A great way to get attention as an insurtech start-up is to find a new area of risk to model, particularly when the founders have a deep history in insurance, analytics and technology.
The founders of Describe Data are Michael Crawford, Gerard de Vere and Mick Cooney and in this episode Matthew talks to them about their experience of being on the Lloyd's Lab cohort 2. After day two Michael already realised that "it was going to be amazing". When we spoke they were coming to the end of their time in the lab and they weren't disappointed by what they had been through.
In three weeks they had built a prototype for a product to help underwriters assess their D & O (Directors' and Officers' liability) risk. During their time at the lab they had 10-15 companies from Lloyd's engaging with them and they finish the cohort with a strong pipeline of opportunities.
The Describe Data team are usually based in Dublin, but travel to London frequently. Gerald De Vere is a founder member of Insurtech Ireland, which like InsTech London is a member of the Global InsurTech Alliance (GITA).
Transcript for this podcast
00:00 Michael Crawford: Do you have the buy-in from your board? Do you have the ability to get things done if we ask? And they went, "Absolutely." That was a game changer. On day three, we went, "Oh, this is just going to be amazing."
00:15 Matthew Grant: Hello, this is Matthew Grant, one of the partners at Instech London. In this week's episode, I'm talking to the three founders of Describe Data: Gerald De Vere, Michael Crawford, and Mick Cooney, at the end of their session in the Lloyd's Lab earlier this summer. Now, although based in Dublin, the team are regular visitors to London, we often see them at Instech London. Gerald is founder of InsurTech Ireland, which is part of the GITA group, of which Instech London is one of the leading companies. Now, it was really interesting to hear how the team have been testing out different propositions for insurers before landing on one specific area for where they were going to build their business, and this is in the space of D&O, directors and officers. Utilising Lloyd's Lab, the team built out a tool to help underwriters assess the risk and manage the exposure to this $10 billion area of risk.
01:10 MG: Now, their experience in the Lloyd's Lab was clearly a big success. Lots of engagement and support from Lloyd's and the underwriters. Describe Data was one of 10 companies to be selected from over 250 that applied for the second Lloyd's Lab cohort, and they provide some great advice for companies starting out, building technology solutions for insurers, whether through Lloyd's Lab, or indeed, anywhere else.
01:40 MG: Here we are in the Lloyd's Lab, sitting in a padded cell and all looking remarkably casual for being in Lloyd's, and of course, no sign of alcohol anywhere. Joined by Gerald de Vere, Michael Crawford, and Mick Cooney, co-founders of Describe Data, who are just about to wrap up their cohort in the Lloyd's Lab, I think working over the weekend to pull together their presentation for next week. So, great, and you were up on stage talking about cyber Instech London earlier this year. Is that still what you're doing with Describe Data? Or have you pivoted a bit since then?
02:14 MC: The answer is a bit of both, and we basically built a risk engine to look kind of agnostically at various areas of risk, and the first thing we looked at was actually life and health information from the UK, and we built a product for, basically, HR departments and quantified HR to help companies figure out what their employees look like on an aggregate scale. We then applied that to terrorism and kidnap risk, which is what we applied to the first cohort of the Lloyd's Lab with. And then, one of the things we looked at, because we're all technologists, is cyber. We looked at a few of the tools that were out there, and we talked to a few people, and we realized there was an opportunity to start looking into cyber and using the same approach that we had of looking at new data sources and using technology.
03:00 MC: And we did a lot of work on, basically, building a very unique cyber data set. And when we actually looked at the market, and we kind of done, after three or four months of looking at the market, we realized there was an awful lot of stuff going on. People are flooding into that market, and it would be very, very difficult to actually make an impression in any kind of time scale. So we reviewed what we were doing and looked at looking other areas. And one of the things we've looked at, and what we've done for the second cohort of the Lloyd's Lab, which we got into, is using the same techniques on directors and officers insurance. And that's a much more tractable problem in the time scale that we have in the lab, which is 10 weeks.
03:33 MG: Interesting. I'm sure there are people out there doing D&O, or Directors and Officers, but I'm not familiar of many companies doing that, so does it still feel like you've tapped into a market that's got a big potential?
