The Five Pillars of Data Analytics Strategy in Insurance

by Juan de Castro, COO Cytora

This a shortened version of Making Risk Flow podcast, episode: “The Five Pillars of Data Analytics Strategy in Insurance". In the first episode of the new series, host Juan de Castro is joined by his good friend and Deputy CUO and Head of Third Party at Inigo Insurance, Craig Knightley. During their conversation, the pair discuss just how data analytics and insights can play a pivotal role in gaining a competitive advantage in insurance and ensure underwriters are more accurate in risk selection. Plus, Craig outlines the five pillars of Indigo's strategy: great data, great analysis, a great underwriting workbench, portfolio underwriting, and partnering with clients.

Listen to the full episode here

Juan de Castro: Welcome to another episode of Making Risk Flow. Today, I've got a very special episode and I'll tell you why in just a second. In previous episodes, we talked about data and how decision support provides a competitive edge to underwriters, both in terms of productivity and also risk selection, and today I wanted to bring a guest from a carrier that makes data analytics one of the pillars of their strategy and competitive differentiation. So today I'm joined by Craig Knightley from Inigo. And before I ask him to give an introduction, I've got to say, first of all, he's one of the best people I've ever worked with. And I've told this to you, Craig, in private before, and we both worked together back at Hiscox on the fifth floor for a few years. But as importantly, I think he's an extremely rare example of somebody with a mix of a deep technical knowledge of insurance, an incredible strategic vision, and an ability to drive change. So, Craig, it's a real pleasure to have you join me today.

Craig Knightley: Oh, thanks Juan. So you've already embarrassed me at the start of this. I was slightly anxious and now I'm even more anxious. We used to work together and I share your opinion, you're one of the best people I've worked with too. So I appreciate the kind words.

Juan de Castro: And now I put you on the spot so you have to say that back to me. So let's start with an overview before we jump into Inigo and some of your strategic vision there, give us an overview of your current role, what you've done in the past and your background.

Craig Knightley: My background, I started off as an Actuary at PwC, so very much from the sort of analytical, technical background. After about three or four years, I ended up moving to Hiscox. I had a great time at Hiscox because I found an organisation that gave people like myself, in the early part of my career, a lot of different opportunities. Having come from a technical, analytical background, it was great to actually focus on more of the commercial side. And so I did various jobs, and ended up obviously working with you in about 2014, 2015, but it was a great place to learn. And as I moved out of the sort of actuarial roles, I moved into more commercial roles. I ended up working for the C-suite of Hiscox when we worked together, but then also I ended up as the CFO of the London Market Business Unit, which was the biggest part of Hiscox. And then finally I ended up heading up the casualty division of the syndicate, which was what I was doing until I left there in 2020 to become a founding member of the team at Inigo where my job is actually relatively similar. I'm the Head of the casualty division. And then more recently I've taken on the deputy CUO role. So I've got a bit of visibility across the various areas of Inigo. And more recently I've been helping sort of coordinate our data analytics strategy and working with lots of different teams in Inigo on that basically.

Juan de Castro: Obviously Inigo has been a huge success in the last few years, but tell us a bit more about how it started. What's the vision of Inigo? What's different about Inigo?

Craig Knightley: So the mid part of 2020, I got a call from a guy called Richard Watson, who you and I both used to work closely with at Hiscox and I think he’d been away and, COVID had struck, he'd been away with his other half. And I don't know if she was fed up with him or not, but they'd been away for three months and then he basically decided actually, he would see all sorts of market dynamics back in London and so he decided that it'd be a really good time to launch a syndicate. And he got together with Russell Merrett and Stuart Bridges and I met up with them for lunch and they were trying to look at the opportunity to build a syndicate that was basically primarily focused on 10 products, and not do 25 to 30 products that most syndicates do, but just do the core 10 products and overinvest in data analytics to differentiate ourselves. So what we've tried to do is pretty much we don't delegate at all. 40% of premium Lloyds is delegated. We do less than 1%. And what we try to do is leverage data analytics in the 10 core products to hopefully outperform the market average. You know, touch wood, that's what we're shooting for and hopefully, we can achieve that.

Juan de Castro: And what's your current size of GWP?

