Top 5 use cases for graph databases

So what are the emerging use cases of these new capabilities? Here’s how five industries are taking advantage of graph databases’ extremely fast relational query performance across distributed data stores.

1. Gaining a 360-degree view of customers

Interactions between companies and their customers or sales prospects tend to be complex, with many touchpoints. Ideally, these should yield sales strategies that continuously adapt to customer needs. Such 360-degree scenarios quickly incur many-to-many relationships that, using a relational database, would require laborious modeling and cumbersome table joins to yield actionable insights.

This is the sort of situation where a graph database shines. UnitedHealth Group (UHG), for example, has adopted a graph database to help improve the quality of care for over 26 million members while reducing costs. The largest healthcare company in the world by revenue, UHG uses a massive graph database to track more than 120 billion relationships among members, providers, claims, visits, prescriptions, procedures and more.

UHG has developed various GUI applications atop its graph database that, among other benefits, provide a consolidated view of member interactions between physicians, pharmacies, clinical labs, health advisors and UHG itself. Over 23,000 users access the database every day, enabling providers to determine better care and wellness recommendations based on the latest member activity in real time. UHG predicts that the cost savings may eventually run into the billions.

2. Transforming financial services with AI

The exponential growth of data has been the biggest enabler of AI/ML, which requires large quantities of data to surface meaningful patterns and improve the accuracy of decision-making. Few industries are more data-intensive than financial services, but as with other industries, data originates from many different sources and typically ends up in relational database silos.

In bridging those silos, graph databases can help AI/ML deliver superior predictive analytics, risk management, fraud detection, anti-money laundering, insider-trading monitoring, automated recommendations for customers and more. Also, a graph database coupled with AI/ML can help ensure data is clean in the first place, reconciling anomalous differences in customer records and financial product attributes that could lead to inaccurate results.

Intuit is using graph database software in combination with AI/ML to transform from a product company into an AI-driven expert platform company. A key part of this journey is the creation of knowledge graphs, which enrich data and surface insights from clusters of related elements. Intuit combines knowledge graphs with the most advanced form of ML, deep learning, to power Intuit’s chatbots and in-app recommendations. Normally, it’s hard to determine how deep learning arrives at its outcomes; a key benefit of Intuit’s knowledge graphs is that they add “explainability” to deep learning.

3. Optimizing supply chains

Among the lasting effects of the coronavirus pandemic has been the realization that global supply chains can be alarmingly fragile. With or without disruption, manufacturers are acutely aware of how complicated many supply chains are to maintain and optimize.

Consider the day-to-day challenges faced by auto manufacturers. The first requirement is to accurately forecast customer demand to determine the number and types of parts to order — down to the various models and options buyers are expected to choose. Those predictions need to sync with the availability of parts from hundreds of suppliers, along with estimates of manufacturing efficiency and supplier risk.

Jaguar Land Rover (JLR) chose a graph database solution because it could span the many data silos that needed to be tapped for supply chain analysis — and explore the matrices of relationships among data elements. The primary goals were to increase the average profit per unit sold and to reduce aged inventory, along with minimizing the effects of supplier disruption. Some key supply-chain planning queries at JLR now take 45 minutes as opposed to weeks and, more importantly, management can answer questions it never had the opportunity to ask before.

4. Improving online retail operations

Retail ecommerce firms face growing competitive pressure to deliver better customer experiences built on accurate customer details and purchase histories. That foundation enables everything from dynamic pricing to product recommendations to personalized special offers, all of which draw on data accrued along the customer journey.

Graph databases can help in a number of ways. Consider the possible relationships — between customers and payment methods, customers and brands, products and return rates, promotions and sell-through rates, and a whole lot more. Say you wanted to run a query to determine which promotions were most effective for a certain product when pitched to a subset of customers defined as loyal. With a relational database that would take a long time, but a graph database can return the results with very little latency.

The seemingly simple act of reliably identifying which customers purchased what can be improved by a graph database, which can aggregate and reconcile all associated customer data regardless of the payment method or point of sale. In a three-month test of a graph database, one large ecommerce company discovered 12 million new account connections across its five different retail websites. The company estimated an efficiency saving of nearly $3 million and predicted a 17.6% increase in sales.

5. Improving the accuracy of fraud detection

We’ve all witnessed the evolution of fraud detection through our bank, credit card and telecom companies. Early rule-based efforts tended to miss dubious transactions or flag innocent transactions as fraudulent. When the financial industry adopted graph databases to augment their AI/ML efforts, however, the accuracy of fraud detection improved noticeably.

Graph databases coupled with AI/ML improve the accuracy of fraud detection, reducing false positives and detecting anomalies that might otherwise be missed. Machine learning must draw on many different data types to model a customer’s normal behavior — location, device, payment type, authentication method and so on. Plus, what’s defined as normal behavior patterns must be adjusted on the fly in response to legitimate change. Graph databases support that dynamism and enable AI/ML to traverse customer interactions to identify significant variances.   

Financial services giants JP Morgan Chase and Intuit have both adopted graph databases to boost their AI/ML fraud detection efforts. JP Morgan Chase uses a graph database to help protect more than 60 million households in the U.S. According to Intuit, graph-based machine learning has enabled the company to detect 50% more potential fraud events and has reduced false positives by approximately the same percentage.

These are just a few of the most common uses for graph databases. Customers are also using graph databases to optimize business processes, improve healthcare outcomes, sharpen digital marketing campaigns, identify cybersecurity threats and even manage energy grids. New applications emerge on a regular basis.

The mission of the graph database is to open a whole new window on relationships among data elements, delivering analytics that can identify fresh business opportunities, flag wasted motion, and provide a nimble foundation for AI/ML initiatives. When given access to multiple enterprise data stores, graph databases can offer entirely new insights and capabilities.

Yu Xu is CEO of TigerGraph.