by Paula Rooney

Jaguar Land Rover gets more from graph analytics

Dec 03, 2021
AnalyticsDigital TransformationGraph Databases

Facing factory shutdowns, the multinational automaker retuned its analytics forecasting strategy around graph database technology to establish more precision and profit than expected.

Conceptual trend lines track + monitor data analytics [forecasting / future / what's next]
Credit: SolarSeven / Getty Images

The pandemic has been a perfect storm for Jaguar Land Rover’s core business, including a two-month factory shutdown, a semiconductor shortage, and profound challenges with both supply and demand. But thanks to advanced data analytics, the British multinational automotive company not only weathered this storm, but has done so with more precision and profit than expected.

JLR’s 40-person data science and analytics team has developed an innovative forecasting engine atop a mixed proprietary/open source stack to the tune of £100 million in revenue during each of the past three years, with £2 million in profit directly attributed to JLR’s data team in 2020 — despite a disastrous global pandemic, says Harry Powell, director of data and analytics at JLR.

“One of the key parts of our strategy has been implementing graph technology in the business, and we’ve had some reasonably good results applying it to the supply chain,” says Powell, noting that JLR’s use of graph database technology from TigerGraph has been critical in reducing the automaker’s supply chain planning from three weeks to 45 minutes.

JLR now plans to deploy graph database technology to address quality improvement and pricing applications for its automobiles.

As opposed to relational and non-SQL databases, graph databases detect, capture, and leverage connections among data stored or actively in use in business processes in real time, making them superior to relational databases when tackling challenges involving “incidental and unpredictable relationships,” says Carl Olafson, a research vice president at IDC.

“A supply chain tracking system is a perfect application for a graph database,” Olafson says.

Gearing up for graph

Both graph and relational databases support relationships among data, but the difference in the nature of these relationships is key, Olafson says.

“In a relational database, the relationships are definitional in nature … An order can’t be associated first with one customer, then another,” he explains. “By contrast, relationships do not impact the meaning of nodes in the [graph] database. The combinations of nodes (data entities) and edges (their connections) can change ephemerally at any time. This is easy on a graph database, but more challenging using relational [databases].”

Harry Powell, director of data and analytics, Jaguar Land Rover

Harry Powell, director of data and analytics, Jaguar Land Rover

Perhaps not surprisingly, leading social media firms Twitter, Instagram, and Facebook lean heavily on graph databases. Gartner Group predicts that graph technologies will be deployed for decision-making in 30% of organizations worldwide by 2023.

“We see the graph as a central technology for our digital transformation,” says Powell, adding that JLR’s success in managing inventory and supply chain headaches rested largely on its pivot to graph-based analytics.

“We were one of the first firms to think about data as a network because it allows you to see connections and relationships, adapt to change, and take action in a way not conventionally faced,” Powell says.

JLR has moved an “enormous” amount of data to the cloud but like most enterprises still has pockets of legacy systems and an occasional developer’s desktop with “Do Not Touch” notes stuck to it. Still, graph technology has enabled the company to join sources of data together and make connections within data they previously thought they could not join.

Implementing this technology, Powell says, allows JLR’s data team to answer questions that, for the past 20 years, they didn’t think were possible to ask.

The automaker is now moving “towards this aim of a digital twin that we can reflect the world in our data by digitizing it into a graph format, which will give us enormous power to make our businesses better, our customer relationships better and make our products better,” Powell said during a recent keynote at a TigerGraph summit.

Digital twins, which establish real-time virtual representations of physical objects and processes, are often used to monitor operations or plan predictive maintenance and can empower analytics teams to run “what if” scenarios to make more accurate projections involving supply and demand, for example. The technology is gaining traction across a number of industries, helping companies glean timely, actionable intelligence about key business operations.

Pandemic pivot

In addition to graph database technology, JLR uses Google Cloud Platform’s BigQuery, Matillion ETL, and Qlik’s Attunity data replication services, as well as Tableau to surface information to the business. JLR also has a mix of traditional legacy systems and open source tools in the Python ecosystem to build data analytics, machine learning (ML) models, and UX applications, Powell says.

JLR’s graph system can combine 12 data sources in a graph equivalent to 23 relational tables, spanning auto parts supplied by hundreds of suppliers through the particular model and configurations’ bill of materials to the manufacturing build sequencing and order forecast for those cars, Powell explains.

To enable that, TigerGraph is installed from the Google Cloud Marketplace and directly feeds the output of the queries back into the enterprise data warehouse powered by Google BigQuery, enabling the queried information to be integrated with the carmaker’s Tableau reporting system and produce quality results in days, Powell says.

Naturally, the pandemic forced Jaguar Land Rover to rethink its traditional way of using data after its factories shut down from March 2020 to May 2020.

“We had to reorient the way we thought about business forecasting and planning in a completely different way because I don’t think we’ve ever started from nothing,” he says. “That was the big difference, right? We never shut down everything and then started everything from nothing.”

Because forecasting relies heavily on historical data, this disruption had a profound impact on JLR’s operations.

“You couldn’t use the past to forecast the future,” Powell says. “Before, we had monthly reporting models about what cars we were selling each month. Everything was stable. Then suddenly, we had to start thinking about daily sales. What happened yesterday was never a question [for] a company like JLR with its 12-week order planning systems.”

Still, JLR, which has been owned by Tata Motors since 2008, has been aiming to become a “100% digital company” well before the pandemic and is using its graph platform to manage today’s more urgent issues.

“We were able to respond in our supply chain when, for example, dealing with the big elephant in the room right now: semiconductors,” Powell says. “There are thousands of chips in a car and even not having one of those might stop me. So we’re using graph to understand where the risks are, adapting to that, and taking action.”