Craig Knightley: So year one we did about $430 million of premium. In the second year, in 2022, we pretty much doubled it to about 800 million or so. And we're on track this year to probably hit in or around 1.2 billion. So pretty much growing 400 million every year over the last three years. That's kind of slightly ahead of the original business plan. So we're really pleased in terms of the top line. The bottom line is on track again to deliver as per the original business plan. It's been a great start. We're now at about 160 to 170 people. Again, one of the things I would say is how we differentiate ourselves compared to what appears probably about 40 to 50 of the 170 people who are employed at Inigo are in an analytical role. Be it a data scientist, a CAT modeller, or a price actuary, we've got almost 50 people who are focused on trying to help us leverage data analytics. So we've over-invested almost a third of our workforce on the data analytics side. So we have put our money where our mouth is in terms of actually committing to it. And it's not just something that we talk about, it's something that we actually live by and we're trying to really make inroads on that.

Juan de Castro: You've mentioned a couple of times your vision of using analytics to outperform the market. And it sounds like a third of the employees are in those analytical roles. What do you mean when you say using analytics to outperform the market? How are you thinking about it?

Craig Knightley: Yeah, great question. In the last few weeks, we've settled on a goal. So our goal is to enable outperformance through unrivalled client insights. And at the most basic level, I think there are two things in that phrase. One is to say, we want to understand our clients' risks better than anyone else. So we really want to understand these and be able to price this risk better than anyone else in the market. And then hopefully, if we're able to do that, it should result in us having superior performance relative to our peers. Our goal is, like I say, to enable outperformance through unrivalled client insights. In terms of actually what that looks like, obviously it's hard to make that happen. And execution, I think lots and lots of insurers want to be great at data analytics, and I think we won't be alone in that. There are lots of people out there trying to do exactly the same thing. Where I hope we can differentiate ourselves is actually in the execution of that. And so there's five pillars that sit underneath that goal of how we actually want to make that happen. Pillar one is to have great data. So to have more data on those potential clients than anyone else. And when I go back to the start of actually what Inigo was trying to be, because we're only doing 10 products and we're not really delegating, and we're focused on especially insurance classes, what we can see is there's probably only about 10,000 companies that we would ever insure. And I think that's quite different to most insurers out there because they're spread between loads of retail customers, loads of delegations, 30 different products. And so their customer base, their specific customer base might be one to two million different customers, right? And obviously, we know back at Hiscox that the customer base was way beyond that number. For us, we're trying to really understand those 10,000 clients better than anyone else. We're not trying to sell all the world's mysteries. What we're trying to do is say, for those 10,000 clients, we've got better data than anyone else. So that's part one, have great data on those 10,000 companies. Part two comes back to a third of our company being focused on data analytics. We want to then use that data to do great analysis. So part one is great data, part two is then great analysis. So really it's about how do we then use that data to understand our clients better than anyone else? Part three is then we've got lots of underwriters at Inigo and we write about 3000 accounts currently. And so every time an underwriter makes a decision on each one of those accounts, part three is how do we provide them with a great underwriter workbench that showcases how that client looks relative to its peers, and what the pricing loss ratio is. But really how do we show those insights back to the underwriter at that point? And obviously, I know that's something that Cytora is really passionate about doing in terms of the technology part. So for us, really, it's about how we differentiate ourselves with technology and showcase those insights back to the underwriters, that's part three. Part four is if you're making those good decisions with each of those 3000 restained out, part four is then how do you step away from all that and look at this sort of macro? So how do you look back in the ivory tower and say, right, there's these 30, 40 different segments that we do that our 3000 clients add up to, how do we look at the profitability in each of those segments and how is each one performing compared to how we thought it would? And can we optimise there? So we call that portfolio underwriting. And the final part, the fifth part, which I'm really excited about is how do we partner with clients to help them work with us to better understand their exposures, better understand their risk? Is there anything that we can provide back to them to differentiate our offering so that it isn't just simply a commodity of insurance? We'll write you this policy and we'll pay your claims. How do we actually provide a service that is slightly different to other insurers? So actually we try to provide some of that insight back to clients. And so that's our fifth pillar, which is how do we actually build the brand around data analytics and partner with clients to share information back and forth?