03:43 MC: Well, the market globally is about 10 billion in premium, 50% US, 50% rest of world. Within the London market, and Lloyd's specifically, there are a lot of people underwriting D&O, and it's still quite a manual process. They use a lot of computers, obviously, to gather the information, but they collate the information manually, they build very simple reports, they look at financial information from companies, and then they use their gut feeling and basically their market knowledge to underwrite that. And what we've done is, we've taken that process and we've basically automated as much as possible, but still leaving the underwriter in complete control. So what our idea is to give them a better understanding of the risk that they're facing, being able to project it better, to be able to understand what emerging risks are coming down, and also to do something which is quite unique, which is look at their portfolios of risks and stress test them, and understand how does a new risk affect my portfolio.
04:35 MC: So these tools are really, really... They don't exist at the moment, and we've talked to the market over the last three or four weeks by building a focus group. We probably talked to maybe 30 directors and officers underwriters, and most of the meetings we have are along the lines of, "This is really interesting. We really need a tool like this."
04:53 MG: Very interesting. So why do you think no one's tackled that successfully so far?
05:00 Gerald De Vere: I think because it's a particularly difficult problem that looks like a particularly easy problem. For example, Mick, our CTO, went back to look at all of the ticker data for publicly traded companies in the US over the last 20 years. That sounds like a relatively easy data set to grab. It's actually not, it's very complex because companies merge, their symbols change, there's a lot of activity in that space. So that's an example of what seems to be a simple problem that isn't. So we were attracted to this because we're good at taking large, unstructured data sets and distilling useful insights out of them. And to be frank, we were a little surprised that somebody wasn't doing this. It was only when we scratched the surface and saw why people weren't doing it and there were some pretty difficult problems to overcome, and we think we have.
05:46 MG: And anything coming out at this stage where you're seeing themes that you might not have been identified previously by underwriters, or are counter-intuitive to what people might previously have thought?
05:56 GV: We're still learning and we're still extracting the data. We have noticed that there seems to be, and we've yet to prove this. There seems to be a good correlation between corporate health and cyber health. So in the work that we did initially on cyber, it's interesting to align companies that are run well, in other words, they have no class action lawsuits or no corporate issues. And the implication is, of course, that if the company is run well, then it is also run well from a cyber perspective in terms of CISO and business continuity planning. So that was one thing that we've noticed that we're trying to pull a little bit more information out on.
06:31 MG: Okay, and Michael, perhaps just in terms of the particular insurance classes, D&O, it might be worth more just for anybody who's not familiar with that to say a couple of words about it. And then also interested, just in terms of thematically, how you're looking for the risk to the balance between looking at the corporate risk, but actually also the people themselves, how much is that influential in terms of assessing the risk?
06:53 MC: Yeah, sure. Directors and officers insurance is basically a class of business purchased usually by companies to indemnify the directors of a publicly quoted company, or sometimes privately quoted company, against a lawsuit that basically is just filed against the company. Otherwise, the directors would have unlimited liability. And a lot of the time, let's say in the States, for instance, it's more or less a prerequisite because it's very hard to attract board talent if you're basically saying, "You come on board and work for me, and if the company gets sued, you're going to go bankrupt." So it's in place as a safety net for them. And it's actually quite... It's quite a niche area of business. It's something you probably wouldn't have heard of if you're outside the insurance world or the corporate insurance world, but it's actually quite a big business.
07:36 MC: Now we've found, when we're talking to a lot of underwriters, which... What we've done over the last minute. When you talk to them, all of them say, there's a couple of things, strength of the D&O. It's like how well the company's run. Basically whether it's a... If it's a large public company, there are certain things that will trigger a class action suit. You can't really do much about things like stock drops, which might be because they've done something in terms of, the company's had some kind of failing, but there are also malicious cases that come out and they all have to be defended. But fundamentally, what we hear from underwriters is that, D&O is a people business, you're underwriting people. So understanding what those people's probity are, what connections they have, and what other work that they've done is a very important part of that business. And that's what we're starting to look into the connections.
08:22 MC: Now, that information is kind of quite... It's relatively easily available. You can find out what other boards people are on and what other things like that. So the question is, at the moment, that's a process of... It takes a long time for companies to extract that information, and what we're doing is building tools, very simply, that do the simple stuff well, and very, very quickly.
08:40 MG: And the sources of data for that, when you're looking at people, what kind of...
