Internal data holds untapped potential for insurers, says Munich Re

As insurers grapple with a multitude of emerging risks, expertise and judgement remain crucial in properly leveraging data.

“I tend to say it’s like a car without a proper driver inside,” said Thibault Imbert, Senior Insurance Solutions Manager for Munich Re speaking at InsureTech Connect Asia (ITC Asia) in Singapore last week. “I think expertise is core. We have analytics, it’s nice, but it’s nothing if you don’t have the proper expertise and judgement.”

Insurtechs or insurers looking to enter new markets may have to overcome the challenges created by the limited availability of data, so competitive intelligence plays a crucial role pricing risks as market players look at trends and competitors, said Imbert.

“Competitive intelligence is key as a sense check,” he said. “I can have whatever I have in my models, I can have the best model in the world, outperforming everything that has been previously done, but I think what the market is doing and how it should be is really important.”

Insurers must ask some fundamental questions.

“Does it make sense, whatever I am offering, in terms of rates? And do my rates make sense?” Imbert said. “Understanding the market is fundamental, because I need to go launch and train and then readjust, and competitive intelligence will help me (with) this.”

Insights into customer behaviours and risks have traditionally driven how business is priced, said Elaine Tan, Assistant Vice President and Regional Actuary at MSIG. “Apart from looking internally, we often also look externally to build our competitive intelligence,” Tan said.

Tan underscored the need to exercise judgement in understanding the limitations and potential of available data and assessing where valuable insights can be derived or where data may be insufficient or unreliable.

“When our dataset is a bit more limited, then more simplistic models might actually be more efficient,” she said. “You don’t want to use a sledgehammer to crack open a nut. So, you always need to understand what kind of data you have and, from there, move to select the best data model or technique.”

Insurers and insurtechs can often look inward for data — from policies and claims — to enhance their understanding of the market and price risks, said Imbert.

“What we notice is that there’s untapped potential in internal data,” he said. “Sometimes yes, it’s unstructured data, but now with GenAI and large language models we can do a lot of things to structure this data (for actuaries).”

Imbert said that he prefers a structured approach for both internal and external data sets.

“There are a lot of public sources of data,” he added. “Of course, (they) will never replace internal data, proper insurance data, but having this extra layer is important. Purchasing external data also brings value, but some technical questions cannot be overlooked. How do I collect (this data)? Is it being maintained?”

Capture every single data point

To prepare for emerging risks, insurers should also have processes in place to capture every single data point available, said Peter Tilocca, Head of Underwriting at Life Insurer NobleOak. “It could be hundreds. It could be thousands,” Tilocca said.

“From an underwriting rules engine perspective what can you capture there? What did you capture in the policy administration system and what did you capture in the claims?”

“Combine the three to get a great understanding of the person and the risks that they pose. Without digital capabilities (it would be difficult).”

Insurers have not spent a lot of time on emerging risks, like how global warming has affected mental health. Tilocca cited research from the American Psychological Association that found that the risks involved in global warming will be dramatic from a human perspective.

“From a global warming perspective, most people relate that to general insurance, P&C, weather events, flooding and what have you, and that’s pretty fair,” he said. “But what studies have found, is that as the global temperatures increase, the levels of aggression increase for humans.”

Alongside anxiety and depression, that would pose a risk in markets like Australia, a market heavily dependent on disability type products, where around one in four customers applying for life insurance say they have a mental health issue. “That is a significant risk from a life and health perspective,” he said.

Preparing and investing in preventive care has also helped. “There’s multiple points of intervention,” said Lee Sarkin, Chief Analytics Officer in Life and Health for Asia-Pacific, Middle East and Africa at Munich Re.

As customer interventions for life and health wane after the purchase of an initial policy, this has led to a rise of various wellness interventions. “Can we play a role in actually moderating the risk and actually improving societal health outcomes over time?”

Tilocca said predictive analytics with AI help identifying people who were more prone to mental illness. “Post-natal depression is a known thing, and the partner would also be impacted as well,” he said. “We saw a spike in the data in terms of when depression hits.”

Predicting these trends more accurately for policyholders would mean offering services that mitigate these risks and lower them, he added.

Ultimately, to manage emerging risks, insurers will need to take a little bit more risk, as using AI models may increase the risk of error.

“The first place risks emerge is where models make errors,” Sarkin said. “In particular, they make the wrong underwriting decision.”

What will eventually be critical for insurers is that senior underwriters will have to sign off on models, especially for use cases that might increase future claim risks.

“The discussion has now gone significantly beyond data science teams, to a path to production of a model that will not deteriorate our plan,” he said. “And that’s really where we bring modern technology together with our knowledge to do that.”