our team at Arboretica are experts in network analysis. It is the core of our technology, powered by our best-in-class open source data collection, data structuring capabilities, and natural language processing algorithms.
Let’s take a deeper look at how network analysis works and explore some of its potential applications.
What is network analysis?
Network analysis begins with data, which is transformed into a network representation consisting of nodes (objects) and edges (relationships or connections between objects). The resulting map provides previously unseen insights.
Once the data is represented as a network, we can customize its details to uncover various insights including relationships, influencers, and future trends.
Network representation is a powerful data visualization tool. It communicates insights that might be difficult to articulate with words, and it is something that is easily understood by people without technical expertise.
Before: Data organized in a spreadsheet
After: Data represented as a network
An example of network analysis: identifying Twitter influencers
Network analysis helps understand conversations and how information is flowing.
Imagine we want to identify influencers from a set of tweets. When putting tweets into a traditional table of data, we are only able to count the number of followers per user or the amount of engagement per tweet.
But when we create a network that connects two users who have mentioned, replied, or retweeted each other, then we see behavior from another angle. A center of activity begins to emerge, where particular users are mentioned by many people, or get lots of replies to their tweets.
This activity identifies influencers, but these might not be influencers in the traditional sense of the person with the most followers. Depending on the business need, the influencer might be the person who has the most engagement, the most interactions, or the most mentions. It also helps identify sub-communities within the overall network, as shown on the network map below.
In this network map, a connection only exists if a person (represented as a bubble) follows another person. The size of the bubble indicates the number of followers while the color of the bubble indicates the engagement (darker blue color = higher engagement).
The expertise required to create a map for network analysis
The science of network analysis is well-established, and the algorithms that are applied to surface connections are fairly tried and true as well. Creating a network map, however, is more challenging than meets the eye. There are two areas of “secret sauce” in creating the network map:
- The first is determining which data points will serve as the map’s nodes and which as its connectors. This is a manual task that requires domain expertise.
- The second is to create a map with the right level of detail: too many nodes connected and the visualization becomes too dense; too few nodes connected and the visualization loses its value.
Network Analysis in the Wild
In one of our most recent engagements, our team worked with a luxury durables manufacturer to develop a social network analysis. The goal was to understand the types of conversations in which the client was mentioned. These insights could drive product and marketing strategy.
We aggregated and analyzed social posts that referenced the company, and created a network map focused on those conversations’ keywords and topics. We found that most conversations were focused on two key topics — neither of which was a brand message the manufacturer was currently emphasizing.
In doing a similar analysis of social posts referencing competitors, we found that those competitors came up in a much wider range of conversation topics. Those competitors’ names came up in conversations about lifestyle and in association with other luxury brands. This meant that social content about those competitors was reaching a much wider audience with higher engagement–something our client wanted to achieve as well.
Network analysis of social media conversations allowed us to recommend to our client the following:
- Shift messaging towards high-performing existing topics and away from ones that were not generating conversation. We were also able to make recommendations on specific words to retire.
- Start conversations about the product in association with a certain lifestyle.
- Start conversations about the product in association with other luxury brands.
With network analysis, we are providing a new lens through which to look at information, and furthering our mission of bringing value from public data to the masses.
There are many applications of network analysis across industries. In the environmental industry, for example, network analysis can be used to track financial flows between sustainability investments or map out the environmental status of a company’s supply chain.
If you’re interested in learning more about network analysis or seeing how it could be applied to your business, reach out to our team.