Imagine being able to answer these six questions with your database

Graph databases are changing the data paradigm by enabling a new way of looking at information. By Harry Powell, Head of Industry Solutions, TigerGraph.



Since the inception of the relational database in the 1970s, the world’s information has been marshaled into vast libraries of tabular data, and it’s difficult to imagine managing the vast amounts of data in the world without them. Conceived of as a way to automate traditional paper-based ledgers, they were brilliant at the storage, retrieval, and aggregation of data, increasing accessibility to information and reducing error rates.

However, relational databases have inherent limitations that keep us from answering more difficult questions - questions which, if we could answer, would provide invaluable insights.

It’s not enough to retrieve single records and examine them in isolation. Organisations need to understand their data in an increasingly interconnected and complex world. With graph databases, you can do this, because graph databases change the storage paradigm. Rather than fit data into rigid structures, graph databases allow you to create a data structure that more closely mirrors real life, storing data in nodes and links.

In financial services, for instance, we can represent a transaction between two entities, Alice and Bob, as two nodes with a directed link showing a payment made by Alice to Bob. To query that, you would simply follow the link from Alice to Bob.

In a relational database, you would have a table for accounts with a row each for Alice and Bob and a table for transactions. To explore that transaction, you would have to create a join table to show Alice paid money to Bob.

The advantages of the graph approach become apparent when you scale up your queries. For instance, show me all the paths that lead from Alice to Bob, even if they lead through other entities. Or show me the entities that interact with both Alice and Bob, or the community of entities that surround Alice and Bob and how they interact.

With graph, you can exceed the limitations of SQL and perform deep-link analytics on your data, which enables you to answer a whole new class of questions. Let’s look at some.

How do I get a holistic, 360-degree view of my customers?

Given the sheer volume and complexity of data involved in sales and customer relations, many businesses struggle to get a holistic view of their customers, but some organisations are achieving this with a graph solution.

Traditional solutions are built upon relational databases, which store information such as account, contact, lead, campaign, and opportunity in separate tables, one for each type of business entity. Relational databases are good at indexing and searching data, supporting transactions, and performing basic analysis, but they are poorly equipped to connect across

the tables and identify hidden relationships and patterns across multiple leads, campaigns, and opportunities.

And you can go further, using graph analytics and sophisticated algorithms to extract insights about customer behaviour on an industrial scale to better understand the buying patterns of both individuals and entire demographic groups.

Among a cohort of influencers, which are more influential?

Influence is a measure of the degree to which entities link to a given entity, and it was the essential question behind PageRank, a graph algorithm invented by Google to find the most important pages on the internet based on the number and quality of inbound links.

PageRank revolutionised search, but it has found numerous applications in social media, fraud detection, supply chain management, and customer recommendation.

How likely is this transaction to be fraudulent?

While traditional fraud detection systems rely on scoring transactions based on what’s known about the transaction initiator, more complex solutions driven by graph allow you to quickly analyse the account holder’s connections to other known sources of fraud. This enables financial institutions to not only look at the parties to a transaction but also the context of the transaction, including the relationships of the parties to other entities that could be involved in fraud.

Today, four of the tier one banks in the US are using graph algorithms to improve the quality of their fraud scores and provide actionable intelligence to their investigators.

Is this phone call likely to be spam?

With milliseconds to make a decision, telecoms companies would like to be able to help their customers identify calls from suspicious numbers in real time. Blocklists are easily bypassed and take time to initiate, so automated systems such as those developed by China Telecom using a massive graph database are the way forward.

China Telecom considers more than 100 factors in assessing the risk that the number from which a call is being initiated is likely to be a scammer. Mapping the phone numbers and factors against a graph database, the company uses graph algorithms to assign a risk score to a caller. If the score exceeds a threshold, a message is flashed to the recipient’s phone to alert them that the call has been flagged as suspect, allowing them to decide whether to accept the call.

How can I re-optimise my manufacturing operation in response to shortages of critical components?

In complex manufacturing operations, where it’s possible to make many different products, a shortage of one component doesn’t mean the entire assembly line has to grind to a halt, but it can take time to work out how. Jaguar Land Rover used a graph-based solution to reduce the amount of time it needed for this from three weeks to 45 minutes.

The company did this by layering a graph database on top of its existing supply chain management systems, to create a single connected overview of the entire operation from suppliers and inbound shipping through assembly lines and internal logistics to the sales order book. In this way, it optimised a complex manufacturing system – which supports the custom assembly of cars comprising some 4,500 parts each from an inventory of more than 30,000 components – and save around $100 million in the first year.

Where is the fault in my very complex IT system and what are the downstream implications?

A leading British retailer has a large and complex IT system that is essential in maintaining the flow of products and information between warehouses and several thousand shops. The IT system requires more than 20 development squads to build and maintain nearly 100,000 IT processes, which do everything from updating databases to generating management reports. If any one of these systems fail, it can have unpredictable and far-reaching knock-on effects.

Initially the retailer adopted a graph solution to manage service level agreements and track KPIs as part of the IT department’s agreements with business analysis teams. If something broke, it needed to quickly identify what it was and which development squad was responsible for fixing it.

The fault tracker was very successful, and the retailer soon began to find other uses for the graph database system including identifying and triaging potential single points of failure to take a more proactive approach to fault management.

This is an exciting time for graph database technology as it changes the paradigm for data storage and analytics to enable organisations to take a fresh look at old problems and generate new business intelligence from existing data.

By linking up legacy databases and deploying sophisticated analytics and algorithms, graph database technology is helping organisations gain insights into new and existing markets in a way that relational databases simply could not, enabling them to increase revenues, reduce costs, and manage risks.

Graph technology is rapidly finding new applications that improve operational efficiency in supply chains, energy management systems, and IT and telecoms networks. It is the foundational tool behind many cutting-edge solutions in machine learning, artificial intelligence, and geospatial and time series analysis.

It’s no wonder that organisations in industries as diverse as advertising, financial services, healthcare, retail, and manufacturing, among other information-intensive sectors, are embracing graph technology to amalgamate and process vital business intelligence.

By Harvey Lewis, Associate Partner, Client Technology & Innovation at EY.
By Fred Crehan, Area Vice President, Emerging Markets, Confluent.
By Dr Sina Joneidy, senior lecturer in digital enterprise at Teesside University International Business School.
By Andrew Watson, Chief Technology Officer, MHR International
By Helena Schwenk, VP, Chief Data and Analytics Office at Exasol
For many years, ownership of customer loyalty within an organization has traditionally sat with marketing, customer services and, more recently, customer experience departments. By AppDynamics EMEAR CTO, James Harvey
By Graham Jarvis, Lead Business and Technology Journalist.
Simon Spring, Operations Director of EMEA at WhereScape, discusses some of the common misconceptions surrounding data warehouses, data lakes, and data hubs, and why a proper understanding of each is key to building effective data infrastructure.