One way to think about the application of data science and machine learning is that it’s a tool to aid the conversion of information (data) into action. In this context, machine learning is applied to enable better and more efficient decisions, as well as identifying previously hidden risks and opportunities. Essentially, data science helps an insurer to perform significantly better, whatever their goals.
The application of advanced analytics is already well ingrained in the world of insurance pricing and underwriting. However, it is only more recently that it has begun to exert more influence in claims operations.
In the overall insurance value chain, substantial resources and effort have been applied to better understand a customer’s risk and purchasing behaviors to help charge the most appropriate price. Fresh benefits still to be mined in the pricing and underwriting space are relatively scarce. In contrast, huge untapped value is waiting to be realized by insurers reducing their claims spending or better understanding and optimizing their claims processes.
Low-hanging fruit
Although machine learning is increasingly recognized as a tool to reduce claims costs and deliver significant value to an insurer, this remains an area where many have yet to realize value. This means there is plenty of low-hanging fruit to be picked in the claims space, such as the benefits to be realized from providing a better, more tailored, faster service to the customer. These benefits can, for example, be seen by the speed at which claims are settled and how an insurer’s Net Promoter Score (NPS), the global benchmark for client satisfaction, can be improved.
Claims processing already uses a lot of external data, including integration into third-party sources such as operators in the automotive sales market for vehicle values, demographics and sociodemographic information, and various other vehicle information to inform repair costs. Machine learning makes it possible to link all these separate threads and help insurance companies more accurately predict future outcomes and identify earlier changing experience.
Internal impact
There is also the positive impact on the internal organization that has the potential to be equally transformational. Machine learning can be thought of as a tool, a superpower to help claims handlers and claims teams make better decisions. Individuals can upskill, and new roles will be created, all helping provide measurable improvements to customers and vastly improved profitability.
At the same time, it is important to understand that machine learning will not give the perfect answer to every question. Each individual algorithm built will have both strengths and weaknesses. That being said, it is still possible to build and improve models based on an understanding of these strengths and weaknesses. More importantly, it is by understanding how best to leverage what an insurer has, as well as how best this can be applied and integrated, that will determine the value gained.
See also: A Behavioral Science Scandal
Collaborate or fail
This is especially true when it comes to using data science to leverage unstructured data. Using an insurer’s deep domain claims expertise is key to shedding light on unstructured data and translating this into something that actually makes sense. On the application of data science in claims operations, by far the greatest risk in terms of success and failure is the ability of both sides to collaborate effectively. By bringing together an insurer’s in-house claims expertise with their data science and machine learning experts, it becomes far easier to approach problems in a way that leads to a joint successful solution.
Near future
It can be tempting to focus on the short term and doing whatever is needed to make one solution work once. But it is worth keeping in mind the end state, where one insurer’s claims models will be competing against another insurer’s models. In a world where hundreds of models are competing, the ability to move at speed, scale for efficiency and be the most sophisticated will be needed to succeed.
Data science is not the absolute, all-encompassing, magic solution to every issue an organization will face. Instead, being able to fully leverage machine learning means bringing together a multi-disciplined team that combines an insurer’s existing in-house claims knowledge with cutting-edge analytical and data capabilities to deliver next-generation claims processing that optimizes costs and transforms the customer experience.
