CEO & Co-founder of Kettle.
In the second episode of the new series, I was honoured to be joined by the inspirational entrepreneur Andrew Engler, CEO and co-founder of the insurtech Kettle. Andrew is a fantastically intelligent human being, and has amazing insights into using machine learning and AI in innovative and new ways.
In an ever changing world, impacted by climate change and natural catastrophes unlike anything we have seen in the past, it is essential that the insurance world adapts and use new ways and modern thinking to protect their customers. Kettle is at the forefront of this change, and Andrew shares the way he developed into the person he is today.
Subscribe to our YouTube channel and follow us on LinkedIn to keep up to date with all the Fintech and Insurtech news.
Andrew’s Recommended Read
William Laitinen: Hi, I’m William Laitinen, and welcome to the Talent Equals podcast. In this episode, we’re going to get into how an insurtech can help mitigate the disruption in wildfire prone areas through the use of really innovative AI technology. But first, just to jog your memory of what these fires are actually like for the people who experience them firsthand, we’ve gathered some footage and news reports from Southern Australia and the West Coast of the US.
News Report Speaker 1: In Nowra, the fire jumped the Shoalhaven River.
News Report Speaker 2: I’m scared witless.
News Report Speaker 1: By the side of the Princes Highway the main escape route was a bonfire.
News Report Speaker 3: Tonight in California fires burning at both ends of the state, blazing temperatures and strong winds giving birth to this a massive firenado.
News Report Speaker 4: It is incredibly dangerous because of how erratic the fire behavior is. I mean, think about it, you actually have a whirling column of air on fire.
News Report Speaker 5: Meanwhile, 1000s of residents in Oregon fled their homes to escape flames that have already incinerated more than 230,000 acres. And Washington’s governor says that the state has seen more fire damage in one day than it typically sees in a year.
William Laitinen: So imagine you lost your home, in one of those devastating wildfires I just played in that clip. Who do you turn to, to help rebuild your life? You turn to hopefully the insurance company that you had a policy with. But who is backing up that insurance policy, ensuring that you can get paid? It’s a reinsurer. But there’s an issue, that reinsurers are currently leaving the wildfire market because they can no longer accurately predict the likelihood of wildfires happening because of the intensification of wildfires with climate change. My guest today is the CEO of a brand-new company called Kettle, that is looking to address this very problem. Kettle are working with some highly advanced AI technologies to more accurately predict the likelihood of wildfires happening in given areas. And thus, being able to provide the insurance policies that homeowners need to rebuild their life should fire happen. My guest’s name is Andrew Engler. And Andrew is one of those people who is fundamentally rethinking the way this 600-year-old insurance industry is working. Andrew is, what we can think of as a next generation of talent. He’s what talent can equal when we take the best of technology has to offer and apply that to some fundamental problems that our sector is facing. I hope you enjoy this conversation as much as I did. And without further ado, I give you Andrew Engler. Andrew Engler, welcome to the show.
Andrew Engler: Thanks for having me.
William Laitinen: Absolute pleasure. Thank you very much speaking time speaking to us today. So, Andrew, you are a co-founder of Kettle. Right. So what is Kettle?
Andrew Engler: It’s a good question. And normally the first thing is people say why are we called Kettle and it’s because of a whole different kettle of fish. So Kettle is a reinsurer that uses deep learning to better price and to understand reinsurance risk in climate affected regions and really try and bring protection to people that are in these climate affected areas.
William Laitinen: Okay, so why is reinsurance so important in these climate affected regions?
Andrew Engler: Reinsurance is this incredibly old and storied industry, it’s about 600 years old. It started back in Italy, and really its primary purposes is protecting insurance companies and really protecting people against big climate disasters. So, like wildfires, hurricanes, earthquakes, and it’s actually kind of a common mistake that insurance companies everyone thinks, you know, you get your homeowners insurance from, you’d have it in Allstate, a State Farm, and that person’s covering you if there’s a huge wildfire that destroys your house, and that’s not actually correct. There are these massive companies that sit behind it like Munich Re and Swiss Re, but essentially, when this event destroys your home and destroys the 2000 homes next to you, they’re the ones who end up paying out the claims and footing it. It’s because those primary carriers that they can’t handle the loss of losing 5000 homes at once. And so, it has been a very important obviously, there’s been hurricanes and typhoons and tsunamis since the beginning of time, so it’s been a very critical function. And only recently has it really started to experience some, some incredible difficulties because obviously, the increase in climate change has increased that frequency and severity, the amount of events and how bad they are and how often they happen.
William Laitinen: So, maybe it’s helpful for the listeners to understand a bit about your journey into this, like how, and why you ended up trying to solve this problem around predicting wildfire and then helping to, you know, underwrite those risks.
Andrew Engler: Yeah. I mean, I, it is, like anyone in insurance, it’s a really weird and interesting journey. I, very few people go start out as a kid, I’m sure, and go, “Oh, want to get into actuarial modeling, and then I’ll go to school somewhere and, and get into risk management”. So, after I got out of college, I had a history degree, and I’d always been a closet nerd and kind of a hacker and which basically just meant I broke my dad’s computer a lot as a kid, and had an opportunity to work with an agent at Allstate and help them build a book of business. And it seemed like a great place to earn money at the time. And as I got more and more involved in insurance, I realize it’s this incredible industry, that there’s a litany of problems, I, it’s very undeniable, the fact that most people don’t like paying their insurance bill, but you essentially have this industry that’s protecting you against the worst outcomes that can happen. So, so preventing your worst fears from happening, and if they do happen to you know, making you whole again afterwards. So, if your house burns down, you don’t just lose your life, basically, and have no way to ever recover. There’s someone there to financially back you up and protect you. And so, yeah, started with the agency there and ended up using a little bit more modern data science practices that then were being used at the time and a little bit of software to, to grow that into the second largest book of business in California, and then took over the commercial book of business for the state of Arizona and help them build an online quoting platform and then was lucky enough to spend the last five years as Vice President of Digital for a company called Argo group. And really, with this focus of understanding, there’s such a massive opportunity to bring software and specifically a what we call like a data science practice or machine learning into that the insurance industry because you look at insurance, it is literally the original data business, I mean, it is all based around understanding Core Data concepts and variables about your clients, so you can better price it and understand how to, how to run a profitable company. And yet, when you look at it, compared to its big brother, we call it like, the capital markets and equities, it is very far behind from quantitative aspects. You have many quant funds and modern quantitative technologies that are proliferated through the capital markets. And in terms of insurance, and especially reinsurance, they’re still using, you know, old school stochastic modeling and a lot of these technologies that, that haven’t advanced over time. So, it creates this huge opportunity to really use modern software to one, you know, make a better experience for the customer first and foremost, and transacted easier. But two, the second part that that has really become, you know, life obsession of understanding how we, we predict the probability of these events happening and price them and model them better.
