Ryan Kottenstette founded Cape Analytics in 2014 to help companies identify building characteristics based on aerial imagery.
Five years on and Cape has analysed 80% of the US population, in terms of physical property risk, with the resulting data being used to help price and manage insurance policies and reinsurance portfolios by over 30 companies. Today the company ranks as one of the best known technology companies in insurance to have emerged in the last few years.
Ryan and Matthew's discussion points include:
• the challenges and successes he has seen when building a company
• using aircraft, satellite and drone data for more accurate pricing
• use of third-party data
• dealing with uncertainty
• the growth in 'insurtech' and insurance innovation
• understanding wildfire risks through artificial intelligence
• approving new class plans and rate filings
• the importance of focussing on customer needs
If you want to hear more from Cape Analytics, they will be in London at the end of January and we are hosting a breakfast event for them. Details in the Instech London newsletter.
Transcript for this podcast
00:00 Ryan Kottenstette: We have been successful with our customers, getting new class plans and rate filings approved. And to date, we've been approved in multiple states to actually price insurance policies with our data.
00:12 Matthew Grant: Hello, Matthew Grant here. Great to have you back, or if this is your first time listening, I'm delighted to have you join us. Well, with winter approaching us in London and the nights drawing in, we're importing a little bit of sunshine by talking to some old friends from California. This week, it's Ryan Kottenstette, founder of Cape Analytics and described by one well-known industry figure as their favorite CEO. Now, Cape was founded in 2014 to help companies identify building characteristics based on aerial imagery, which is then analyzed by Cape. Until now, they've been focusing on North America, providing their data to insurers and reinsurers, but they successfully crossed that dreaded Jeffrey Moore chasm on the technology adoption curve and now have a number of well-known insurers and reinsurers as clients. We're also delighted to welcome Cape Analytics as a member of InsTech London and we'll be hosting a breakfast event for them in January, when they're coming over here. So if you are our insurer, a reinsurer, or a broker, and you're interested in attending, do let us know. Contact details will be in the episode notes.
01:30 MG: So Ryan, great to have you joining me today. I think you're calling in from California, your head office?
01:39 RK: Yeah, that's right, Matthew. Thanks for having me. I'm based here in Silicon Valley, in our headquarters office in Mountain View.
01:46 MG: Good. Well, as we roll into winter here it seems to be a very sensible place to base yourself. And actually we first met in California last year at the PIR conference.
01:55 RK: That's right. That's been a conference we really enjoy and recently, we've become a sponsor, and so I'll be there again this year in just a couple of weeks.
02:03 MG: Great. I think I described to somebody recently, as the thinking persons InsureTech Connect.
02:10 RK: It's a smart group, it's a much more intimate setting. We've been there, I think, probably going on four or five years now, and it's one of the conferences we really enjoy.
02:18 MG: So Ryan, you started off your life as a mechanical engineer as well. It's really encouraging how many people, given we both did this, seem to have ended up in really interesting roles having started off as engineering. Then you went on to BMW and now you ended up in insurance. And I guess most people would say that would be a fantastic place to not just start your career but end your career working for BMW, but just maybe interested just to hear a little bit about the journey that took you from BMW to Cape Analytics.
02:46 RK: It's not the most linear path you'd imagine at first. I've had a few stops along the way now. I've been involved in early stage kinda technology companies, been involved in venture capital investing. I knew what we're doing as Cape is sort of coming full circle in so far as... You know, there is a real... We have a bunch of great talented engineers, lots of PhDs in our company, but there's an embedded desire to really solve meaningful problems with gravitas and they deal with the complexities of the real world. And I think in that sense at least, in the automotive industry, sports cars or racing, you deal with a lot of those kind of solving real hard technical challenges, but that have kind of real impact and in that sense, I think I see a lot of commonality.
03:33 MG: And insurance, in a sense, having a role to help people understand the risk and therefore be able to reduce risk is something people don't always think about immediately, but it has, obviously, has a really big impact on that. And we'll talk a bit about the wildfires in a minute and some of the things you've been doing there. But probably, first of all, just useful to get a description of what Cape does.
