Delivering analytics and ML innovation at speed and scale
Platform cost reduction
Increase in sales
Increase in NPS score
Storebrand is in the business of helping millions of Norwegians plan for their future through data-driven asset management. To achieve this customer-centric mission, they strove to make data analytics and ML a transformative focal point to enable more relevant and personalized interactions. However, delivering on this goal was easier said than done, as they were hampered by legacy on-premises infrastructure that was complex to ingest large volumes of diverse data and costly to scale analytics. With Azure Databricks, Storebrand has been able to accelerate innovation at scale to delight customers with product recommendations while protecting them from fraud — giving a boost to revenue while lowering overall costs.
Explore, scale and unify
At Storebrand, Jeroen van Zeeland leads the analytics team that is responsible for supporting the entire business with an ever increasing number and diversity of analytics projects – from increasing revenues through sophisticated marketing initiatives to streamlining claims handling and improving fraud detection analytics. “There is an ever increasing demand for analytics projects and products. We want to be data-driven and we have a huge need for analytics that can actually help us make better business decisions,” said Van Zeeland.
One of the greatest challenges Storebrand faced was caused by a fragmented approach to data analytics in the face of ever-increasing demand for analytics projects. Exploring large datasets was difficult due to multiple data formats and poor integration with other commonly used tooling. Van Zeeland wanted to create the ability for Storebrand data analysts to explore large datasets more easily. “We started using Hadoop. It felt like we were building a tank and something far too complex.” Instead they needed a simplified workflow that would also achieve both governance and sophisticated collaboration. “I was looking to change to a cloud based platform agnostic infrastructure that scales quickly so we could consolidate and create a common workbench that would unify teams’ toolsets, processes and projects,” said Van Zeeland.
A focus on business value, not technology
Storebrand partnered with Cognizant and Databricks to unify data and analytics across teams and business areas. “Databricks is core to the ecosystem we have created, it’s the part that changes where effort is directed. Rather than a focus on keeping technology running, we can now focus on how it creates business value,” said Valdas Maksimavičius, IT Architect at Cognizant.
“Together with Cognizant we went through all the compliance and legal hoops,” said Van Zeeland. “We spent time reassuring Storebrand on the safety and security of the infrastructure,” Maksimavičius added, “Vague excuses around security shouldn’t be a blocker. We asked the hard questions, but also ‘Why not?’ We can clearly demonstrate why security isn’t an issue and show that it delivers business value,” explained Maksimavičius. The Databricks Data Intelligence Platform offers a powerful toolset for Storebrand to achieve a faster time to market and an ideal environment for rapid testing and prototyping. “Delta Lake makes it easy to use and integrate with data storage and provides a good cost saving over other previously employed data storage,” says Van Zeeland. “Previously, accessing data would take days, with multiple people involved. It now takes just a few minutes. We’ve also reduced data duplication through combining Databricks, Delta Lake and Azure ExpressRoute, achieving a cost reduction of at least 55% compared with the old setup where maintenance was onerous and we relied heavily on SQL procedures and old scripts that didn’t deliver the benefits of using a formal language like Python”.
Databricks interactive notebooks serve as a central hub to share and trade resources, increasing productivity and making it increasingly easy for teams to collaborate across projects. “Connecting directly has created tremendous increases in workflow, we can be asking for engineer help and yet continuing to work. It was impossible previously,” said Van Zeeland. “The collaboration Databricks enables is actually more important than the technology– and the only model of value,” added Maksimavičius.
From an engineering standpoint, Databricks has simplified workflow, providing both data governance and accelerating data pipelines for downstream analytics and ML. This was a big win for the engineers. The new flexibility and Python user interface has led to better insights and breakthroughs in machine learning.
Satisfied customers, increased sales
With Databricks as the backbone for their end-to-end data analytics workflow, data science and engineering disciplines now work together as one integrated team with analytics connected alongside.
The team was able to quickly create product recommendations for customers across 45 insurance products, boosting customer revenue and retention. “Our Next Best Actions approach has led to an 9% increase in sales where applied and 29% higher customer satisfaction,” said Van Zeeland. They’ve also tackled common claims handling fraud with analytics. “When a customer has a valid claim it needs to be paid ASAP – that’s beneficial for everyone, but fraud is a real and expensive issue and every claim needs to be checked,” explained Van Zeeland. “In terms of the insurance claims process, in some instances we have reduced a multi-week process down to a single day. This has reduced costs. Less time is spent on each claim, but also payout is speedier, a significant benefit for our honest customers.”
By using the Databricks platform, Storebrand now engages with customers more intelligently, “Databricks has given us a simple unified platform that enables effective collaboration, speedy development, increased oversight and simplicity while retaining the freedom that our data-scientist and engineers so much enjoy,” stated Van Zeeland. “The ability to use data effectively for the benefit of the business is key to success and with Databricks that’s a great position to now be in.”