Juan de Castro: That's going to be really useful to understand the different dimensions you think through to understand analytics. So you mentioned great data, the right analytics, a great underwriting workbench, the macro view of the portfolio, and how you partner with clients to help your clients better understand their exposure. Let's deep dive into each one of those. I think there's a lot to unpack in each of them. In the first one, you said you have great data, your understanding of the given risk and you have a limited number of accounts you would be willing to underwrite. What does that look like? What type of data are you looking at gathering from those clients?

Craig Knightley: Of those 10,000 clients, I think it's important to say that most of them will probably have a revenue that's over $1 billion. Not all of them, but quite a few of them, prospective clients and current clients, they're of that sort of scale. Now the good thing about those clients is typically, not always, but they're often publicly listed and there's quite a lot of information out there in the public domain. So there's quite a wide range set of data we can get on clients. At the most basic level, we want to know about the claims history for those 10,000 clients. And if they're a large corporation, there are four products that typically account for somewhere between 60 and 80% of their expected claims. So they are D&O, GL, Property D&F, and Cyber. So we would really like to understand the claims, the loss runs, for those four products for those 10,000 clients. Now typically what happens for most of these risks is, as a broker submits them to us, they will send through their loss run. And having seen this at other places, but what often happens is that it sits in a box, literally in their email inbox. They have a little look at the spreadsheet, you know, they might put it in a pricing tool, they might not put it in a pricing tool, they eyeballed it and then it kind of gets forgotten about. And sometimes it will be saved down in a pricing tool, but often it's not saved down in a database that is easy to interrogate at a later date. What we've tried to do is find a way that we efficiently coordinate and source that data for those four main products, those 10,000 clients. So the loss run hits us, we store it somewhere and then we can look at it later on. So that's sort of the slightly more boring insurance data that we already can capture. I think some of the things we're talking about are slightly more interesting. So for our property D&F account, if you are a large industrial company, you might have anywhere between 20 to 200 properties that you own. And often for these large industrial companies, what they'll have is an engineering report about the risk that's at that property. And so again, what the underwriter gets is they get sent a sort of zip file with 52 engineering reports, which is great. And each report is maybe 50 pages long. Unless you've got an extremely diligent underwriter, it's very rare that an underwriter is going to read and understand and be able to quantify 52 engineering reports at 70 pages long, right? So what we've tried to do is say, right, that's pretty valuable data in there. Particularly when you look at the red flags over time for some of these engineering reports, are companies closing down these red flags quickly, or are they letting them hang there, or are they increasing, what's going on? And what we've decided to use is some AI to rip out the sort of key 29 data points from each of those reports. And again, we put that next to those relevant sort of clients, you know, those 10,000 clients and we can actually see and track those engineering reports. And we play it back to the underwriters of CMPLA-3. So that's sort of an example. Some of the other weird and wonderful things we start to do is we've got a hypothesis that culture does affect the likelihood that a company will claim. We think there's some companies that are really, really diligent, that have got long-term ownership, that have got a seasoned board, that take risk management very, very seriously and overinvest. And our view is that if you've got that sort of culture, we believe that probably across products your risk decreases. So one of the projects we're now working on, and potentially line up with Cambridge University, which is the MBA, is to basically partner with them, do a project around how do you understand the culture of a company and does that correlate with the claims? And going back to my earlier bit that's sort of slightly boring about capturing all these lost runs, the reason that's so important is that if you haven't got the lost runs, you could say I think a culture of a company is good or bad. But unless you understand then how that correlates with claims data, you can't actually quantify what the increased risk is, right? So what we try to do is look at all kinds of things, right? From the boring of the just give me the claims data, show me what historic claims that you've got to the slightly less boring but pretty interesting engineering reports to the sort of slightly out-there culture. How do we look at social media sentiment, Glassdoor survey results? How do we look at a very broad set of data that actually tells you about the culture of the company?

Juan de Castro: And I think that is perfect. But actually, that specific point is something that we discuss quite often with some of our clients and insurers, which is some of the obvious traditional data points are quite straightforward to make a decision to start capturing those properties like parallel information, flood scores, etc. The tricky bit comes when we know we want more data to have a more granular risk selection or better risk selection. But where do you start? And I think this process that you just described is often what we discuss with them, which is you first need to be able to capture from your life submissions, or historical data, risk information and claims experience. Then you start with hypotheses, start gathering data around those hypotheses, in your case, the client's culture. Does it correlate with the loss history you've seen? But I think those steps are quite fundamental.