08:45 MC: Oh, I mean, the things... I mean... You've got all the... Basically, if you look at the EDGAR database, particularly for the US, it's a massive kind of trove of data around company directors, and you can basically build network graphs of who's worked with who and that sort of thing, it's very, very indicative. But there's a massive information in there. The problem is, it's a very, very dense and huge data set. Even just trying to deal with it, it is actually... It's relatively big data, so you're into engineering problems very, very quickly. And then, on the other public side of the actual claims data, for security class action suits, that data is readily available but it's... Stanford have a large corpus of security class action suits, but it's effectively text data. It's basically court proceedings, etcetera, etcetera. So that's a very, very useful resource, but again, it takes work to actually extract quantitative information out it.
09:34 MG: And then, the brokers are obviously key in this, given that it's Lloyd's. Are you also, as part of the work you've been doing, engaging directly with brokers, to get their input into this?
09:42 MC: Yeah, we've talked to a lot of the primary insurers and we've talked to a few brokers, and they have very slight... They have... I mean, they're all interested in this because the D&O market is actually... It's hardening, the premium rates are going up. So people are actually interested in this as an area where people are starting to look into, to basically come into. The brokers tend to be more kind of like looking at it on a one risk basis. They're interested in this risk that I'm placing now, what price can I get for it? And what we've done, because we've done some quite... Even relatively straightforward modeling, we can actually figure out what the underlying rate of... What the underlying claim rate and the severity and frequency of a D&O claim is. And we can project forward and we can basically give you, more or less, a kind of value bet on how much are you going to... This is going to cost you if you take the premium.
10:32 MC: The brokers kinda look at it from a one risk point of view, whether they can get a deal. But the insurers have, A, they want to look at the risk, one risk, but they also think, "Okay, what portfolio, what aggregations do I have? Where am I lacking? What do I... Do I want to write business? I've got a lot of business in one sector, do I want to write it in another?" And we can start looking at that quantitatively as well. And that kind, it doesn't exist at the moment. That kind of non-geographic aggregation, it's probably bread and butter in things like hurricanes and storms, but not so much in things like D&O and casualty.
11:00 MG: Got it. And then, so you've built... Or you're building a tool, how do you anticipate that the underwriters would actually get access to the analytics? Do you have to build a separate platform for that? Or do you link it up through an API? Or do you partner with somebody else?
11:12 MC: Well, what we've done is, we built a prototype product, which is the full stack, effectively. It's got a front end on it, just 'cause we know that, to actually... The people who are going to use this tool are basically probably not the people who are going to buy the tool, because the IT department or the underwriting people are going to buy... Are basically going to be buying it, but the actual users are the people on the shop floor, effectively. So we've more or less built the whole stack of the product, but what we want to do eventually is offer this as an API and people can put it into their own systems. But we'll happily sell you a front end as well.
11:44 MG: Yeah, I think that is a really critical path for anybody building technology. There's two things there. One is, in my experience, a company doesn't need to have a front end that people can understand, visualise what's going on, but also recognising that not everybody will necessarily want to use that front end. So you need to be able to create the flexibility to go straight to the data using the API, and at some point, maybe the front end disappears, but it's very hard to sell something on the basis...
12:07 MC: You can't. We work in a start-up accelerator in Dublin and it's cosponsored by Google, and they come in once a month and do mentor meetings. And we've had some really interesting mentors from Google, who basically have said, "If your product takes more than half a day to train someone on, you're dead in the water. Solve the first five problems well." They're very straight. These are their ideas of like, "We do this for a living. These are the things that you have to hit. You have to build a front end. It doesn't... You can't sell an API."
12:34 MG: Yeah. And then, you mentioned the lab in Dublin. So that kinda bring us on to the Lloyd's Lab. What's the primary reason for applying to Lloyd's Lab? And how has that worked out in practice?
12:47 GV: The primary reason, as Michael said, that we applied to Cohort 1 as a very young startup was really to answer all the questions that the application process forces you to think about. So, what corporate structure would we think about, what products, how would we go to market, all that kind of information. So it was nice to know that we were all pretty much on the same page when those questions came through. We really didn't expect to do as well in Cohort 1, as we did. We almost got into Cohort 1 as a very young company, and we learned a lot from that process. So when the time came to apply to Cohort 2, about six months had passed, we could show some pretty decent progression in terms of the risk engine. Michael had already been on the stage for the pitch at Cohort 1, so he was quite familiar with the process. And I think it went really, really well. We got a lot of very positive feedback on the pitch for Cohort 2 and here we are.
