For nearly two decades, the Moneysupermarket Group has helped UK consumers save money on insurance, travel, mobile services, utilities, and other everyday expenses. In 2019 alone, Moneysupermarket sites served 13.1 million active users and delivered an estimated £2 billion in savings to consumers through its four brands:
“Almost 80 percent of online adults in the UK visit one of our websites every year,” said Moneysupermarket’s Chief Data Officer Harvinder Atwal. “That’s more than the number of people who use Facebook in the country.”
According to Atwal, it might seem like an easy job getting consumers to a price-comparison site—after all, who doesn’t want to save money? But it’s actually quite challenging, and involves sophisticated data science to be effective. Machine learning models inform every part of the company's work: optimizing customer journeys and experiences, personalizing offers and content, negotiating pricing with partners, and more. MoneySuperMarket uses the Domino platform to develop and deliver models faster and more efficiently than would otherwise be possible. The Domino platform saves each data scientist nearly 3/4 day each week through fewer manual processes and dependencies on platform admins.
Several years ago, MoneySuperMarket launched a massive technology re-platforming effort, moving from traditional on-premise IT to a multi-cloud strategy so it could more rapidly innovate and unlock new market opportunities while reducing infrastructure costs. However, as data scientists migrated their work to the cloud, they faced bottlenecks spinning up cloud resources and installing the tooling they needed.
"We had a key challenge faced by many companies: How do we give data scientists fast access to the tooling and capabilities they need in the cloud?," asked Atwal. “How do we give them control so they can choose the types of virtual machines they need and deploy their own libraries without waiting for approvals? How do we make it easier for them to share knowledge and bring new data products to market faster? There was a lot of waste in the process—in terms of time, talent, and money—that we wanted to address.”
MoneySuperMarket evaluated several data science technologies including Domino Data Lab, Databricks, and Dataiku. They searched for one that would facilitate their journey to the cloud, centralize data science, and accelerate the pace of model development and deployment. "We chose Domino because it is designed specifically for code-first data scientists and massively reduces friction in their workflows,” said Atwal. “It has become a major part of our DataOps ecosystem, helping us reduce wasted time and enforce better control.”
Twenty data scientists were up and running on the Domino data science platform in a matter of weeks, and immediately saw efficiency gains due to four key capabilities:
MoneySuperMarket runs Domino on Amazon Web Services (AWS) today to support development of ad hoc analysis, dashboards, and self-service tools for business staff across the organization. MoneySuperMarket will soon allow data scientists to run Domino on Google Cloud Platform (GCP) too. This will allow data scientists to build models on Domino that use an event-driven Kubernetes architecture, such as product recommendation engines that are integrated into customer communications.
"Domino’s Kubernetes-native version is designed to support our multi-cloud strategy and will enable us to develop models more efficiently regardless of what platform we use,” said Atwal.
To maintain its leadership position, MoneySuperMarket must ensure its prices remain competitive in the marketplace. Its Commercial team works closely with partners, such as insurers, airlines, and banks, to identify opportunities where they can reduce rates for consumers while increasing profitability. To support this work, MoneySuperMarket's data science team developed a pricing tool on Domino that enables the Commercial team to model the impact that different price reductions might have for its business partners, including how each potential price cut will affect customer conversions and, ultimately, the partner’s revenue.
"There's a tremendous amount of data science that goes into identifying how to reduce prices," said Atwal. "Using the pricing tool, our commercial teams can now have much more fruitful and intelligent discussions with partners. And our data scientists don't have to be involved in every single conversation.”