03:54 RK: A couple of different things. First of all, there was more and more access to geospatial imagery. So satellites, drones, aircraft were capturing more and more imagery. And about 2012, there were some major breakthroughs in deep learning and artificial intelligence specifically around extracting, using computers to automatically extract information from imagery. And so we basically took these two tail winds and said, "Where can I apply massive access to geospatial imagery plus deep learning analysis on that imagery, where I provide a real answer and a hard ROI to a real customer need?" We quickly focused exclusively on insurance and reinsurance.
04:40 MG: A lot of people claim to be able to use satellite imagery, and drones, and aircraft to capture data and it seems that the cost of acquiring that is going down. How do you distinguish yourself in terms of what you're doing versus all the other organizations out there?
04:58 RK: When we started Cape, InsureTech was not in the popular vernacular. We started in 2014, I think. We've run 80% of the US population in terms of physical property risk, and consequently, since we're able to cover 80% of the US population and go back multiple years in time, we're able to run loss impact studies that form the basis of actually demonstrating the monetary value associated with the risk answers that we're able to deliver.
05:27 MG: You have primarily been looking at aerial imagery, and how much can you tell from aerial versus having to do a kinda ground survey or using... Do you use third-party data to compliment what you're learning from the aerial imagery or can you just do it from the air?
05:44 RK: For us, it's really a question of what are the unique things we can do that are differentiated and high impact. And so we started by exclusively extracting data from imagery, so there's a handful of folks out there that, large incumbents and otherwise, that have a veneer of geo-spatially derived data, when in practice what's happening is they're delivering geospatial imagery and they're simply appending data source from tax assessor data or other structured data sources. And that really misses the whole value impact because there's a ton of value when you can capture a more recent image and understand what's changed and what's new about a given property, and so that's really been where we've focused. Now, we do append more and more data or look at different alternatives, but the critical distinction is, I think, a handful of folks out there, even large entities, have really been doing for a while what I consider a bit of a bait and switch.
06:41 RK: Showing a geospatial image but then appending data from a tax assessor file which may not be up to date or as current as the imagery, and often is not. So I think that's a pretty important distinction for the in industry to understand.
06:54 MG: Are you tapping into other third party sources of data and beyond just tax information but...
07:01 RK: We're increasingly doing that. You should expect to see more and more of that from us. We've announced recently a partnership Hazard Hub, so we're delivering hazard data. For some time we've been delivering other third party data sets. And really the focus for us is not to append the additional data, but it's for us to apply our unique data set plus our unique machine learning and science expertise to really improve the overall quality and accuracy of some of those other data sets and apply basically the same thinking, rigor, and processes that we apply to deep learning derived imagery extraction so that we get a more and more complete data set that we deliver to our end customers.
07:42 MG: This is really interesting. So you sort of cut your teeth with the satellite imagery, but the lessons learned from that means that you can add value to the third party data and then presumably link it up with what you're getting from their imagery.
07:56 RK: I'll give you a really good example of that, with every data point we deliver, not only do we deliver a specific score, let's say a roof condition rating, but we deliver a confidence score or basically an actuarily sound probability of that class label being correct. And what that means is now an actuary can take all of our scores for all of our locations and intersect that with the probability of that score and get an accurate weighted average and you see that across an entire book and it holds. So it's what I call an actuarily sound approach to confidence scores. For a long time this industry frankly was... Data was delivered with no confidence associated and increasingly as confidence scores have started to be delivered by some vendors, it's really a heuristic, like highly confident versus not very confident as opposed to some real probabilistically rigorously modeled confidence score.
08:53 MG: Makes complete sense. That's from the catastrophe modeling days. That's what people were looking for. Everyone recognized that it's not perfection and they need to get the visibility of the uncertainty 'cause they can work with that. But presumably still need to validate the data in some way. I guess the longer you're in business, the more data you've got compared to real life experience from claims or surveying. But how do you... If you're going in to a client and they want to validate your data sets, other than going out and surveying individual properties, how do they get confidence?