13:40 MG: And is there a fast track process if you make it... If the companies that try for Cohort 1 or I guess those that try for Cohort 2, then the next time around they get to the front of the queue? Or it just depends on a thing like this?
13:49 GV: I don't think so. From what we've learned from the good people at Alphanumerics, the process is pretty linear and very well structured. So, it starts with an open call. I think about 250 to 300 companies have applied typically for each of the two cohorts so far. The market itself, through a number of various connections, helps to distill that down to a long list of about 60. Of that list, about 20 make the short-list and they're invited to pitch. And of the 20 that pitch, 10 are offered places. And in Cohort 2, there were two other companies that were already involved in Lloyd's who were also offered places. So there are 12 actually on Cohort 2.
14:28 MG: Great, so I'd be interested to hear how the actual experience on the ground works. So you turn up the first day fourth floor. You don't have to wear a tie and a jacket, so you wear some jeans. So what happens over the period of time you're actually part of the lab?
14:41 GV: We were very good about being suited and booted, and even cuff-linked, I think, on day one because we're quite familiar with the Lloyd's process. The first week was incredible, it was full on, there was a lot of information. We had speakers in from the LMA, from basically every different aspect of Lloyd's. We got to do the Lloyd's tour with the head waiter, really bringing people up to speed with how Lloyd's works and what it is, the essence of it being a marketplace, and the history, and the etiquette and so on. And then we really hit the ground running. We went straight into working and meeting our mentors and trying to get as much out of the process as possible.
15:23 MG: And back to your point, Michael, about engaging with the underwriters. So how does that process work? 'Cause everyone's busy, clearly if it's to be valuable you actually need to get some time in front of these people, initially you're probably not going to have anything much to offer them. How does that work and how willing did you find people were to talk to you at this fairly early stage?
15:42 MC: Okay, I mean when we signed up to the Lloyd's Lab, you sign up to a participation agreement as a company. And that basically says you can share your IP with Lloyd's and Lloyd's can share data with you. Lloyd's also has a mentor agreement, which is also signed where basically the mentors can come in and they're basically hand-picked from various companies. They self-select themselves to come in. And they have a very similar agreement, so they can share data with us and they can talk very freely and frankly about what their problems are. We signed this document and we thought this is great. This is actually how it works out. We walked in on day two to our first mentor meeting. There were seven mentors from across the market, all marquis names. They shut the door, and literally they basically went, they opened their kind of hearts to us like, "These are the problems that we have and this is we want to give you data."
16:28 MC: And we kinda said to them, 'cause we've been in a lot of these meetings like, "Are you people... Do you have the buy-in from your boards? Do you have the ability to get things done if we ask?" And they went, "Absolutely." That was a game changer. On day three we went, "Oh, this is just going to be amazing." So then getting access to people was no problem. You can say, "I would like to come in and have a meeting with you, can you put me through?" And no problem, just bang, bang, bang. And that has been facilitated by Lloyd's and LMarks, they've been but absolutely brilliant at that. So it is a very grown-up incubator. They don't really hold your hand, but you're basically given the tools and the access and just told, "Knock yourself out, kid."
17:05 MG: And is that, you think the experience of everybody else on the cohort? Or was it because you got a specific application that people really see as this value thing?
17:11 MC: A combination of things. I think because we're not youngsters, we know the way this market works. We have a lot of connections anyway. And we just came in and knew that what we wanted to do was come in and basically make as many connections and get ourselves and our product in front to as many people as possible. And that's what we've done. We've just taken every meeting to the extent that we've pushed off anything that wasn't core to the business, such as we've had a lot of interest from VCs and seed capital funds, and we've had to say to a lot of them, "Look, can you come talk to us in July? At the moment, what we're doing now is we're making hay while the sun shines in Lloyd's."
17:46 MG: And you're now at the point where presumably you start looking to get people to pay PIC or some kind of consulting project. How is that looking in terms of getting some revenue coming in?
