Before joining BGL, I worked in investment banking and central government. In my government roles I had the opportunity to grow and lead multi-disciplinary teams of analysts, building models to inform decision making across a disparate number of areas (from modeling spend volatility and informing fiscal events, to exploiting data attributes and designing initiatives to identify and prevent fraud).
I joined BGL in 2016 in a dual role, leading the motor and home insurance technical pricing team as well as the life insurance pricing team, before switching to life full-time. BGL seemed (and proved to be) a great place to drive value through modeling and data science, where analytical teams have the autonomy to try new things, with strong support from the rest of the business.
Since joining, I’ve changed the way we price for life insurance, introducing automation, best practice, machine learning, and data science to improve our processes. This has involved integrating different data sources, as well as internal data, into new, in-house, machine learning models to help us understand what drives customers and competitors behaviors, and to inform our pricing strategy.
Most recently, we’ve been expanding both our pricing and data science capabilities. We’re using the resource and knowledge we’ve built up over the past few years to optimize what we do on customer experience, risk modeling and selection, product development and marketing.
People sometimes talk about actuarial and data science and how the two come together, or not. In my head however, when it comes to analysis and actuarial science, it’s all just maths – and maths is great! I think there’s sometimes a risk of being a bit territorial. For example, where does data science stop, and decision science begins? To me, it’s all different ways to address analytical questions and problems using data and analytical tools.
I’m a huge believer in diverse, multi-disciplinary teams, tackling projects together and learning from each other in the process. At BGL Life, the pricing and data science teams work extremely collaboratively, which makes us much more effective than we would be if we were in silos. The important thing to me is the drive to innovate, to challenge the status quo and established ways of doing things – while making sure it’s all underpinned by strong, analytical rigor.
I think there are three things:
It’s important to have an experimental mindset and to fail fast. Learn from your mistakes, be resilient and move on when things go wrong, but also never make the same mistake twice – be open and transparent.
It is also very important to have executive buy-in and investment. You need investment in data and data science to be able to try new things and innovate.
One of the things that’s changed recently is that approaches which typically required heavy coding and programming are almost readily accessible with new technical and data solutions.
That makes it a very exciting time, but it’s also very important to try and understand what these techniques actually do.
What this ease of application means going forward, is an expansion of the areas of where data science and machine learning is being applied. As great customer experience sits at the heart of our business, this enables a more personalized experience for individuals. So there will be more different pockets of opportunities and applications than there are now, including customer segmentation, customer acquisition, fraud etc.
I think it’s very important that these teams are as closely aligned to business strategy as they can be and shape the strategy as much as possible – being there from the beginning rather than at the end, if you like.
Three things. Firstly, go for it, but always try and understand the basics. The fact that you can build a machine learning model with very few lines of R/Python code is great, but if you don’t understand the maths or stats behind it, then it’s very hard to determine whether the models actually work and are doing the right thing. Always try and explain and validate the results of the analysis.
The second thing is to always remember that all models are wrong – by definition! So, the trick is to always find the least worst one. Never fall into the trap of ‘it must be right because the model says so’.
The third thing is to be resilient. Sometimes you will fail, sometimes the models will fail, and you need to be okay with that - always learn and move on.
Within BGL Life, that’s a huge focus. I’m part of the inclusion and diversity group and we hire someone for who they are and we want everyone to be themselves at work. Diversity and inclusion need to be a part of the work environment. Everyone needs to feel that they can bring their authentic self to work, rather than what they might feel they are expected to be. Diversity, in this sense, must be both respected and expected. This also means diversity of thought, experience, belief etc.
In my current role, it was introducing machine learning into life insurance pricing, which is something I hadn’t seen before. I borrowed techniques and approaches from previous experiences to try and enhance what we had in life insurance to better understand customers, competitors, and the driving factors behind their movements and behavior. Using this information, allows us to develop a pricing strategy that drives growth in the business while making prices for our customers more competitive.
