Insurers face ‘black-box conundrum’ as risk modelling evolves: Swiss Re

Sophisticated models can help insurers see risks that historical claims cannot, but fragmented data and legacy systems risk obscuring the picture, Swiss Re’s Jonathan Rake tells Monte Carlo Today.

Key points:
Insurers face ‘black-box conundrum
Legacy systems fragment risk views
Models can flag concentrations early

More data and increasingly sophisticated models will not necessarily give insurers a better view of risk if fragmented systems and poor data foundations leave underwriters unable to understand what sits behind the result. 

“There’s way more complexity in the underlying connectivity,” Jonathan Rake, CEO of Swiss Re Risk Data Solutions, told Monte Carlo Today. “The systems are talking to systems, data is talking to data.”

Rake described this as the “black-box conundrum”. As insurers connect claims and exposure information with external data, models and analytics, understanding the source and purpose of that information becomes increasingly important.

“The future is not necessarily human judgment versus data-driven decision making, but the two working together.”

“Having more data does not automatically mean having better insights,” he said. “The challenge is connecting the data, models and analytics in a way that produces a more coherent view of risk and helps support more concrete, better decisions.”

Untangling legacy data 

Much of the challenge starts within insurers themselves. Data quality and understanding its source and purpose remain significant issues, particularly where different parts of an organisation are working from different information. 

“Having more data does not automatically mean having better insights.” 

“The data foundation is super critical,” Rake said. “Otherwise, you have different areas and departments all working with different data sources.”

The problem can be particularly acute among insurers that have expanded through acquisitions, leaving parallel legacy systems and data sources that were not designed to work together.

“Many organisations have been built on legacy systems, especially those that have had a big acquisition trail,” Rake said. “They’ve been buying companies up and they create a ‘spaghetti effect’ of systems and data. And some may not have adopted industry standards yet.” 

Looking beyond past claims

Fixing the internal data foundation is only part of the challenge. Historical claims show insurers what has already happened, but understanding where future losses could emerge requires them to combine that experience with external data and modelling.

“Insurers need to combine their own exposure and historical claims data,” Rake said. “It needs to be combined with high-quality external and proprietary data as well as advanced models. No one company is solving that on its own. It’s how you combine insights.”

That can help move insurers beyond a backward-looking reliance on claims experience towards assessing the frequency and probability of future losses and how they might accumulate.

“Get insights into what’s the frequency, what’s the probability of these potential losses,” he said. “How do they accumulate? Using multiple modelling perspectives can be valuable.” 

As one example, Rake cited flood risk in Florida, where Swiss Re combined historical claims data with the flood-modelling capabilities of Fathom, the UK-based water intelligence company it acquired in 2023.

“Combining historical claims data with [Fathom’s] capabilities has helped us identify segments with disproportionately high modelled loss ratios,” he said.

The analysis identified locations where modelled flood losses were unusually high relative to premium, allowing insurers to focus underwriting attention on those risks while giving clients a clearer view of where stronger flood protection might be needed.

Risk models need judgment 

Identifying a modelled loss, however, does not determine the underwriting response. Underwriters still need to decide how much weight to give the result within individual portfolios and market conditions.

“We have sophisticated models that need to be understood in the context of individual portfolios, market conditions and insurers’ strategic objectives,” Rake said.

Technology can process increasingly complex information and identify patterns more quickly, but Rake said experienced risk professionals remain essential to interpreting what those patterns mean.

“It can process complex information. It can surface patterns at speed, while experienced risk professionals can provide good judgment and context needed to interpret these insights.”

That combination will become increasingly important as insurers move from periodic reviews of historical losses towards analysis that reflects changing exposures more quickly, potentially identifying concentrations before losses materialise.

Better information could also allow insurers to differentiate between risks more precisely rather than respond to uncertainty by restricting capacity or withdrawing from a market.

“The better informed you are, the more confident you are,” Rake said. “The future, therefore, is not necessarily human judgment versus data-driven decision-making, but the two working together.”

Jonathan Rake is the CEO of Swiss Re Risk Data Solutions.

For more news from Monte Carlo Today, click here.

Did you get value from this story? Sign up to our free daily newsletters and get stories like this sent straight to your inbox.