Craig Knightley: 100%. And I think it's a good thing for us at Inigo, where I think it's helped us, where we’re pushing open doors, as we've hired all the teams or we've hired these 170 people, we've all talked to them about how we want to be highly analytical. We've not inherited a culture where people say they will be analytical, but when it comes to it, they're not quite sure what that actually looks like. I think we've been quite fortunate that not only in the sort of 40 to 50 sort of data analytics guys that we've hired, also in the sort of other 120 people, they've all bought into that this is really important. So the reason I mention that is when you actually try and capture that great data, really, to a certain extent, it's not your data analytics guys that do that. They're the ones that are going to use it. Often it's then your technology teams that have then got to make sure this database is set up to do that. It's your underwriters that know they've got to then push the data to a certain inbox, so it actually goes into that database. It's your operations team that are excited about the prospect of actually capturing this data. It really does touch lots of different functions, not just people like me, sort of 10 years ago that are your numbers guys. It's really lots of other functions are required to create that great data. The people who are going to use the data often aren't the ones who actually are accountable for making it possible to store that data. It's often all these other teams. And so I think we talk about the culture of other companies and how we want to look at them in terms of their likelihood to claim. I think what's also really, really important about data analytics is having a culture whereby it's really important across functions. It's not just one function that thinks data analytics is important, typically data analytics teams. It's then all the other functions because otherwise, you know, it does sort of fall over, I think, and executional instincts are really, really difficult.

Juan de Castro: We've almost moved to the second pillar. You mentioned better data. The second one was great analysis or decision support for underwriters. And specifically, when you're doing that analysis, are you looking at analysing a given risk in isolation or are you also thinking about how the risk compares to peers at a similar company? So how are you thinking about it?

Craig Knightley: On this sort of analysis point, what we try to do with the teams, each underwriting team, is sit down with them and say, right, systematically, which rating factors do you have in your tool and why? What do you think drives risk and why? And then make sure that using the data in the sort of pillar one that we then use in sort of pillar two of analysis is to make sure that we then assess whether those hunches are correct? And sometimes the underwriters have got great hunches and we can show that data supports it. And sometimes those hunches don't always play back through the data and still there's some subjectivity there, right? And then, obviously, what we’re trying to do is always is then think through more and more rating factors you can add in and to give an example of that, you know, for our DLT team, more recently, we've had a few workshops where we've been chatting to them about what drives security class actions. Historically, we've looked at things like market caps, the size of the company, which industry sector they're in, and also when they go public, so how mature is the company? One of the things we've started to look at more recently is the short interest. So how much is that company shorted? And so we've then been looking at how do we analyse companies that are actually pretty similar on the face of it? But actually, if one is being shorted much more than another one, actually, we can sort of see in historical claims, all else being equal, they’re much more likely to have a claim. So then when we play that back and this sort of starts to touch on then, pillar three of how you play back to an underwriter, what we try to say to the underwriter is, here's a risk that looks like it's a pharmaceutical company or it's an oil and gas company, here's its peers, here's its loss record, but also how much it's being shorted relative to its peers. And actually, all else being equal, if it's shorted much more, we think here's the increased risk that drives in what we should then increase the price by. Or you converse that is actually, it's being shorted a lot less than its peers, actually, the short interest is decreasing. Okay, actually, all else being equal, you can decrease the price of that one because it's slightly high risk. What we're finding is that underwriters really value playing back to them and this really starts to touch not on pillar two but on pillar three, playing back to them, here's a company that you've written previously that's very similar to the one you're about to write and here is its loss record or how risky it is or isn't relative to that risk you're about to write. And often underwriters will talk about their guts, what we try to do is provide that sort of gut feel back to the underwriters analytically so they can kind of see that real time in how it compares. A lot of stuff isn’t revolutionary, it's just kind of putting the data in the hands of the underwriters in a seamless way that enables them to make really efficient and hopefully accurate decisions.

Juan de Castro: In those examples, it makes a lot of sense. How do you convince underwriters to trust that data as you're adding new data points or new insights or some type of metric that says this is a higher risk compared to its peer? What is the process of convincing the underwriters to make use of that data?