William Laitinen: So why do you think that the industry never modernized in the same way that why didn’t the insurance industry modernize in the same way the capital markets did? What was it about the business that stopped them doing that, do you think?
Andrew Engler: They didn’t have to. And again, you know, the common misconception to most people in insurance is like, “Oh, you know, insurance companies work really well, if they just underwrite, you know, if you can say who’s going to have an accident and who won’t, you’re going to make a profitable company”. Actually, the reason why insurance companies are so profitable is because of what they’re, they’re called a float, and that’s why Warren Buffett is Warren Buffett, Warren Buffett is Warren Buffett because he owns Geico. And Geico makes billions of dollars of premium a year, which he can then invest and, and get return off of. So over time, you know, it has been a kind of core understanding that as long as we’re running an okay, loss ratio, as long as we’re making, you know, 95, or we’re making five cents to every dollar that we put out, then we can invest it and make a lot of money, which kind of creates this, this innovators dilemma where you’re not forced to do massive innovation and really push forward the core concepts of like underwriting in your business, until you have an event like climate change, which fundamentally pushes you into the red and now makes it unprofitable to underwrite which makes it really, really hard to make all your money back on investment, and it really stresses the investment portfolio side of it.
William Laitinen: Mm hmm. Yeah, sort of, in this instance, nature really has been the, the mother of invention here. It’s, it’s, it’s really forced the hand of the industry to figure out how to deal with these natural catastrophes now that are happening much more often. And I can see that in what you guys are having to deal with wildfire because you know, you’d have to basically have lived under a rock not to have known in 2020 that with every other crazy thing that happened there was record wildfires going on in California and in Australia, it was a really terrible year for wildfire. So maybe given this is your sort of wheelhouse now, but many people don’t know about wildfire but, and every time I speak to people, they are “everyone’s talking about wildfire in the insurance industry right now”. So what is so difficult about insuring wildfire?
Andrew Engler: Yeah, so I mean, let’s take it back down to like basics in the way that, that wildfire used to be looked at. And so when you’re using stochastic modeling, essentially, to very much oversimplify it, you’re saying, then let’s use California as an example, I’m going to look at Los Angeles. I’m going to look at, you know, the entire state of California, and I’m going to take a historical data set of 500 years, I’m gonna say how many wildfires happened? Okay, if I want to price Los Angeles, I’m gonna say how many major wildfires hit Los Angeles in the past 500 years? Well, that’s happened twice. So that gives me a pricing to understand that there’s a one in 250 chance that a major wildfire will hit Los Angeles this year. That’s correct, unless you’ve had a complete nonlinear increase in the frequency and severity of those events over time. So now you have this core problem where you used to be able to get actually really accurate results from stochastic modeling and say, “Okay, here’s the frequency, and we know if there’s a one in 10 chance of it happening, we can price it right and everything can work out”. Now, if you have no idea what the frequency of these events happening, if the Camps are or the Tubbs fires are a one in 100 event or one in 10 event, you have no understanding and no core on how you’re supposed to price that risk. Because if you price it as a one in 100 event, and it happens to be a one in 10, you’re going to lose a massive amount of money. And if you price it at a one in 10, maybe no one’s going to buy it because the client, the insurance company is looking at it as a one in 100. So, you have this complete breakdown in the marketplace of communication between price transparency and extreme price dislocation happening. You know, there, there is a very good move by, by Gavin Newsom, the governor there, and what he said was as these wildfires started to increase over time, and the severity of them started to really increase, he realized that there was an un-insurance and under insurance crisis about to happen, which is, if you can’t predict how often these will happen, or you don’t have any idea how to price it anymore, the only thing you’re allowed to do is leave the market. So what he told the primary insurers is, you’re not allowed to leave the state, if you’re in a wildfire zone, you can’t non-renew on people’s policies, because what would happen is just that private insurers would just start saying no, we, we have no idea how to price as homeowners, we have to move away from it. So, what happened is, is it created a moratorium on the ability for the primary insurers to leave. What didn’t happen is there wasn’t a moratorium for all those reinsurers. So, at the end of the day, remember all that catastrophe risk gets offloaded off of a primary carriers’ balance sheet onto a reinsurers balance sheet. And now the reinsurers are still having the same problem of saying this is way too volatile, we don’t know if 2018 is going to happen every year are going to happen every 20 years. So, we no longer think we can price it correctly, we’re just going to price it as high as possible to feel comfortable or we’re going to leave the market. So, what you have is all the supply for all that catastrophe risk start to move out of the market, which creates a huge demand spike in the front for all these insurers who will pay anything to offload that risk off of their balance sheet. And so again, you, you exacerbate this problem of price dislocation even more. And you get to the point of saying, alright, unless we can understand the probability of these events happening in very geo specific locations, unless I can understand like for this home, what is the probability of it burning in the next year, then we’re going to have a serious issue and understanding how to insure it and how to protect it in the future.
