DATA ANALYSIS

Get started with network graphs—what they are and why they matter

Amanda Derrick

Written by Amanda Derrick

 

Multitasking? Great! You can listen to this blog post by clicking above or find our podcast, Intelligent Data Exploration, on major podcast platforms.

Businesses spend a lot of time, effort, and resources to collect and store data. But the information they gather doesn’t make an impact sitting in the cloud. It’s the relationships and drivers that the data can uncover that guide transformative business strategies. With datasets growing exponentially in volume and complexity it’s becoming incredibly difficult for data scientists and analysts to locate and understand the most impactful data relationships—and ultimately communicate them to stakeholders.

Businesses spend a lot of time, effort, and resources to collect and store data. But the information they gather doesn’t make an impact sitting in the cloud. It’s the relationships and drivers that the data can uncover that guide transformative business strategies. With datasets growing exponentially in volume and complexity it’s becoming incredibly difficult for data scientists and analysts to locate and understand the most impactful data relationships—and ultimately communicate them to stakeholders.

That’s where network graphs come in. 

Network graphs are visualizations that capture the connections and relationships in data. Those relationships are shown using nodes and edges, where nodes represent what is being analyzed and edges show how those nodes are connected. 

Where traditional BI tools like spreadsheets and 2-dimensional graphs can show limited relationships, they lack context and the ability to compare one relationship to another. Network graphs are capable of showing many relationships—and the magnitude of those relationships—at the same time. 

Why do network graphs matter? Your data has important things to tell you about your customers, equipment, logistics, and outcomes. Data-rich network graphs enable teams to identify root causes, manage risks, and make impactful decisions based on the communities and behaviors they discover. In short, great network graphs are your map to identify and refine successful business strategies.   

Great data science is being held back by outdated technology

The science of network graphs has been around for the last couple of decades, but they have relied on a lot of heavy lifting from data scientists, complicated technology stacks, and flat visualizations that don’t fully communicate the rich insight within. As a result, few organizations have started to take advantage of them as part of their data and analytics strategy. 

Most exploration is still done using traditional BI analyses, which don’t capture the relationships between multiple data points. Limiting the data you explore, and how you explore it, introduces risk and reduces value since you’re missing the relationships within that portion of your data. Your data's quantity and complexity are an asset, as long as you explore all the possibilities. 

From pivot tables and charts to networked communities

Consider the example of a dataset from Pew Research Center that surveyed people about their social media use. They gathered responses to 40 questions, including demographics (age, gender, marital status, etc.), their current internet or cable TV providers, whether they have a smart device, and whether or not they use a multitude of social media platforms. 

Some of the Pew Research findings looked like this:

Interesting information, but you’re missing the interactions between dimensions. Who is using Facebook most frequently, and what resources are available to them? What else do TikTok users in their 40’s have in common? Do Twitter users share a similar profession or education level? 

We have a lot of data but our knowledge of user behavior is still pretty light. If you’re an advertiser or marketer trying to reach a specific audience, you’d benefit from having more information about the communities hiding in your data. What’s the point of having so much data collected (remember, 40 data points!) if we don’t look at the relationships between responses?

We took this dataset and used the AI within our platform to create a network graph showing the communities of respondents. Our Network Extractor quickly sifted through all the data, grouping respondents together based on their similarities and differences to build nine distinct communities.

Our network graph showed us things like:

  • Young people make up the largest community, even though they were underrepresented in the survey, suggesting that their responses are consistently very similar. We’re confident that their behaviors are a good standard for their community. We expect this community to use all of these platforms, and to use them frequently.
  • There are three groups at the top of our network graph: Doesn’t Use Internet (green), Low Internet Use (peach), and Low Income (light blue). These groups are all characterized by low or no internet use. The “no use” community is very tightly grouped, sharing many of the same behaviors across respondents. 
  • There is a group of people who identify as regular readers. You might hypothesize that this community would be less likely to spend time on social media, however, the group’s positioning with the other communities characterized by internet use suggests that they are still regular internet users.

The power of network graphs is the ability to see all of the relationships underlying a business problem in their full complexity, identify the most valuable relationships, accurately understand the drivers of those connections, and use AI to discover all the what-ifs.

 

Practical examples of network graphs

Network graphs are incredibly versatile, showing relationships between a massive variety of data points. Here are just a few ways they can be used:

Manufacturing and energy

Consider the incredible networks and relationships within the transportation industry. A commercial airline is responsible for making thousands of decisions, from fuel planning to supply chain selection to pricing strategies. Network graphs can help them see how choices and trade-offs impact their overall goals.

Financial services

Banks handle massive numbers of users, transactions, and applications each day. The related data includes a variety of demographics, characteristics, and behaviors. Financial providers could use network graphs to find unbiased trends in loan applicants, detect fraud, or target ideal borrowers.

Supply chains

A supply chain can be impacted—for better or worse—by so many variables. If you want to expedite deliveries or introduce a new route how do you know what that might impact? Network graphs can illustrate those relationships, and find more that you might have missed.   

If you’d like to learn more about these examples and the power of network graphs combined with AI-guided exploration check out our free e-book A Beginner’s Guide to Network Analysis or set up a live demonstration to see our platform in action.