Graph Databases: Why Your Business Needs One

Posted 2020-08-27 Posted by Tom O'Connell

What is a graph database?

The structure of a Graph Database is easy to understand; they consist of nodes and relationships. Each node represents a unit (a name, a person, a place) and the relationship defines the connection between the two of them. For instance, the two nodes, “Titanic” and “boat” would have a relationship that states that Titanic “is a type of” boat.

When looking at popular use cases, social media sites such as Twitter and Facebook are able to utilise Graph Databases to understand the connections between their users. As they have millions of users, they enable them to clearly see which accounts follow which, if they are followed back, and those most viable to connect the two mutually. Google’s search engine works in much the same way, connecting billions of data points to each other.

In traditional relational databases, data can be accessed or reassembled without disrupting the structure. These databases store facts and information, whereas Graph Databases go beyond, defining and describing the relationships between these data points.

It’s not a coincidence that Facebook asks if you know those lost connections you haven’t seen for 15 years. They’re building a web of connections to recommend the people you’re more likely to know by using a database that can join the dots. They keep their users engaged and crucially, reduce churn.

Pivigo teamed up with the Ditchley Foundation, an organisation bringing together a unique mix of people from politics, technology, finance, academia to discuss wider issues in society and the world, to build a database that can use connections from social media to guide them when selecting individuals to invite to conferences.

By using Twitter data to create networks, algorithms can identify the critical individuals in those spheres, isolating the most viable to invite to conferences to influence conversations.

What makes graph databases crucial to the future of business is the limitations of relationship queries in traditional databases. As the scale of data ever-increases, traditional solutions are unable to store ever-growing lists of connections and performance can grind to a halt.

As Graph Databases are scalable, they’re the perfect investment to facilitate big data applications, giving business easier and faster access to vital insights. Throughout the last decade, JSON databases established themselves as the dominant player in their field. Given that some of the hardware limitations that plagued Graph Databases throughout that period are now redundant, it is likely they will become the clutch database of the 20s.

One of the most powerful applications for Graph Databases right now is fraud detection. This industry is over-reliant on straightforward checklists. Whilst this may result in successful detections, these simplistic methods are unable to pick up on subtle attempts. Graph Databases, however, can unearth unusual connections or clusters of suspicious activity.

These activities may not flag as anomalies on their own, but when the relationship between them is established, a string of financial transactions that are unusual when connected or show distinct routine patterns can alert to instances of money laundering.

Pivigo, through its S2DS programme, where analytical PhDs work on data science solutions for 5 weeks with project partners, worked with a leading biopharmaceutical company to build a Graph Database to look at the sentiment of its brands on social media. The S2DS team delivered a tool that can gather valuable insights into their products and identify potential customers.

To see how we built Ditchley’s Graph Database and how we used data science to solve their other business challenges, watch our Webinar below.

Get priority access to Pivigo news, features, events and networking opportunities