William Laitinen: Right. And so, there’s these homeowners, basically, going through these most, you know, harrowing experiences of wars of fire sweeping across their land, maybe killing livestock, incinerating their homes, destroying sort of everything they, you know, everything they had, and maybe killing people. These are really massive events. And then you know, insurance is there to help build back those communities. Right. That’s the point of them. But then if they’re left not being able to get that insurance, it’s devastating. So, you’ve kind of really well highlighted this fact that the government can step in and in stop the insurers leaving, but then there’s this whole network of people behind you need to be there, the reinsurers. So, there’s a problem. And there’s a problem with the way that wildfire was being able to be assessed, because the stochastic models just weren’t doing a good enough job anymore. So where did you come in, Andrew? Is this, were you observing this and thinking like, there’s a problem there, I want to solve that so…?
Andrew Engler: Yeah, I, you know, I’d, I’d say it’s just become like part of a strategy that I constantly use and the way I guess the lens of my mind views the world. So, I, three years ago, we moved out to Bermuda as I was working at Argo at the time. And as I started to dive deeply into the reinsurance industry and look at it, I realized there was these systemic issues that were building up over time. So, this idea of that, you know, we can keep using stochastic modeling, to understand these events. And yeah, they, they trend back to the way they used to be over time, which isn’t the case that we’ve seen. So, in, in my background understanding like the more deeper concepts of deep learning and large datasets and asymmetric information that you’d find in satellite imagery, the weather data, understanding that if you could do a very good job at taking this massive amount of information that is out there, everything from MODIS, LIDAR, Landsat, all the NASA satellites, they’ve been running for 20-30 years, some of them. And this is incredibly good ground truth data and be able to create a fully automated pipeline that translates it because you know, at first, you can’t just take a satellite feed and give it to machine learning model and say, “Alright, predict what’s going to happen”, it can’t read it. So, you have to be able to translate all this information that comes in into a computer readable format. Then if you do that, you, you end up with what we had, which was about a 7-billion-line data set. And we look at this data set, we said, all right, the traditional machine learning models that are out there. So, to get a little bit on the techie side, when you look at the majority of machine learning models that are out there today, they’re all based off of a human mind, really, they’re the majority of them are, and they’re very good for a multitude of tasks. They’re not great when you’re dealing with something that’s steeped in chaos theory. So, where you have billions of variables that might be coming in, and you’re trying to use a statistical probability of an event happening.
William Laitinen: So, what is chaos theory? I thought to interject here for a moment, because it’s quite an important theme to understand, because it affects so many disciplines in insurance, where we’re trying to model very difficult risks. And it has its application in many other sectors as well. But the metaphor of the butterfly is probably what most people will know, when it comes to chaos theory, the idea that a butterfly in Peru can flap its wings, and that will create the storms in the Atlantic months and months later. It’s this idea that in seemingly random events, there are actual underlying patterns and deterministic laws which affect those random seeming events. Chaos Theory. Hope that helped.
Andrew Engler: And so, this is where like, you know, the second part of our journey started after we did the kind of mind-numbing amount of work to translate it into computer readable format, and create that fully automated ETL pipeline. We started in this next venture of saying, alright, we need a complex enough deep learning architecture to really understand the complexity behind these events. And we ended up doing a lot of experimentation and the off the shelf stuff we found, it works okay, it’s better than, than what’s out there right now. But it wasn’t the orders of magnitude better, you want to put a billion dollars limit out on.
William Laitinen: So, tell us about the model that you did end up working with a Kettle.
Andrew Engler: We ended up finding our way into a breakout branch called particle swarm optimization, which is far more utilized in robotics and nuclear particle physics modeling. And Dr. Kennedy, one, one of the researchers who founded it, he is brilliant. What he said was, instead of using the construct and architecture of a human brain, when you build your machine learning model, if you have something that’s steeped in chaos theory, why wouldn’t you use a hive mind structure so less something you’d see like in a beehive or a termite hive, and the point of that being is, if you have an incredibly complex structure, you want to have an incredibly complex output. Beehives and termite hives are incredible. So, there’s 100 million termites in a hive. Each one is kind of dumb and simple, but very good at its little simple job. And the beauty is when you get 100 million of these little actors to communicate with one another and dynamically influence one another’s decisions, you kind of get this groupthink. And all of a sudden in nature, you get these insane outcomes that no single individual would ever find possible. And I’ll, I’ll let everyone nerd out on their own and won’t bore them hear about West African termites in the 30-foot mounds they build with fully automated air conditioning systems. But what we recognize is, instead of having one giant neural network, we should have 32 separate neural networks that work concurrently with one another and dynamically influence and communicate with one another in real time, and when we switch to this newer architecture, we suddenly saw this huge jump in our f1 score and, and we abstracted back and had to understand why is this working so well comparative to other, other algorithms. And it made sense. And it’s because when you deal with something like a wildfire, something that’s steeped in chaos theory in these climate events, you have 50,000 wildfires every single year on the west coast. Maybe one to 15 of them become a Camp, or Kincade or a Woolsey or something of that size. And the reason is, you have to have billions of small little variables lining up in an absolutely perfect storm of everything going wrong for these events to break out. And it’s because you have to have all the wind speeds, the brush color, you have to have the elevation that humidity and, and a litany of other variables all lining up to say this can break out into a massive 5000-acre wildfire. And that’s what you’re really concerned about are, are these big events, we’re not concerned about the one or two random fires started by a lightning storm and, you know, burn for 30 minutes, and then they’re gone. So, so, that’s really where our breakthrough came through. And where we really found that, that there was massive opportunity to start creating a new form of, of reinsurance underwriting.
William Laitinen: Amazing, thanks. Andrew, I’m gonna have to go back and double click on a few of those things. Because there’s, there’s a lot, there’s a lot there, definitely. So, you started talking about stochastic models, a lot of people listening to this are in fintech and insurtech, they may know, but those who’re not, you know, stochastic models are ways of predicting the likelihood of an outcome, right? This is like the way that you can predict something happening. And then you did mention something like an ETL pipeline, I wasn’t quite sure what you meant, what that is, what is an ETL pipeline?
