Beware of blind spots

Insurers are driven by data. The ability to model data and use data-driven models to predict outcomes allows them to both help their clients reduce the potential for future loss, and for the insurer to price insurance more accurately. Even when faced with perils that occur regularly, the insurance sector can still be caught out. Despite their experience, insurers still suffer from blind spots.

When that happens at a catastrophic level, with industry losses in the billions of dollars, blind spots can lead to extensive losses. Furthermore, those voids prevent insurers from gathering the essential learnings to better understand these complex events such as hurricanes, hail and flood. 

We are reaching the end of the 2024 hurricane season in the US, the period from June to October when major hurricanes occur in the Atlantic Ocean, impacting the US and Caribbean. In July 2024, Hurricane Beryl made landfall in Jamaica. It was the second named storm and the first major hurricane of the 2024 Atlantic hurricane season. US insured losses from the event are estimated to be anywhere between US$2.5bn and US$4.5bn. 

This was just the start of a season that scientists had expected to contribute to natural perils losses for insurers that are now exceeding US$100bn a year. That represents a doubling of the average annual natural catastrophe losses over the last 30 years. They are expected to double again over the next ten years. 

The good news is that, alongside these predictions of increased losses, industry developments mean those blind spots could be receding.

“We already have cat modelling strategies. We use risk management solutions but surprises still seem to get us, every year,” admits Deepak Badoni, Co-founder and President of EigenRisk Inc, noting that, despite the sophistication of those models, Beryl still caused anxiety across the insurance community. 

The reason insurers are still often in the dark is that the models being used are considered to be limited in application for some situations. “The word model is typically synonymous with the big, probabilistic models with a lot of detail and uncertainty. The problem comes when you try to use those models in the context of live events,” Badoni warns.

The answer, he suggests, is to go back to basics and build out alternative ways of looking at the risk. “We do much simpler, deterministic modelling. We focus on getting the best possible view of the hazard and having simple assumptions on top of potential damage and then, finally, a full financial model to pull it all together.”

Getting the best possible view of the hazard during a hurricane season requires more than just observing the current hurricane. It means delving into a deeper understanding of what could happen, given a broader set of variables that are impacting the range and likelihood of hurricanes and which change day by day. 

This is where companies like Reask come in. By leveraging advances in AI and numerical weather prediction and modelling to train algorithms on proprietary data sets, Reask believe it’s possible to model individual events better. Furthermore, such accuracy can be leveraged for wider insurance considerations such as parametric. 

“Understanding events at a [single] event level is quite limited in existing modelling, so could [we] simulate these events in different environments that can tell me about what could happen?” explains Reask’s CEO, Jamie Rodney.

Rodney goes on to explain how Reask takes this data and then combines it with risk analytics. By creating events at a very high level of granularity, these can then be simulated in different climates to show, across the whole risk curve, how likely there is to be a Cat5 in Florida for example, given different climate conditions. Other questions include: are there likely to be bad seasons in Japan and the US at the same time? 

“With event level [data] and risk level [data], you can now create data sets that allow people to quantify the uncertainty in between” Rodney reveals. 

Understanding how various Nat Cat events are likely to behave is one thing, being able to translate it into the language of the insurer, and the underwriter, is another. “The value of the EigenRisk platform is that it [shows] Reask’s wind speed says this, it means this for a particular use case, whether that’s underwriting, loss cost, how it will affect the portfolio, how much should be reserved.”

Marrying data points in this way seems simplistic yet it is bringing together information in a way that sheds light on previously unanswered questions. It’s revealing those blind spots. 

Rodney explains that a client may wonder why they have such high claims in certain location. With the addition of the wind speed data it’s possible to see that those claims happen in a particularly exposed area and are more vulnerable. One finding from Hurricane Beryl was that although that event itself was not the most ferocious to hit the US, it had impacted densely populated areas, so the total size of loss was higher. 

Critical to building an accurate picture of claims exposure is making sure data is easily accessible and comparable. “We have built a technology ecosystem that allows all our data to sit close together – our bound portfolio, claims data, third-party data – is all accessible together,” reveals Tim Spencer, Head of Analytics, Vave. 

“We upload via API on a daily basis which means that when there’s an event coming, we’re not scrabbling to assimilate our exposure from various books of business,” he continues. “We have a really good idea pre-event and immediately following it what locations are at risk and we can make those initial loss estimates.”

Using that information to inform future pricing is equally important, looking at how well certain risk attributes performed when applied to underwriting. “The claims we’ve incurred helps us learn really quickly and we can pass that learning back to the algorithm so we can update our pricing faster and be more resilient for the next event.”

Understanding single events is clearly important but the data has just as much value in building future resilience. “As you start to unearth correlations that you never saw before – how wind speeds impact other phenomena, for example – you can get a pretty good predictor.

“If we can model events accurately, we could create a three-to-five-day early warning system which unlocks value we didn’t have before,” Rodney adds. “Our industry is very supply and demand driven. [Insurers can now] accelerate their loss adjustment approach ahead of time and reduce cost. If you’re prepared before an extreme weather event, that cost to benefit ratio could be as much as one to 13.”

Increasingly, clients are looking to their insurers to provide answers before the fact. Derek Thrumble is Managing Partner at Alesco Risk Management Services. His clients are always looking at new areas for their business, particularly in the field of energy and renewables. Making the transition to renewable energy is receiving a push from governments, it’s important that organisations can make sizeable investments with their eyes open.

 “They’re being exposed to new asset classes. This can often mean the way the more traditional perils have been modelled doesn’t really fit their risk profile well,” Thrumble reveals. “With a transparent modelling process in their locations, peril by peril, and converting it into damage and loss estimates, we can generate probable maximum loss numbers with a real audit trail.”

While this is primarily aimed at longer term investments – in contrast to an insurance sector that typically works year to year – clients are also making use of these insights in real-time – “they like to see what’s coming their way,” Thrumble adds. 

Have all blind spots finally been eradicated? Not quite Thrumble says, “There’s a huge amount that could be done. We’re talking to risk managers who are only doing this for a small part of their job because most of their role involves buying this year’s insurance, rather than [preparing] what they might need in five or forty years’ time.”

“One blind spot at a time,” Badoni agrees, “we still have some ways to go.”
For more detail on solving insurance’s blind spot problem, watch our video interviews with Deepak Badoni, Tim Spencer, Jamie Rodney and Derek Thrumble from our ‘Climate and property blind spots revealed’ event here.