For the past three years, a data team in PFA has worked to build more dynamic offers using the full toolbox of data science. Now, PFA has teamed up with NNIT to scale their efforts and create a more efficient enterprise-level framework with a clear blueprint for implementing data science operations.
With more than 1.3 million customers interacting across a multitude of touchpoints and channels, the pension giant PFA has access to a huge quantity of data. For three years, a dedicated 10-person data science team has worked diligently to utilize that data to create more dynamic offers for customers and more efficient internal processes.
The objective of these data science efforts has always been to provide the company with an advantage in an increasingly competitive market. As PFA senior data scientist Johannes Waage puts it, it's all about managing the offer chain to expose the right customer to the right offer at the right time:
– Even though we are a 101-year-old company with fairly safe and predictable products, the pension industry has become a lot more competitive over the last decade – and data play a big part in that. We want to use the full toolbox of data science to create value for our customers and avoid bothering them with irrelevant communication, says Johannes Waage.
Building an organizational data framework
For the first couple of years, the PFA data science team has enjoyed a high level of autonomy, spearheading business-driven projects in an organization with a fairly conservative approach to financial IT – a natural consequence of managing 600 Billion DKK worth of pensions. Now, the team is working with NNIT to bring PFA's data science efforts to the next level.
– We need to professionalize our data science setup and move on from our initial startup mode. Not in the sense of moving out of the lab, because we have a well-functioning platform and a lot of production models, but more as an organizational process. To avoid getting slowed down by important, but time-consuming requirements for governance, compliance and operational stability, we must create an enterprise-ready framework for implementing data science operations, says Johannes Waage.
The framework built by NNIT and tailored to PFA's needs will serve as a blueprint for taking products from development to production, providing a better internal overview of roles, profiles and skills needed. The framework will also define responsibility for tasks such as server management, model development, source control and testing, ensuring that non-data science tasks can be effectively delegated outside the data team.
A shared mindset
With years of experience and a well-rounded mix of skills, including machine learning, data science, analytics and operations, the PFA data team has already gained a high level of maturity. Still, the team is thrilled to join forces with NNIT.
– It is a godsend to work with people from NNIT who not only possess the necessary knowhow, but also share our mindset in terms of an entrepreneurial approach to IT. It provides a strong foundation for anchoring this in a solid way in terms of enterprise and operationalization. And since our systems are hosted by NNIT, we have access to inside knowledge about the day-to-day operations, says Johannes Waage.
Claiming the data mandate
The cooperation between PFA and NNIT also has cultural benefits. Because data projects often have a hard time fitting into existing frameworks and challenge conventions, they affect the relationship between the data team and the rest of the organization. Having NNIT onboard as a third-party partner provides validation of the perspectives and choices made by the team and helps manage expectations with other stakeholders.
– You have to be opinionated about best practices and which tools to use. If you lack this centralized tooling governance, you lose efficiency. The process with NNIT has helped us align with the rest of the enterprise, consolidate the issues and claim the mandate to run data science, Johannes Waage says and continues:
– The biggest challenge is not architecture or math, but ensuring a well-managed process, so you have buy-in, a clear roadmap and well-defined KPIs. Data science is not an IT discipline, but a separate business discipline, which requires a unique skillset with participation from both IT and business.
NNIT Scale Data Science
Scale Data Science is an offering from NNIT focused on business consultancy and platform-neutral AI governance, built in close collaboration with their customers.
Scale Data Science covers:
- Culture and mindset
- Methodology
- Platform and tools
Scale Data Science is part of NNIT's program for corporate entrepreneurship, Business Innovation Growth (BIG).
Read more about Scaling Data Science and Co-Creation with NNIT