17:57 MC: What we did was, when we settled on D&O week one or two of the lab, when we basically ran it by our mentors, some of them write D&O, some of them don't but they said they'd basically back us and they'd come and give us a help and they'd introduce us to people. So, we had a trade show halfway through this process at week five. And what we did is, we actually built a prototype in three weeks of our product, a real-life functioning product. It looked very ropey because it was a prototype, and deliberately so, 'cause you don't want a performance to produce a finished product. We had a trade show where we'd probably about 150 people come into that. And we got massive amounts of interest from that, and that drove an awful lot of conversations. From that, we actually went back to LMarks and Lloyd's and the LMA and said, "We'd like to put a focus group together of D&O underwriters." And they basically opened the books to us and basically contacted maybe 10 or 15 companies. And most of them have come back and we've had one-on-one meetings with them. We've then crowd-sourced what they like and what they don't like about that. And then in the last three to four weeks what we've written is that.
18:55 MG: And any sort of early indications of what you can charge for it?
19:00 MC: One of the things we came in here was actually what business model works as well. So we've asked people what business model. 'Cause people say, "Oh, let's charge per use or charge per license or charge... " And basically, we've actually asked those very specific questions. If you were to buy this, how do you wanna pay for it? So that's actually almost... What we try to do with this process in the Lloyd's Lab is basically build a product but also get product market fit and find a business model that works. So, if we'd come out of this with a product and investor ready, then this is, you know, that we've absolutely hit all our marks for what we wanted to get out of the Lloyd's Lab.
19:28 MG: Yeah, and I guess great advantage being here, you're either going to succeed well or not at all because this is the nature of the way the business works here. You're probably going to see more than double-digit users of the product or none at all.
19:43 MC: Yeah.
19:43 MG: It's unlikely that you and the business wouldn't survive anyway, but this is the way the market works. Well, congratulations on... It sounds like a very productive and effective use of time. So just switching back a little bit to Instech London, you've been fantastic supporters of ours, you've come across from Dublin. I think even especially sometimes to come and see us, just a bit interested to get your perspective on what you've made of that and any advice to anybody that hasn't yet been to our events, but what they can get out of them.
20:09 MC: Well, we kinda started coming probably, I mean I think I bumped into Robin at the digital garage in 2015 or something like that. I think this was like Instech London had just started. And slowly but surely started coming over to London, 'cause that's basically where all our business was at the time, we had a small consulting company doing bespoke analytics work for insurance companies, and it was just we knew nobody in London and it was an incredible way to make a network very, very quickly. And coming from Ireland, it's a small country, networks are really powerful. And people, I think people underestimate the power of a network and contacts, and that's one of the reasons we come back. It's also, you know, we met lots of people we know, we've made lots of friends, we've made lots of contacts through it. It's just been... I highly recommend coming to Instech London, but just in general, in any business, go out and find who are the companies that are doing something and go to those events and this how you... This is where you get your next job from, or your business from.
21:02 Mick Cooney: Yeah, sorry, one thing I like to say about this Matthew is, some people ask me like, Why should I go? And what I always say is, Because companies spend tens of thousands of pounds for market research, and you can just come to these events and well, for startups it's free, for the people there's a nominal fee, but you get fed as well and, you know I mean, I'm not... Sometimes there's free beer if you fancy it. So, it's the kind of thing that marketing companies charge a lot for and you just get it and you show up and you see a whole bunch of companies or just see what the trend is going on.
21:30 MG: Thanks, Mick, yeah. And actually, if anybody wants to see the Describe Data team in action, we can look at the, probably thousands of photographs now on the website. And, yeah, literally see how effective they are at networking.
21:44 MG: Thanks very much for carving out some time. I know you've got a lot going on. But it's been a pleasure and best of luck, and look forward to hearing how it all goes after you leave the lab.
21:54 MC: Thank you, Matthew.
21:55 GV: Thanks, Matthew. Much appreciated.
22:02 MG: Gerald, Michael, and Mick are now back in Dublin, but they are over in London regularly. Come along to our Instech London events if you want to meet them, or drop in and see them if you're in Dublin. Talking of events, we have an action packed autumn season coming up now. Our next event will be on the 24th of September where we'll be holding a reverse pitch for insurers and other large corporations to tell you what problems they are trying to solve. Come along, you may find your next client. Details of this event and/or other events plus soon to be released, the rest of the events for the remainder of the year at www.instech.london.