A new type of insurtech sandbox licence has been granted to Bermudian-based start-up Kettle Limited, which is using deep-learning and sophisticated statistical models with the aim of more accurately predicting wildfires in California.

Kettle Limited is a subsidiary of Our Kettle Inc, and has been granted a Bermuda Innovative Agent Licence.

It will use Our Kettle’s machine learning technology to analyse data from 47 different sources, including Nasa satellites, weather satellites and laser-based Lidar mapping sensors, in order to predict more accurately the likelihood of California wildfires in any given square mile.

By doing so, it aims to better predict risk of loss and thereby better source, price and distribute reinsurance coverage to insurers across the state of California.

Andrew Engler, one of the co-founders, is an insurance industry veteran who knows Bermuda well, having lived here for a number of years when he worked for Argo Group.

"Bermuda is a second home for me. And Bermuda is at a point now where it has a massive opportunity to be the epicentre of helping to protect against climate change,“ he said.

Mr Engler regards insurance and reinsurance as the most important tool to protect from climate change. While Kettle is utilising innovation and technology to better predict wildfires in California, it has a goal to target other regions and perils, such as storms and floods.

Team-building in Bermuda is under way. The first announced appointment is William Haddrell as chief underwriting officer. He was most recently with Sompo International.

Looking ahead, Mr Engler said the vision is to bring together the best technologists from the US and the experienced underwriting talent found in Bermuda.

He also answered the following questions from The Royal Gazette about what makes Kettle stand out in the insurance and reinsurance world, why it set up in Bermuda, and what it is seeking to achieve.

Q. Why did you choose to bring Kettle to Bermuda?

Bermuda is unlike any place in the world. The two most important factors to a start-up are: speed and access. From a regulatory standpoint, Bermuda is by far the most advanced insurance regulatory regime in the world when it comes to setting up a technology first start-up. The sandbox regulatory framework allows start-ups to minimise time and effort in understanding a complex web of regulatory rules, and instead allows for direct access to the governmental offices to execute as fast as possible.

You previously worked with a traditional re/insurer. What made you decide it was the right time to create an insurtech?

Argo was an incredible company to work for and one of the most innovative speciality insurers in the market. But as a native Californian, I had recognised the mass dislocation of pricing currently evolving in the CA market on a personal level when friends and family weren’t able to get insurance or were under insuring their home. The catalyst was meeting a mutual friend, Nat Manning, who had an extensive history in start-ups and a very fateful meeting with Jasmine DeSilva at the Bermuda Business Development Agency. I still remember coming to her with the idea of starting a reinsurance company that utilised the most advanced deep learning methodology being developed for chaos theory, and thinking ‘am I crazy’. It took one meeting with her to recognise how achievable this idea was especially with the support of the BMA [Bermuda Monetary Authority] and BDA.

Is Kettle exclusively focused on California wildfire risk?

For now we are. Our current model utilises 47 separate satellites and weather data sets that equate to over seven billion lines of data, and the vast majority of them include geospatial imagery and climatological factors that are easily exportable to other perils. Also the proprietary deep learning system we developed and patented called ‘swarm neural network methodology’ is incredibly accurate at understanding the probability of event likelihood for anything steeped in chaos theory. While we focus primarily on wildfire today, the overall goal of Kettle is to use advanced deep learning to target any region/peril that is significantly affected by the complex evolution of climate change driven risk. From early R&D we can already tell that convective storm and flood show significant promise.

Is there a danger homeowners in California will no longer take out insurance because of the escalating costs due to premiums surging as a result of the wildfire?

This is already happening at an alarming rate. We have seen an incredibly large amount of policies that are being underinsured or self insured at this point because of the limited options left in the market. The overall tragedy is that this is a completely insurable risk. Last year is the perfect example: 2020 was the third worst year in history for wildfires in CA, where 4.5 million acres burnt (about 4 per cent of the state) and yet out of 14.2 million structures, only 11,466 were destroyed. That is less than 0.1 per cent, yet we are commonly led to believe that almost the entire state seems to explode in flames every year.

By having a cutting edge system that holds an 84.7 per cent accuracy rate in identifying the most dangerous zones in California, we can bring significant stability to the market and help identify which areas are most dangerous and which are being completely misrated.

How are you able to model vastly more simulations than most reinsurers?

Our system runs on the most advanced available technology from Amazon Web Services and Google Cloud Platform. We currently have 32 separate neural networks that work in dynamic unison to horizontally influence one another in real time, much like how a bee hive or a termite swarm communicate with one another.

The only way we can accomplish this level of computation is we own 3,414 separate ‘virtual machines’ that run together as one giant super computer essentially. We are able to rent these machines at a fraction the cost (probably 15-20k minimum each) because of the advances in cloud computing offered by AWS and GCP. This means we can run 43.2 million simulations over the course of an hour instead of months utilising traditional server set-up.

How do you see Kettle's premiums comparing with others for the same risk?

When we are able to surgically price the reinsurance and probability of a wildfire starting/spreading in every single home in California, we are able to get a far more granular view of risk. That said, it's still aggregated into larger treaties like most reinsurance.

There is definitely about 20 per cent of the state that we find to be nearly uninsurable because premiums needed are so high, so those homes are better suited for the FAIR plan. That said, there are extremely large portions of California that are being mis-rated at this point.

The main issue we have seen is that modelling currently takes micro factors instead of macro to correctly rate an area. For example, a home's distance to brush is important, but not nearly as important as understanding the probability of a 70,000 acre wildfire burning uncontrollably within a 0.25-mile radius of that house. Then it doesn't matter if your home is 100 feet or 10 feet from brush. These wildfires can throw embers the size of your fist over half a mile that can burn at extreme temperatures for over five hours. At that point it melts anything in its path. This allows us to have a far more favourable rate for many areas that are being misjudged right now and bring stability back into the market

How is Kettle doing with regards to attracting funding? And is Bermuda and its reinsurance sector playing a key part – or is likely to in the future?

We feel quite blessed in terms of funding and how our fundraising has gone. Our entire premise was ‘Reinsurance is literally the last line of defence against climate change’, and when we went out to the top venture capital firms in Silicon Valley and were overwhelmed with their response.

We have received funding from some of the top VCs in the world that helped build everyone from Space X, Facebook, Peloton, Tesla, Plaid, and so many more. Having this backing has given us two incredible advantages. We are able to pull the absolute best talent from the most groundbreaking teams in Silicon Valley as well as knowing our capital is focused on us trying to accomplish our incredibly large mission to essentially create the greatest safety in history against climate change.

Even more gratifying has been the incredible support of the traditional reinsurance market in understanding the role we want to play in protecting society. Without this absolute marriage of the most advanced areas of technology in SV and the most respected names in reinsurance, it would be impossible to take on such a big mission.

Where do you see Kettle going as far as a Bermuda presence (offices/staff)?

Bermuda is absolutely critical to our movement forward. We have already made two on-island hires and will significantly expand our footprint over the next five years. The real excitement to me personally, as someone who absolutely loves Bermuda and the culture, is the ability to bring some of the brightest minds of Silicon Valley into Bermuda to help create what we are lovingly dubbing ‘silicon island’.

Nowhere in the world has a higher concentration of reinsurance expertise, knowledge as well as innovative thinking. This presents the perfect opportunity to have some of our team who have graduated from places like Stanford, Columbia, Harvard, and worked at firms like Google, Space X, Nasa, and many more to come to the island to share their knowledge. By bringing these two worlds together, we are able to launch an ecosystem that the insurance industry as a whole hasn’t seen before.

Kettle has a website at https://ourkettle.com/