Andrew Engler: I, the, the easiest way to think about it is like as the data comes into the satellite, we extract the data from the, you know, the database, so it all gathers inside of the satellites database, like sitting in a server, and we extract that information out. And that information is kind of useless to machine learning model and we translate it so just think of it as like a pipe you created that translates it into a computer readable format, so creates it in ones and zeros and makes it so the machine can actually understand an image. You have to use a computer vision algorithm again, because like if I tried to show this image and say, “alright predict, you know, which one of these images is, is William and which one’s Andrew”, it doesn’t know that inherently. So, you have to literally break down the image pixel by pixel. And you can say like, “Okay, this exact color of white in the top left corner of my screen is pixel color 010000. And the machine can read that it can read the 01000, it can’t read off-coloured white with a shadow on it. So, creating this pipeline is critical, because you have to be able to translate the data and bring it into a machine-readable format, to actually even begin the process of using machine learning on satellite imagery, on satellite data and weather data maps.
William Laitinen: Yes, that’s it’s definitely a case of right solution and the right time, you got to be there at the right time with both of those things. So, thank you for explaining a bit about satellite stuff. And you mentioned, mentioned about the f1 score, and for maybe those people listening, this is like a measurement of accuracy for an experiment, right? So, this is the f1 score, the closer you are to 1? If I’m correct. Is that right?
Andrew Engler: Yeah, precision over recall. And so, you know, you have to use you can’t just be like super easy with accuracy and machine learning, especially in chaos events, where you go like, Oh, it’s 99% accurate, because technically, machines are smart, the way that they’re dumb, I say. If you have an event, like you know, you’re trying to predict, a great one is always like, “oh, if you have a million people you give loans to can you predict which ones are going to default?” Well, the machine would be correct 99% of the time if it just said no one defaulted, because the likelihood is only like 1% of those people are going to default. So, the way you have to skew your accuracy is you have to heavily reward for a true positive. So, like in wildfire, we heavily reward the system, when it guesses that there was a fire and there actually was a fire, we give it a little bit of a tiny reward, if it says there wasn’t a fire and there actually wasn’t a fire. And we heavily punish it, if it says I don’t think there’s gonna be a fire and there actually was, so you have to skew the way that you measure it.
William Laitinen: So, I just wanted to pause here a moment again, to really focus in on what Andrew is talking about here with punishments and rewards for AI and machine learning. This topic is called Reinforcement Learning. Reinforcement Learning is the training of machine learning models to make a sequence of decisions based on rewards and punishments. The agent, and the agent is what they call the model they’re creating, achieves a goal in an uncertain or, you know, highly complex environment through creating these artificial reward and punishment criteria. And that’s why Andrew rightfully was very, very focused on ensuring that the model that Kettle created was giving reward in the right way to ensure that they got an outcome that was meaningful. This is important because once the hive mind of the AI, the ML model agent that they’ve created, reaches an outcome, it’s impossible currently, for us to go back and understand exactly how the computer made that decision. So, the only way that you can calibrate later on is to then add, again more of these rewards, and punishments. So, I’ll make some show notes and links in the show notes to reinforcement learning for you, I think it’s really interesting area to understand. And I hope you found that useful. Anyway, back to the show.
Andrew Engler: If you look at it, you know, when we say we have an 84.7% f1 score, so that our precision and recall that the best way to put that back into like real numbers. So, when we did our prediction for 2020, we modeled the entire state of California, and we said, “Alright, here’s our prediction of what will happen”. At the end of this year, we looked back and out of the 14 major wildfires, 11 of them we predicted in the top 10% most dangerous zones that the model predicted. And if you move that metric up to like it, so what the model predicted is the 20% most dangerous areas in California, we were 14 for 14 for predicting where those fires were.
William Laitinen: Oh, wow. And that’s, that must have been a game changer moment was that like, late at night event? Like you’re you press the button, you’re waiting for it to spit out its results? Or, you know, was it, as it sort of, you know, cinematic, that’s what I’m asking, or was it just one of those things, you know, “Oh, I got the email. Yeah, that worked. That’s good.”?
Andrew Engler: I wish, it was like, one of those moments. It’s, it’s much more like what you put out the prediction in February, then you go like, “Alright, let’s ride” and we have confidence, so you have a mathematical confidence and, and what your statistical ability is, because you’re training on a historical set, you know, it’s mathematically correct. But everything all, anything can change in the real world when it happens. So, it’s more like watching each day, and each week, and each month and going, “Oh my God”, and you know, you hear about the fire that happened in Lake and we go, you log on it, you know, 6am, and you’re like, holy, holy ass, like, let’s see where it is. And you look here, like, it’s very surreal moment to be honest of looking at and going, “Oh my God, the model was like showing this is one of the top dangerous areas, we were correct at it”. And then there were ones that we missed, you know, we’re 11 for 14, if you look at the 10 percentile, so if you look at the CZN fire, or sorry, CZU, which we looked at, and then we had to go back and say, well, we still predicted it in the 20% most dangerous areas, but it wasn’t like in our top 10% most likely, and we look at it, we say okay, what was wrong, okay, the, the lightning randomization factor in it wasn’t high enough. And so, this was a strange event. Now we have to like kind of retool the model towards itself. It is very much it’s an emotional experience, you’re looking at your baby, and you’re looking at, you perform incredibly well, in real time that other times going like, you always want to be perfect, which will never ever happen. So, it’s a, it’s a fun pursuit in perfection that you’ll never read.
