Data analysts and actuaries have been around practically for centuries to bring mathematical models and a data-driven approach to the insurance industry. Yet the wave of new technologies and techniques available today — including data science, machine learning, and artificial intelligence (AI) — have the potential to propel the field even further, delving into more sophisticated (and accurate) predictive modeling techniques for everything from claims prediction to fraud detection.
Building the Team
We spoke with Tom Spencer about how he built the Customer Data Science Team at Aviva from the ground up to build the company’s AI initiatives, as well as with Ayca Kandur about what the team is like and how they are able to execute so efficiently today.
Although we say we are a center of excellence, a lot of our work is done side-by-side with line-of-business teams to make sure we’re not in an ivory tower — that we’re working on things that matter and projects that matter.
Tom Spencer Head of Customer Data Science at Aviva
It all started two years ago when Spencer joined Aviva to lead the Customer Data Science Team, which is a data center-of-excellence heavily focused on a consultative approach. Hiring a team of data scientists from scratch today is no easy feat — the shortage of data talent (and data scientists in particular) is well documented.
But Spencer’s philosophy around the role of data scientists as partners with domain experts allowed him to ultimately find the right staff. Data scientists at Aviva work quickly using agile methodology in order to move the needle, but perhaps more importantly, they also understand and make sure to maintain a deep connection with the business.
The business connection is particularly important for the Customer Data Science Team at Aviva because most of the data scientists don’t actually have prior experience specifically working in insurance. This approach allows data scientists to bring creative solutions, new techniques, and new data sources to the table, while those on the business side bring extensive industry knowledge.
The answers often aren’t from insurance, they’re from other disciplines out in the world. Our job, then, is to partner with our colleagues that have deep insurance expertise for a data-driven outlook.
Tom Spencer Head of Customer Data Science at Aviva
Keys to Success
At Aviva, good data science and a productive data science team come from:
1. Good data. Upstream to downstream, one of the most important contributors to great data science is great data. For Spencer and his team, that means not only high quality raw data, but having a staff that knows what data is, what it means, and where it comes from.
2. Focusing on the customer. One of the Customer Data Science Team’s most celebrated projects is ADA (Algorithmic Decision Agent), Aviva’s personalization AI, which helps the company be more specific and relevant to its customers. The AI, built using Dataiku, helps the company understand its customers better and delivers tailored marketing experiences based on their needs.
As a customer data science team, we’re always looking at how we can make things better for customers. And happily, that also tends to drive profit.
Tom Spencer Head of Customer Data Science at Aviva
3. Proper tooling. When Spencer started building the Customer Data Science Team at Aviva, his first priority was getting great people, but a close second was getting the tools in place that would allow that team to work together and to be its most productive. Today, the entire data science team uses Dataiku for every step of the data pipeline, from connecting to data to data preparation, building models to deploying them to production, and everything in between.
“The most beneficial thing about Dataiku is having everything in one place, so you don’t have to go from one program to another to another and have them work all at the same time. Dataiku takes away that hassle.
Ayca Kandur Data Scientist at Aviva
4. Staying grounded in results. The team at Aviva recognizes that data science is neither fun nor useful if the business ultimately isn’t actually using what they’re producing, so they have a strong focus on pushing to production (not just playing in a sandbox) for real impact.
An important thing for us culturally is that we are absolutely part of the business. Our drive is commercial results — we’re not trying to build things that are really advanced to make us look good. We’re trying to drive business results.
Tom Spencer Head of Customer Data Science at Aviva
5. Staying Agile. Delivering value fast is also important to Aviva, which means the Customer Data Science Team strikes a balance between quick, exploratory R&D and more structured push-to-production strategies.
For us, when it comes to very quick R&D work, we like the fact that we can use Dataiku to have a very quick model and see the output. […] It just speeds everything up if you have very limited time to see if this approach is going to work or not or make a quick decision to see which way we need to go.
Ayca Kandur Data Scientist at Aviva
I didn't want to limit my talent acquisition to just Python or just R — I was looking for a workbench that was data science-ready so we could do some of the advanced techniques that other vendors wouldn’t provide, something that would level the team so that we could collaborate on the same platform, and something where we could increase overall effectiveness.
Tom Spencer Head of Customer Data Science at Aviva
Aviva & Dataiku
When building his data science team, Spencer wanted to make an investment in the team that showed both the new staff (as well as the rest of the business) that data science was at Aviva to stay and that it was a serious endeavor. Before choosing Dataiku — the platform supporting agility in organizations’ data efforts via collaborative, elastic, and responsible AI, all at enterprise scale — there was limited sharing of knowledge or information and lots of individual work on laptops.
Spencer estimates that projects that would previously take two weeks now take the team two days using Dataiku. ADA, the company’s personalization AI, is no exception.
When we started building ADA … it was taking us quite a long time with the existing legacy systems. But through using Dataiku and the API functionality, we reduced the amount of time from beginning to end to build a model and push out the model into the marketing channel.
Ayca Kandur Data Scientist at Aviva
Ultimately, the data scientists at Aviva enjoy working with Dataiku, which Spencer and Kandur consider to be the “gel” for the team. Given the complex industry and difficult problems Aviva’s Customer Data Science team is tackling, it simply makes their lives easier and more efficient.
If you have the wrong tools in place, you can fly solo – you can get away with inefficiencies and hide your mess a bit. Dataiku changed our team atmosphere and culture for the better through sharing capabilities.
Tom Spencer Head of Customer Data Science at Aviva