09:26 RK: We have the benefit of having started doing what we do before anybody else and I think the highest quality team, the largest customer base, and some of the most rigorous solutions in the industry. And consequently, we've built very strong partnerships with our customers: We consider them clients, we treat them as clients, we have very collaborative interactions. We've had the privilege of customers sharing significant parts of their loss history information with us which we've then been able to back test and validate our variables and demonstrate the predictive power. So that's been extremely valuable for us. And again, we've done this on a national basis across the Continental US with what some of the largest carriers who are our customers. That's been a big thing for us. So when we engage with a new carrier, we're able to show him quantifiable rigorous loss history analysis, some of which we've published that's basically industry-wide. And we're also able to run loss impact study directly on their particular book of business.
10:27 MG: Yeah. No, that's a huge benefit and I think it, again, it just plays to the fact that you've got credibility and therefore people are willing to share data with you, and I think that's hard for someone... A company to do if they don't have the confidence and insurance companies are understandably getting a lot more careful about who they share the data with. But in terms of how people are using it, is it more on the risk selection and obviously, in the US you're slightly limited about how people can price just given that it is a regulated market, or is it more on the portfolio side that people are using this information?
10:58 RK: It's really important to distinguish between primary carriers and reinsurers. We bring more accurate information that is proven to be relevant to assessing risks. And that, at the end of the day, is fundamentally what this industry is about, is assessing a given risk and coming up with the correct price for it. Or in the event you don't have pricing flexibility, as you mentioned in some regulated markets in the US, you make an underwriting decision. Do you want to take that risk or not? If you can't price it appropriately, maybe you don't take it.
11:29 RK: The specific use case and the way that manifests I think has slightly different flavours for carriers versus reinsurers and depending on your particular business strategy within a carrier you might use it differently. So I'll just give you a few examples. An obvious thing for a carrier to do, typical US carrier renews about 85% of their portfolio each year and then replaces about 15% of it. The first place that they can look at us is on the 85% of the portfolio they're likely to renew, they can use us to understand what's changed and identify the risks they want to pay the most attention to and determine whether or not they want to renew. So of the 85% that are going to renew, how do they figure out which 85% they want to renew and they can actively look to ameliorate some of the riskiest business on their books, and do this at a scale and cost that they've never seen before. The second big bucket is on the new quote, so as you look at new business coming in the door, you can apply... Some of our data is available instantaneously, actually all of our data is available instantaneously, and carriers, for some of that they typically have to wait until they've bound a policy to inspect a property and understand some of that information. So by us delivering the same or similar information instantly up front at time of quote, it allows carriers to better price, better underwrite, fewer post-binding adjustments, easier process workflow...
13:00 RK: And so that's pretty important. But then there's additional use cases. One of the ones I'm most excited about is, because we have the rigor that we described and the proven impact that we've described, we have been successful with our customers getting new class plans and rate filings approved. And to date, we've been approved in multiple states to actually price insurance policies with our data. Conversely, we deliver our data directly to reinsurers, who then use this to inform uniquely their perspective on a cedent portfolio.
13:32 MG: So that rate filing one, you mentioned that in passing, but that's really impressive. Those are often really rigorous and that as an endorsement for what you're doing, and then its ability to actually use the data in really meaningful ways is actually incredible. And I think most states actually put that in the public domain. So presumably, anybody that wants to look and see what you've done or wants to inspect it can also go and review their assessment of you for the rate filing.
14:00 RK: For us, I might have buried that in the middle there, because for us, this is one of many things on a continuum of the things a carrier can do when they have better, more accurate data upfront. But from an industry standpoint, it's absolutely a game changer.
14:14 MG: And also, you've done a great job of being able to get your... Certainly some of your clients to be comfortable about you publicizing their name. I see you've got Hartford Kin, Cincinnati on the website, and I think you've got, is it close to 30 companies now who are actually working with you as full clients?
14:30 RK: Yeah, we've got about 30 customers to date. It's not a small thing for a carrier to support, endorse usage of their logo or give you a quote, but we've been very privileged to have a handful do so, right? I think we have announcements including Security First, State Auto, The Hartford, Cincinnati Financial, CSAA, which is a part of Triple A, Nephila Capital, which is now part of Markel, XL Catlin, which now part of AXA XL, have all been announced. And again, as you mentioned, there's many more customers as well.