William Laitinen: Yeah, and I think that’s what people I’m trying to get my head around a while as well, I listened to you talk about what this does is effectively your system is using historic data analysis of previous patterns, you know, weather conditions, random events, like kind of visual imagery, a whole range of data sets, right to predict what will happen in the next year, right? And where it’s going to happen next time. And this is the amount of variables that go into that, like you said, these 7 billion lines of data. These are sort of levels of complexity we’re talking about, then you deploy your, your neural networks, your 32 hive minds, to like, these sort of meandering termites. So, they’re trying to figure out all of the potential variables, and then bring those solutions together and predict where this is gonna hit. And that fire is going to happen in that neighborhood next year. That’s what we say. I mean, that is, I mean, what an incredible achievement to do that. Right? And it must be really exciting. But on the flip side, let’s be honest, like we’re thinking about people’s homes burning down, right, or these areas burning down. So, it’s like, I can understand it as the observer and you’re trying to solve these problems with success you like “Yes!” and then like, “Oh, no”, how does that fit? Because it must be this sort of discord between the two situations getting it right, and obviously knowing what’s happening…
Andrew Engler: Yeah, and so like, just to give you the scale, and I like what you’re talking about, like the, on the on the previous part. Every single time we do one iteration of our engine, we do three tredecillion calculations, which is a three with 43 zeros behind it. And essentially, you’re…
William Laitinen: Tredecillion, that’s a great word.
Andrew Engler: …the model is literally taking every single possible iteration. So, it’s saying on August 18, the wind’s blowing 30 miles an hour, it’s on a 13-degree grade hill, the brush is this exact color, it hasn’t rained in, you know, four days, and there’s a one-way dirt road right here, every single time I see this lineup, I’m gonna guess there’s a wildfire going to happen here. And if it’s correct, it stores all of that, and the millions of other variables that line up there and says I’m gonna use this next time to predict it. And then the next time it says I’m gonna do it for August 19, and I’m gonna do it in a 12-degree slope hill. And so, you can imagine how, why…
William Laitinen: Wow
Andrew Engler: …we get to three tredecillion calculations. To your second question there, we know we can make incredible corporate profit off the back end as a reinsurer, it’s a very good like our, you know, market right now in terms of like dislocation and being able to understand what a good risk is and a bad risk. And, you know, the beauty is that we can literally fund all this research with, with the reinsurance profit, and then essentially take all of this research and give it away for free on the front end, because we’re funding it to the back end, and we’re making our money as a reinsurer so we can give it away for free to governments and, and legislators to say stop building in these areas. We give it away for free to fire departments to say, “Hey, here’s better places to create fire breaks, because these are the most dangerous areas of our model”. And, you know, it also hand this information away for free to like energy providers to say, here’s a better shut off strategy, because this area is much more dangerous, you should, you know, know to shut down your power lines in this specific region. And also, at the end, just hand this information for free back to the insurers, so they know how to price the risk. Because what you want to do, you’re not trying to create more price dislocation, so you can make arbitrage and essentially do better when, when everyone’s doing worse. Wildfire is kind of this beautiful thing where like, we’re all in it, no one does better when there’s more wildfires that are worse, you want to mitigate the risk, because as you mitigate the risk over time, so now as we influence the market and say don’t build in these places shut down power here, all of a sudden, the losses start to decrease over time. And as those losses decrease over time, now you have stability in the pricing, because there’s less volatility in what’s happening, which just means you can get more supply out to the market, which means you can create more and more size and get more and more people covered. So, everyone ends up winning in this kind of like perfect cycle of it, if it’s done correctly.
William Laitinen: It’s wonderful that so you’re able to give this information away on the front end, and you said you’re looking for that. So, you’re looking for a business, which can make money and also do something good as well, that is got to be right, a win-win. And, and I’m thinking about that, as well as like these businesses that can achieve financial return, but help to solve a societal issue as well. Because we’ve, you know, the whole reason that we’re experiencing these problems, in my opinion, with climate change, and such, is because people have been too greedy, you know, companies have taken too much, people have not cared about the conditions and the results of our, our habits of consumption and fossil fuel consumption and a whole range of other environmental degradation processes. And, unfortunately, that’s what’s driving this. So, we need to put something back, we need to do something more. Another day maybe I bang that drum. But um, I really come back with you. rewinding the clock, a bit together, you talked about, you know, being out in Bermuda, which sounds lovely, you know, sipping your cocktail on the beach there thinking about what how you to solve the woes of the world. And you kind of start to think about these problems with your other co-founders and sort of the creative problem. How long does it go from thinking of the idea, having the idea, to going through iterations, to getting that, that, that ticker tape come out of the machine giving you your 80 odd percentile f, f rating, whatever it was, to getting that? How long does that process take for you?
