Earlier in September, AXA completed its acquisition of XL Group Ltd, a leading global Property & Casualty commercial lines insurer and reinsurer. With this acquisition, AXA has added XL Group’s premier speciality and large corporate P&C platform to existing commercial lines insurance portfolio.
While the combination of AXA and XL Group is going to strengthen the offerings in insurtech space, Analytics India Magazine caught up with Indranath Mukherjee, AVP & head of strategic analytics of the newly-christened AXA XL to understand how the buyout will affect overall analytics strategy, the adoption of data science & analytics in insurance industry and growth plan of the company.
Mukherjee boasts of more than 16 years of industry experience with a strong hold of statistical expertise and astute business knowledge. A well-known industry veteran, in his present role at XL Catlin, Mukherjee manages a team that uses the most advanced machine learning techniques to develop business solutions.
AIM: Having worked in the field of risk analysis and insurance, how do you see the role of data science evolving in these fields?
Indranath Mukherjee: Over the last two decades, the role of data science has evolved significantly in the insurance and risk analytics space.. The insurance industry has always been data-centric. Traditionally, the actuaries have been the custodian of all the data related work in insurance. This included work mainly in the areas of loss cost pricing, IBNR reserving and capital modelling. That has changed over the years. Actuaries are still deeply involved but mostly in the work streams which require regulatory reporting. Data science or if I may use the term analytics is being deployed today in almost all the insurance companies in multiple divisions including marketing, underwriting, claims and operations.
AIM: How has the analytics and data science adoption in insurance sector changed over the years, especially in India?
IM: Although insurance is very data centric business, it has been slow to adopt data science compared to banking or payment industries. This is true for both India and other countries. But over the last decade or so, insurance carriers have also gotten their act together.
I remember having a conversation with one of the Indian insurance company in early 2008 when their biggest challenge was availability of quality data to do any analytics. Having taken a small step with developing a response model for one of their marketing campaigns then, they now have a fairly robust analytics ecosystem in the organisation.
AIM: What does your role as head of strategic head at XL Catlin involve? What are the various analytics solutions you have worked on over the years?
IM: My role as the head of Strategic Analytics team in India is to lead a team of data engineers and advance modellers to assemble datasets suitable for modelling, develop and implement machine learning models by improve efficiency and decision making.
Given the fact that our business is mostly in speciality lines, large part of the work revolves around underwriting where we assess some of the most complex risks in the business. Outside of underwriting, we also work on claims.
AIM: How is analytics used in various functionalities at AXA XL? Would you like to highlight a few use cases?
IM: AXA XL has a number of initiatives around analytics where data-driven innovative solutions are being developed to solve business problems. The Strategic Analytics group works on projects which are implemented to drive business values.
I will highlight a typical use case of underwriting models. Growing profitable business in a competitive market is a huge challenge for all insurers. To improve overall returns, rather than wait for the market to harden (either by natural market forces or as a result of a major catastrophe), insurers can seek to gain a competitive advantage through getting carefully targeted, profitable new business onto the books, achieving a superior risk-adjusted price for each risk bound, and improving retention levels and hence lifetime customer value. We developed a number of machine learning models for risk segmentation to achieve superior predictive performance.
AIM: What are the kinds of data science problems that need to be solved and how do they connect to AXA XL’s business outcomes?
IM:. From marketing, underwriting and pricing to claims and operations there are business problems which are being solved through data using the right tool and applying appropriate techniques.
Continuing with the use case I talked about in answering the previous question, the underwriting models we develop are used by our underwriters to assess the risk before acquiring or renewing a client.The understanding of the model results is a key for the end users. Our models help us acquire / renew profitable clients and charge appropriate risk-adjusted premium for the risk being undertaken. This has direct positive impact on the loss ratio numbers.
AIM: Which particular data science techniques are useful in this kind of domain?
IM: In the current competitive market using only linear models to solve predictive modelling problems is not enough. The trend is moving towards using the machine learning and deep learning algorithms to solve predictive modelling problems because of their low bias, higher predictive power and dependencies on more features.
AIM: What are the various kinds of analytics tools that are used at AXA XL?
IM: We are tool agnostic as a team. However, for legacy reasons we still are heavy users of SAS. We have also started using R and Python fairly intensively.
AIM: How big is your analytics and data science team? What are the various kinds of role that you recruit for?
IM: We are a 25 members team globally and growing. Our team has three pillars – business liaising, data engineering and modelling. Since we have very little business in India, our business liaising folks are in USA and Europe. Currently, we recruit people for data engineering and modelling.
AIM: As a seasoned data science and analytics professional what advice would you have for new data scientists working in the insurance and risk sectors?
IM: My first suggestion will be to understand the domain well. This would help drive clarity in understanding the business problems that they would be working on. Another critical aspect is data. I cannot emphasise enough how important it is to spend time in getting to understand what each of the data elements mean. Tools and techniques are important but not sufficient to become a good data scientist. They can be as good as we apply them in solving the problem at hand. Hope this answers the question.
AIM: What are the various challenges you see emerging in the space?
IM: The industry is going through a round of massive hype. Funding in the sector has grown up significantly; we have multiple education/training institutes providing trained talents in numbers, we got to be conscious that we don’t blindly depend on tools and algorithms to become smart data scientists. We need to stay focussed on solving problems and helping organisations take better decisions.
Try deep learning using MATLAB