15:05 MG: Now, well congratulations, and also congratulations on getting their legal teams to approve that. It's a frustration for many that the insurance industry wants to support innovation, but yet the legal team don't want to talk about who they're working with, so no one can really figure out who's really making a success versus who's just making a lot of noise, but not [chuckle] necessarily doing the real work. When you're looking at reinsurers and you've got a very large portfolio, is it hard for them to see the value on location by location risk that you're offering?
15:35 RK: We have gotten asked. Hey, at the end of the day, what we do is correct, more important, more accurate information on a location-specific basis, then by the time you've aggregated that into a portfolio, location-specific errors just sort of all come out in the wash, so to speak, when you aggregate into a portfolio. And what we've found time and again is that's actually not the case. The errors in cedents' books aren't random errors that kind of net out neutral. They're structural errors because different carriers kinda have different human policies, different ways of writing their business, and you can actually see that some of those things move in a structural direction. They might overweight hip roofs or underweight hip roofs based on the way that they underwrite risk or roof condition, for example. And so what we've found is that the distinctions we see do hold on a portfolio-wide basis and can move cedent portfolio risk substantially. So we have reinsurers that are fully live in production actually pricing cedent risk with our appended data and we're able to put that directly into the format that their current workflow consumes as they basically leverage our information.
16:49 MG: And presumably the pricing model is slightly different if you've got somebody using your data for individual property look ups versus a reinsurer with a large portfolio.
16:58 RK: We're able to offer a kind of pricing that's appropriate for the use case. Some of our customers may have a 100X higher quote volume than another because of the nature of the use case, be it reinsurance, or let's say, a marketing use case even, and those have absurdly large quote volumes per the value of a specific address determination that they're making.
17:21 MG: Yeah, we could have a whole discussion around how to price for value and fairness, but [chuckle] we've only got half an hour for this topic. So that was, yeah, that's very helpful. Talking about wildfire now because you're doing quite a lot of work in that, clearly it's very topical. When we spoke before, you mentioned that one of the things you look at is the extent to which there is vegetation around a house or not as one of the measures of wildfire risk. But presumably, you've... Unfortunately, just given what's happened, but you've learned a lot in the last few years, just interested to hear a little bit about where you're taking things next?
17:55 RK: Living out here in California, we are absolutely ground zero for wildfire risk at the moment, unfortunately. Not only is this a huge issue for our industry, it's a huge humanitarian issue out here. We've been having kinda rolling power outages as a preventive measure to kinda cut down on sparking new wildfires, so it's a significant issue. The total damage associated with wildfire has just ballooned the last several years, I think everybody in our industry probably understands that. The second thing people have come to understand is the prior ways that we would use to analyze risk in this space are really just outdated and not effective. So we basically looked at this problem and said, "Well, how can we help solve that?" And there's very clear wildfire standards published by CAL FIRE that really speak about this notion of defensible space. So what have you done to remove fuel loads from your property, in particular, and create a defensible space radius to protect your structure? So these are not kinda new concepts that we have to invent or evangelize. What we've done is basically come up with an algorithmic way to make that information available to our customers for physical structures in their portfolio or in cedent books they might be considering and do that instantaneously at scale.
19:14 RK: So artificial intelligence algorithms will zoom in on a specific property, they'll identify the structures, and they'll identify all the trees, shrubs, and other fuel loads, and then they will measure the distances from the structures to any adjacent fuel loads, which can be trees, can also be other structures, for example. And then give a quantifiable score, which basically says, this is... And it's not just a one to 10 score, it's much more detailed, much more rigorous than that. So, it's really what percentage of fuel load is in increasing concentric ring around the structure, and that then ends up being a way to disambiguate between multiple structures who may be in the same kind of macro-wildfire risk band, but have very different property-specific defensibility measures in place.
20:02 MG: And what about just generally instead of vegetation in the areas where the fire may initiate, so we were talking about the areas in the proximity of the house, but are you also looking at the level of dead trees, which I think is increasingly a problem in California and that is part of the reason wildfires have got so significant. But can you look at it on a sort of bigger scale so you can see where the wildfires might start, how they might spread, as well as just individual property level?
20:30 RK: We have a fundamental tension in our organization between trying to remain laser-focused on kind of the next relevant commercial thing for our customers and all the great stuff we can do to remain a bleeding-edge technology company. We see the major part of the gap has been on the property specific. How do you distinguish between specific individual risks that might share a wildfire zone, but if you stay tuned for the future, you might see some more broad-brush complex things, things coming out of our R&D efforts.