Andrew Engler: No, I mean, well like, some iteration of the idea is bumbling around my head for seven years. And like, yeah, as they say, like the overnight success, it takes five years like the I, I truly believe and there’s a multitude of other leaders that say it that like, the patience level we have towards, you know, we constantly see the stories about like startups with the 17-year-old who like, got out of MIT and like, in you know, started up and now it’s worth a billion dollars. Like that is not the vast majority of companies out there like, it is a slog, and it is a slog of things going wrong over and over and over and over again. To actually like this company alone, like when we really started on it, one of the first things that happened, so the original idea was to use this algorithm and, and in pricing to create an evacuation policy. We wanted to create like evacuation policies. I’m from California, one of my co, one of my co-founders lives there now and I was in Bermuda at the time and he got stuck trying to evacuate from the, from the fires, and I almost had to evacuate from a hurricane in Bermuda and the idea was like we have the means to evacuate, and it’s still actually really difficult decision, it’s very expensive, it’s like $3,000 -4,000, and let alone, like, we’re not in a financial dire situation. So, we really needed to try and figure out a way to create simplified evacuation policies where someone could pay $15 a month and know that if they were in a wildfire zone, and there was an evacuation ordered, they would instantly get $3,000 in a bank account and be able to leave. So, we thought this was a brilliant idea, and, and a no-brainer. Dog food is as they say, like we’d buy it. And we ended up applying for Y Combinator and getting like, very, very far into the process all the way to the last interview, which is actually incredibly hard to do, which is when you go all the way out to well, when you used to go all the way out to like Menlo Park and, and sit at a table and you have 15 minutes to pitch these legendary guys who I love and read nonstop. And guys like Paul Graham and Sam Altman and stuff and tell them your idea, and they basically tell you yes or no, if you get in Y Combinator is this very, very prestigious accelerator where some of the biggest names in history have come out of, and the day before, you know, we said, “Look, we need to really like test our product market fit, as they call it. And so, let’s just go down to Target and you know, ask people, if they buy an evacuation policy”, there’s like even smoke on the hills, and we’re like, this should be a no-brainer. And I just remember, Nat and I went down there. And it was one of the worst experiences of our life, like people were like, “Ew, no”, like, “Get away from me”, like, “What are you talking about? Why would I want evacuation?”, and that was our first sign. And we went through Y Combinator and ended up getting denied, and they said, “We just don’t think you’re going to be able to sell this, like, it’s gonna be too hard”. And that was just like a total like kick in the gut, and like, “Oh, my God, this is terrible”, like, this thing that that we’ve had this idea, I’ve been working on iteration, and or I could tell you 1000 other iterations of it before. And then, you know, we kind of came back to it. But this is like the game of startups. And we went, well, we know we have something good here, we know we have in the algorithm behind being able to price those evacuation policies was the first generation of what we have today, because it was very good at predicting where these wildfires and evacuations were most likely to happen in the next year, so we can price it correctly. And, and so we kind of stepped back and said, Well, we know we have something good. Like, why are we going after like the front end, but let’s go after the big idea. Like let’s just build a reinsurer from ground up which is kind of an insane idea takes a lot of capital, it takes a lot of expertise and a ton of, of data science horsepower, we’re like, Look, why don’t we just take a shot at this one. And that’s where the Kettle itself like really took off. And God bless it, you know, we’ve, we’ve worked like crazy over the past year and built it up. And, and we’re lucky enough to have something pretty fundamental. And that, that is fundamentally powerful, and recognized and raise a seed round off it from wonderful investors. And now we’re well underway of growing the company.
William Laitinen: I’ve got to go back to something you said, which I’m fascinated by. So, you got all the way through YC. As the cool chaps call it right, I hear. And you got all the way to the end and they said, “okay no”, but you want to go and test your idea, so you went with someone to the local supermarket or store and what stood outside and asked people if they’d buy it? Like literally you stood there with a clipboard and said, “Hi, do you want to buy some evacuation insurance?”.
Andrew Engler: Oh no, we went inside and hunted aisles and like even asked employees and stuff and that it was, I, one of my favorite things like I tell the team is like one of the easiest ways to do well in business is just go do what you’re uncomfortable doing like, so if you know you’re uncomfortable going to a store and asking someone to buy your product and you know, there’s something wrong, or like there’s something out there or if you’re uncomfortable having a conversation with an employee about like some issue, then that’s literally the most important conversation you need to have. Like, it’s just natural, we kind of always gravitate towards a place of comfort, and that place of comfort can lie to us for quite a while. So, it’s normally pretty easy to test out assumptions like you, believe me, it’s not fun. It never feels nice, but like it’s pretty easy to get this stuff, test it out and see if you have something or not.
William Laitinen: Yeah, that’s really great. So I can just imagine you in the, in the aisle for like cereal, tapping on someone’s shoulder…
Andrew Engler: Please, by the way, and like, I’m, look I spent the beginning of my career building an insurance agency, so I’ve sold insurance, like over the phone and face to face for the beginning part of my life, so it’s not something uncommon, and that was still one of the most gruesome tasks that you can imagine.
William Laitinen: Hmm, that’s really important, that’s a really interesting point. And I actually think doing some type of sales or tele-sales or some type of role like that door-to-door work early on. in your career very early on, it’s such a great skill to develop, because nobody, you know, till you sold something like face to face to somebody, you really, you know, you really are missing out on a really magical moment of human interaction and learning a lot really, really quickly. I suppose now I just want to sort of talk a little bit about like going on to get where you are now, because you talked about your seed investors. So where are you guys out at the moment with Kettle? Because, you know, I’ve, you know, I’m fascinated by what you guys do, and you got some really impressive people behind you. So, you’re now at seed stage want to tell people and what you’re doing and sort of what’s, what’s happening next for Kettle?
Andrew Engler: Yeah, we were very lucky to have quite an overwhelm of interest in what we’re doing. And we initially went out to Silicon Valley and the VCs thinking, like, “God, no one’s ever gonna want to invest in a reinsurer and balance sheet risk” and the point we made to him is like, look, reinsurance is essentially the single greatest safety net we have left against climate change, either that or like FEMA, government programs, and that you do not want to overload FEMA and government programs with the job of making people whole again, and, and protecting them after a huge climate change event happens. It is not, it’ll create a humanitarian crisis. So, it’s very critical for reinsurance to work, and so we raised our seed fund and had incredible investors come in like, like True Ventures, and Homebrew, Accrue, Anthemis, and it inspired. And now we’re currently growing like crazy, hiring it feels like, every day and like the main part of our business is really deploying, so we set up our operations in, in Bermuda and, and we have obviously, it’s kind of a new world where we’re all distributed. So, part of us are in California, part of us are in New York and part of us are in Bermuda and London. So, spread out quite a bit. But it’s really iterating on the product, so continually increasing the accuracy and data that’s coming in and abilities of our, of our architecture. And then the other portions of it are raising risk capital. So, while we have our own small internal balance sheet, we leverage that with traditional capital providers. And we’ve been blessed in that sense to that there is definitely a lot of interest in the industry from pension funds, hedge funds, and, and reinsurers to really, you know, work with us to deploy capital to make sure supply comes back into the market. And a lot of demand from the front end too obviously, of people wanting to offload wildfire risks. So, we’ve just kind of seen this incredible growth that knock-on-wood, you know, it’s funny in, in our side, and maybe I’ve just been in reinsurance too long, I’m always like, what, what’s gonna go wrong, like something has to go wrong? Because we’re so used to that. And maybe that’s just my own screwed up brain in it. But yeah, we’re in very blessed point right now and running up, and we’re all too to anyone at this point in time, like, it seems it’s very hard to see like in periods like this, and we’re dealing with COVID, and so many things going wrong. And it seems like these are like the down periods in history, and they’re actually not, in an entrepreneurial setting. Just Google online and look at like, top 20 companies created during a recession, you’ll see Airbnb, Uber, all these companies started in 2008. It’s because when these massive catastrophic events happen, everything starts to fundamentally change everything you have paradigm shifts happen. And in that point, you know, Uber started because like, people were losing their jobs from the financial crisis, and going I need to make more money somewhere else. So these actually present themselves as the greatest opportunities for innovation, as opposed to, you know, these moments when everything’s going down, all the good times are over, and now it’s going to be terrible from now on. And that’s, I think, maybe it’s because we’ve worked in reinsurance and insurance so long. So, we’re used to seeing how often these events happen. They’re actually much more common, these black swan events, as they’re called. They’re, they’re much more common than the human brain, lets us believe. So, they just present massive, massive amounts of opportunity for innovation.
