Tom is a Lead Data Scientist currently working within our home actuarial and data science (AnDS) team. This week, he shares a valuable insight into the fascinating world of big data in home insurance and the type of work keeping the tight-knit home data community busy.
We have big ambitions for our home insurance product, which is why we’re growing our home actuarial and data science (AnDS) team. They’re our colleagues who’re dedicated to building models that utilise the highest quality data and the most innovative methods available – rapidly deploying them as the foundations of our technical pricing.
The state of the art
Drawing on experience from inside and outside the sector, AnDS at Hastings has moved beyond off-the-shelf tools, we prefer to get under the hood and engage with our tech stack hands on. Utilising the best of SQL and NoSQL approaches, we curate our own tailored data landscape, and use it to develop our models using a Python, R or Spark based framework. These open-source tools allow us to leverage the most up to date and innovative methodologies whilst also giving us greater transparency and responsibility for our products. The best part is that if we can’t find a solution for our use case, we can just build one – learning lots about production-quality code and best-practice deployment as we do so.
A day in the life
The home team at Hastings are a tight-knit bunch, with the AnDS function working in close collaboration with Risk Pricing, Account Management and Foot-printing. Each day starts with a stand-up meeting in which progress is discussed, blockers are dealt with and ideas cross-pollinated. Then, for the AnDS team, it’s on to project work. This may mean reviewing a potential data source for predictive capabilities, training a machine learning model to classify the latest market trends or monitoring the performance of a recently deployed model.
The scale of projects we work on varies from three days to three months or more, but all follow a common theme. We first agree the scope of the project and the ‘definition-of-done’. Then we collect the necessary data ingredients (engaging with the relevant parties around the business to do so), design a modelling approach, implement this approach in code and then iteratively test and review the results until they’re accurate enough to be published or deployed. Whilst this recipe may sound simple, in truth our projects are rarely trivial and each comes with its own data intricacies and modelling complexities, which require us to architect solutions at a very high level of sophistication.
Pretty big data
Our home insurance data relates to themes including quotes, policies, claims, individual residences, individual people and much more. This scale of data requires us to think very clearly about how we would like to process and model it, and for that we use cutting edge approaches such as scalable cluster computing and highly parallelised processing. Our datasets aren’t getting any smaller either – quite the opposite – and this forces us to constantly re-evaluate our approach and to educate ourselves about the latest developments for efficiently processing data at scale.
Everybody loves a map
One of the best parts about working in home insurance is that geography plays such a large part. Because of this, we’re developing our geospatial analysis capabilities and will go to great effort to understand each customer’s property in the wider context of its local geography. Whether it’s local house price trends, changes in seasonal weather patterns, risk of flooding or emergency service response times, it all matters to us.
A wider community
The home AnDS team forms a small part of a large AnDS community at Hastings. This community is truly special. With members ranging in seniority and background, this group distils the best practices from around the business and meets regularly to learn advanced techniques in data science and machine learning. Using a common technical landscape, this team is able to constantly support and review one another’s work, increasing the quality and confidence with which we deliver for the business.
Join us: If you’d like to grow an actuarial and data science career in a fast-paced company that has access to wide and varied datasets, find out about available opportunities including our Geospatial Data Scientist and Senior Actuary opportunities.