21:06 MG: There's a tension for every successful CEO, which is on one hand, you see all the exciting, interesting things you can be doing, but on the other hand, you got to stay focused and deliver where people are going to spend the money, and getting the balance right, I guess, is the difference between success and failure.
21:23 RK: This goes back to a core philosophical point for us, is really treating customers as our partners. And by establishing and continuing to invest in those collaborative relationships, it helps us ensure that the incremental effort we do spend on an item is actually not only theoretically interesting, but impactful in the real world, and that continues to be our guiding light in terms of how we prioritize our work.
21:48 MG: And on the topic of new things, you have announced recently change detection and property chroniclers to new initiatives, can you just say a few words about those?
21:58 RK: So here the basic concept is, I not only want to know what's happening for a given property at a given point in time, but I'd like to understand the impact, what has happened on that property or in that property associated with that property over a period of time, right? And this can be what's happened in that property pre-dating a carrier's interest in insuring that property or it can be the median policy, homeowner's policy in the US is about eight and a half years. You're looking at a significant amount of time over which that structure can change, things around that structure can change, certain environmental things can change, conditions can change. And so, we do a good job looking, not only persistently at a given location to understand how it's evolving, but we're looking at the surrounding areas and understanding how those things are evolving, and getting a view on how that might change risk. And it can be for the better or for the worse.
22:56 RK: If you get tree growth, this could end up being a wind missile risk, this could end up being a wildfire risk, and a carrier might want to proactively engage with an insured and talk to them about risk mitigation techniques they can take. On the flip side, let's say a property incurs damage and replaces the roof. We know roofs lead to a huge fraction of all losses, and so that is now arguably a much safer property, and both new carriers as well as current carriers might want to be aware of that. The current carrier might want to proactively take steps to retain that client, whereas new carriers might have more confidence kinda quoting that business.
23:35 MG: I just wanted to ask a broader question really, as you look out there at the landscape of other technology companies that have started over the last five or six years, it seems that you're one of the few that have actually been able to develop technology and actually engaged directly with the insurance companies, and also, particularly those that are not MGAs or are actually doing some of the transaction themselves. But how do you see things evolving with people that are building technology? Are we seeing a shift to more companies that are going to acquire partners or platforms to get access to the insurance companies or do you still see as a strong theme of organizations like yourself where you can stand alone and develop relationships without needing to sort of work closely or on the back of some of the larger organizations or platforms out there.
24:25 RK: Most carriers are of a significant scale, reinsurers as well, and with that scale, and for good reason, I think relatively conservative businesses, they get penalized for being wrong on things. The bar is very, very high, so I can't say it's been an easy road for us these past five years. It's been really challenging. The key has been we have brought and continue to bring a highly differentiated, unique, high-impact solution. What I've seen not be successful are solutions that are less differentiated, have a real hard time, and may have a bit more success going through a platform. But this industry absolutely is willing to listen if you have a solution that is clear, and demonstrated, and that has a real unique high impact.
25:00 MG: Well done, but it sounds like you've got a few scars along the way, but given you were founded in 2014 and you've got some strong clients now, I think you're probably through that... That sort of tougher stage, which is getting yourself known and as you say, differentiating yourself from everybody out there and proving the value, but on that theme you're... You've got a team coming across to London in January and we're delighted to be working with you putting together a workshop, but you mentioned briefly the reinsurers. Of course, London is also a big market for writing the E&S non-admitted business, but can you just talk a little bit about what your expectations and hopes are for the UK and more broadly in Europe?
25:51 RK: We do have some scar tissue, but we've come through... I think we've got a very strong reputation, a bunch of referenceable customer successes to point to, and we've really built a significant and growing business here, focused on continental US-based risk. We're a privileged account leading top tier property insurance carriers, nationwide carriers, specialty carriers, super regionals, all of them as loyal customers and partners, as well as some of the world's smartest reinsurers. We have a few customers that are Lloyd syndicates. We've got some significant Bermuda-based reinsurers as well, and to date, we've been largely focused on US-based property risk. Eventually, we will be moving beyond continental US-based risks, but even before that, we are looking at other writers both global reinsurers as well as other Lloyd's syndicate partners that write non-admitted direct or reinsurance as well. This represents our initial push into a London market in earnest, and so, that's something we're very excited about.