William Laitinen: That’s really fascinating. And I just want to actually go back cause I think this idea of innovation and what you’re doing and why it’s so exciting, and something I’ve touched on with previous guests is and Andrew kind of try and try and summarize this if I can best, that what you guys are doing with Kettle is being able to understand, predict the likelihood of wildfires occurring, so you’re more able, accurately able, to price that and now you can take on primary insurance, underwrite primary insurance, with a much, much more granular understanding the likelihood of someone needing to claim and reason investors are so excited by that is because you can create a pool of risks, which you’ve really priced accurately, you’ve got a really nice arbitrage because nobody else is able to price it like you against the market, against the primary insurers. And so that can allow an inflow of capital, which at the moment there’s terrible interest rates, right? There’s like zero, governments are printing money, and so there is a desire to find alternative asset classes. And so, there’s these trillions of dollars floating around pension funds, or everywhere else in a capital markets trying to invest in these things, and you’ll create effectively start to create a pool of risks, which people can invest in, did I do a good enough job?
Andrew Engler: You did better than me even. That was phenomenal.
William Laitinen: Thank you very much.
Andrew Engler: I might steal that.
William Laitinen: I want to pivot here for a moment, like in sort of, some of the last parts of this because you built a team. So, what is it like building a team? And is that been harder than you expected? You know, what if, what has gone into building the team that you’ve got what you’re learning about that? What would you tell other people your experience?
Andrew Engler: Yeah, I mean, I’ve, I’m familiar, I mean, I’ve built teams up to like 40 people in past like roles and stuff like that it’s fundamentally different in COVID times, when you can’t even meet an employee and face our potential employee face to face. I think we’ll see a shift into like, kind of like, “oh, would you be able to consult for a bit and then come like work” because it’s this great like trial period where someone can make money and, and, you know, do multiple things at once. And then, you know, jump on full time, but it has been very, very different from past like hiring experiences, where we have an influx of people wanting to join, because it is such like a core mission driven company and hit so many marks. And you know, we’re lucky to have like excitement around it and have articles in Forbes, and like Fortune and stuff. It definitely, if you have a mission and vision behind it, and people realize like, this is something that I would love to wake up every single day and just do, the hiring part becomes a lot easier, I would say. I take a lot of influence from like, if anyone wants a great recommendation, there’s something called like the, the Valve Employee Handbook, that was put out, there’s a company named Valve, and they own Half Life and Steam, and all these incredible software developments.
William Laitinen: It’s a video game, a video game maker. Valve is a video game platform.
Andrew Engler: Yeah, a video game maker and they actually, own like a video game store now, which is incredibly profitable. But Valve has this beautiful book that they created for all their employees, and it’s really the way we look at it, cause it’s very horizontal structure and the way that they’re structured, and it’s created around, like, giving people massive amounts of empowerment and telling him like, figure out what’s wrong with the company, like jump to the point of biggest need, and jump in there and help out, so like, find your place in the company, and we’re here to support you like find it. But like, it’s much different than like that, we’re hiring you for X role, you’re only going to do X come in and like, do that position and stop like talking to this person stop worrying about these problems. It’s very much more like you have people hunting down the biggest issues and fixing them.
William Laitinen: How have you found to do one of the perennial issues, you’re recruiting in an area, in reinsurance where people are very comfortable, it’s a well-paid industry, right? And actuaries, I don’t if you’re recruiting for actuaries, and people like that, they’re a risk averse bunch. Have you found that to be still quite difficult as a hiring talent, or has that not been an issue for you?
Andrew Engler: There hasn’t been an issue for us, actually, I mean, look, we like to call actuaries are the original data scientists, like there’s an absolute love of what we’ve been doing especially in these, like incumbent companies, they have been nothing but incredibly supportive and great conversations. Because, you know, we’re really just trying to bridge this gap between like the modern data science practices that, that have proliferated and in capital markets and quant methodology, and, and bring it to the industry that, that is the original, like data science business. And it’s really like kind of a beautiful marriage between the two. It always, it’s always weird to me, when I see like startups saying, like, oh, we’re gonna come in and take over the insurance industry and like, kick out the incumbents, like, no, you’re not like, these are incredible companies that have built billion, massive billion-dollar balance sheets, and have been protecting the world for decades. Like they know exactly what they’re doing. And there’s just a match to be made in terms of like this kind of future iteration that happens. So, the companies that’ll succeed are very much in partnership with them, and in no way like, oh, we’re gonna compete and eat their lunch and, and you know, they’ll die out.