26:54 MG: Good, now we're looking forward to it. There is a very enthusiastic and interested audience over here and as you know, London is a little bit like Las Vegas every day, in the sense that if you talk about InsureTech Connect and that you've got 7,000 people or many more talking about innovation all within a square mile or so... Much looking forward to putting together a lively discussion with a number of people from the market when you come over or when the team come over in January. More broadly around one thing that always fascinated me with people like yourself, that have got a huge amount going on and also need to stay abreast of what's happening in the industry, technology, and your clients, how do you personally collect information and process what's out there? What's your favourite means of getting hold of that?
27:43 RK: I try to always be very disciplined with how I'm spending my time. This means less passive listening, but I have a list of things I actively want to learn about and I try to prioritize and make sure that I'm spending my time appropriately and I look at the right sources. But you don't want to be totally close-minded to new things coming in. So just exploring new channels is important, we have a core discipline of, again, treating our customers like clients and partners, and trust and candor are some of the core values of our organization. And so what that means is it leads to a lot of, let's say, high-candor, high content conversations with our customers and partners, and we're able to bring in a lot of information that way. Some of those things just don't fit our strategic road map but some of those things do and sometimes we change our strategic road map to accommodate, and so that's really valuable. I think it's active listening to customers, actively listening to employees, empowering employees to be discovery-driven. Those are some of the tools, but there's not really a secret silver bullet, I guess.
28:48 MG: Yeah, and that is pretty interesting in this world of digital or increasingly digital, how much we all still rely on personal contact, and we're doing it now. And I think that's sort of, it's balancing that input isn't it?
29:02 RK: One of our core principles as an organization is really building high trust relationships, and this extends to our customers, our employees, and our investors. And one thing I do love about this industry is it's real people trying to solve real problems, and we're in highly recurring relationships with our customers, so it's a natural incentive to take the long road. There's a few media sources on this industry that I really like, but there's a lot of junk out there. There's a lot of hype and the good news is we've kind of eschewed that just culturally, but our customers have come back to us and paid us one of the bigger compliments I can think of, at least in my mind, which is, "Hey look, we really love what you guys are doing 'cause when you tell me something's going to happen or you're going to do something, we can believe that that's true." And I think that allows us to cut through a lot of the clutter and the noise. And unfortunately, I do think you see some of these more hypey events or media articles and it just ends up being a lot of noise.
30:06 MG: So Ryan, I've asked you a lot of questions, is there anything you'd like to tell me that I haven't asked you or you would just like to talk about before we wrap up?
30:16 RK: I've always enjoyed our interactions. We've spent some time together, you and I, in London and also over here on the West Coast. I'm thrilled that we're in the position we are and I think we're really well-resourced not only to continue to serve our domestic core market but I'm excited about this next phase for our business and just thrilled to have the partnership and an ability to extend. So, no, thank you.
30:37 MG: Thank you for your support for InsTech London and 25% of our listeners at the last count were actually in the US, so even though we've got London in the name, we're increasingly global. Well, Ryan, sorry, unfortunately I'm not going to be able to join you in Laguna Niguel next week. So you'll have to have a glass overlooking the ocean for me and think of us back here in rather dark and gloomy London, but it's been great carving out some of your time, particularly knowing how busy you are. So, thank you very much and best wishes the next few months. And again, I look forward to seeing your team in January.
31:09 RK: Great, Matthew, thrilled to connect. Thank you so much.
31:14 MG: Well, if you haven't already found it, we're publishing our transcripts from most of the podcasts on the website now, and we're also starting to release some of these as more condensed interviews, both on the website and through LinkedIn. So if you want a reminder of what you heard or just want to send some information on to somebody else, then keep a lookout for those in the next few weeks. We're also building out our scheduled events for InsTech London for 2020 and with a growing list of global members, we've got some excellent guests lined up for our podcast next year. Now, if you'd like to know more about our membership and how you or your favourite CEO can get on the podcast, drop me an email or send me a message via LinkedIn.