William Laitinen: Cool. That’s interesting, I’m glad to hear you having success in those areas. So, you’re so I heard from that when the Valve stuff, you’re taking a model where you’re going to try and put values at the center of the business as well. And create a flat hierarchy where people, smart, brainy people can come and solve big problems by searching out those problems having some agency right? To do things. I heard that right, cool. So, I’m mindful of time and I’m really thankful for the time you’ve given me today as we sort of wrap up the show today. I always like to ask about influences and specifically like we’re in COVID right now, I’m in second lockdown here in England. It’s a pretty stressful time. And I’m sure your, your, maybe your employees are experiencing the same thing. And I noticed from your, your stuff you’re doing your, your spare time, you do some charity work. So, I was wondering, maybe tell us a bit about how you stay well, and like how you help others to stay well? What are you doing, Andrew?
Andrew Engler: Yeah, I think anytime that you’re like kind of focusing outwards and a great mutual like relationship you can have. So, a lot of times like people who are mentoring, one I love mentoring, the entrepreneurship like community, I promise you is one of the most friendly and welcoming like we had so much help getting to where we’re at, there’s no way we would have ever got the funding we had without the incredible support of like a million people I wish I could name. And I am always willing to help out any person, I just had someone randomly reach out to me on LinkedIn three days ago saying they have some climate idea, and I always want to help out in that. So, it’s important to, like, pay it forward in those circles. And one of the other great things you can do too. And when I mentoring, they say, “Well, how do you get like higher in your career, and how do you get to like a VP level?”, one of the greatest things you can do is like go work for nonprofits. So, in your free time, if you have, you know, a bit of digital skills, and you can learn all this stuff online, I, I’m a case study for, of just taking millions of online courses and reading every blog available. And what you can do is start to use like some of these tools that you’re learning in true practice for, for nonprofits, because nonprofits are always looking for people, especially younger people who understand digital technology to sit on their boards and to, and to, you know, help out in terms of like their social media practice, or, or their website development, whatever it may be. And it’s great, because you actually can start sitting in a leadership position. Now, you’re not going to get paid for it, but that’s the beauty of it. Like you’re going to sit in a leadership position, learn more about how to be a leader and be able to give back, incredibly to like organizations and causes that need your help and your support more than anyone in the world. And I’ve been lucky enough to work with great organizations like news, Neurological Musical Association in Arizona and Gabriel’s Angels and places like that, that I got to work with people because generally the other people on the board will be CEOs of giant like companies and people who just want to give back. So great environment to develop.
William Laitinen: And is that what helps you stay healthy, in some ways? Is it that idea of…
Andrew Engler: Yeah, yeah, I do. I mean, that’s the only way to stay sane, I would say is, is constantly like trying to focus outwards. If you focus inwards, then yeah, there’s a lot to worry about and a lot to, especially on what’s going on in the world. And instead, if you’re just focused on like solving small little problems that are around you constantly, then yeah, that there’s a lot more sanity in that I believe, than trying to just read things that you can’t fix.
William Laitinen: I’ve been, I’ve been thinking a lot at the moment on compassion and kindness on for ourselves and for others. And it’s such an abundant state. And by what you know, I think we’re all so fortunate, we have so much abundance in our lives, people like yourself and myself, and I know how lucky I am to have what I have. And by going to help others, you know, you really get a sense of what you can give to others. And through that giving, and it really is like a great, like, restorative force, right? You feel good about yourself; you feel others feel good about themselves. So, and so for those listening, you know, look, you know, you can just go find a great charity to work with. And I can testament to that my work with Rainforest Trust, and it’s just great knowing that you’re helping other people. Well, final thing, I always ask all of my guests, so your favourite book, Andrew, what is your favourite book that you recommend the most or you give out the most?
Andrew Engler: That’s a hard one, I would say the number one that I always recommend, it’s actually a little bit hard to find, it’s only in hardback. So well actually there might be some digital copies now, but it’s called “Poor Charlie’s Almanack”. It’s by Charles Munger is the number two at, at Berkshire Hathaway, who is probably the most brilliant human being on earth. And it’s all about how you use mental models to kind of… there we go. Yes! Oh my god! So, I’m sure you will test, that is one of the most brilliant books ever written that will change your life fundamentally.
William Laitinen: Absolutely fantastic. There’s a hint to this one. I love this book is fantastic. But I bought five copies, so I can give them away.
Andrew Engler: Yeah, I know. Because it’s so hard to get, you have to find it for 100 bucks on Amazon when I try and give them out!
William Laitinen: The way, the way you do is you buy it from the publisher. You can only buy directly from the publisher in the US because of the way I think they did it. But it’s you can find it. Maybe I’ll post a link in the show notes to where you can go and order it direct because Charlie gives away all the all of the proceeds, so it’s um, yeah, a wonderful recommendation, “Poor Charlie’s Almanack”. It’s just such a fun book. It’s massive.
Andrew Engler: It is, it’s beautifully written. It’s yeah, it’s great for a coffee table even but you’ll tear that thing apart over and over again and go “oh my God, how is one human being so smart?”
William Laitinen: Yeah, absolutely. Well, that’s a wonderful way to end the show Andrew, I thank you so much for your time. Thanks for sharing your journey with Kettle. I really genuinely wish you the best of success.
Andrew Engler: Thank you so much. Yeah, absolute pleasure. William, it was really an honor.
William Laitinen: Well, thank you, ladies and gentlemen, for joining us on another show of Talent Equals. A big thank you to our sponsor Exige International. Without the support of Exige we wouldn’t be able to make this show. An even bigger thank you to our production team, Andrea Muraskin and Samantha Smart because without their help and support, the show would not be what it is. Please do join us in the next episode, we’re going to be interviewing the incredible Lisa Lahey, who is and has been one of the foremost thinkers in and around adult development and has been critical in the development of what I think is one of the best change models, the Immunity to Change I hope you will join us on that episode. And until then, have a wonderful day or evening wherever you are.
We won’t share your details with third parties or send you anything we wouldn’t want to read ourselves. Promise.