Craig Knightley: What I think we've got is slightly easier than some of our peers because we've started out with hiring people that really believe in this concept. We have a lot of these teams who are pushing open doors, we haven't got loads of legacy, so I caveat everything I say with that I think it is slightly easier for us than it would be for most others in the market. That said, I think what we try to do is sometimes prototype things that show underwriters what's feasible and what's possible. If we've got a slightly cynical team, first of all push the doors that are more ajar, semi-ajar, so go to the teams that actually are more open to it, hopefully find some wins there and if you get wins there then try to show some of those wins to other teams. The other thing that we've done is in one example, the team was a little bit cynical about a new factor that we found and what we did was we said, okay, fair enough, you're a bit cynical, we understand that and we can see why you're cynical, so we sort of parked it for a period but we catch the data on it and then about a year and a half later we could show actually the fact that we said we thought increased claims or decreased claims, we actually showed it, they play out in our actual experience. Let the team sort of continue as is but then what we tried to do is say, okay, maybe we are wrong, let's capture the data on it and if we're right we'll see that as we look back retrospectively and, fortunately, we could show that factor was a credible one when we look back at the experience.

Juan de Castro: I guess it's easier to do in a short-tail, blind than a long-tail one, right? But it makes total sense.

Craig Knightley: It is, but I think, again, when I was growing up as a sort of actuary, everyone would sort of say to me, oh, long-tail, you don't know until five to ten years what the experience is going to be. In your ultimate loss range that is largely true, right? It does take you typically three to five years for claims to actually be paid. But over time, obviously, quite a few policies are a claims made basis. You write the policy today, it's got a 12 -month period. And so although you don't know how much that claim is going to cost, what you can see is the claims notifications. So more and more, what we've tried to do is say, right, we don't always know absolutely what the ultimate claim is going to be. But I think what I underestimated in the early part of my career was if you can also have a really good assessment of notifications, actually, you can shorten that tail. If you're expecting 10 notifications that will be concerning, you end up with eight. Well, you know you've done better than expected. Your incurred loss ratio might be zero still, but you still know you've had eight claims versus 10. Or if you've had 20 claims, you expected to get 10 notifications, you know you've probably got an issue, even though you incurred a loss ratio, you're still zero upset. If you've got 20 things that are pretty concerning, you're expecting 10 things that are concerning. I think what you can do is a shorter tail. So I think some of that we are trying to track notifications, not just what the ultimate loss ratio is. And I think that helps from a sort of category perspective, but it's almost a separate topic.

Juan de Castro: You're just finding early indicators of frequency, even if it isn't going the full amount.

Craig Knightley: Completely right. That's all we try to do. Early indicators, we actually think the true performance of the account is. And the quicker we can get to those early indicators, the quicker we can assess whether what we thought was happening is happening or actually is something missing. If there's something we're missing, quite quickly what we try to understand is why is that? And then you go back to great data. If you've got great data, it will help you when you do your analysis to quite quickly figure out what is going on. So great data and having really good people to do the analysis is critical in terms of assessing how things are going, which is why those two pillars, I think, are so important.

Juan de Castro: Definitely. You've already started touching on what you call the underwriting workbench, what is the experience and technology that you provide to your underwriters? And I find it really fascinating the setup that you've put together. What does it look like? What is the underwriter setup? They've got two monitors. What's on each one, etc.?

Craig Knightley: I always think what would Apple do as they approach this business? What would they put on the two screens in front of the underwriters? What would they show? And the reason I use them as an example is I think obviously when you use your iPhone, it's just so easy to engage with. It's so intuitive and it's just so easy to navigate. So in my eyes, I sort of picture the underwriter sitting at their desk and there are these two screens there. So for most underwriters, they've got two screens in front of them. And as they get sent for a new submission on my sort of large 10,000 companies, I love the idea of it almost immediately flowing through, so we can see the client. So we don't start with who is the broker, this email is in the inbox. We start with who is the client and which risk we try to underwrite. So build almost a workbench in these two screens around this client. Unrivalled client insights. And you touched on it earlier on, like how they look relative to this. So who is this client? What data do we have on them? What claims experience have they had? How do they look relative to these? We obviously touched a little earlier about this idea of a culture score. Is there an Inigo culture score that's in the top right-hand corner of this screen? But really give you a picture of this client. So on one screen, you'd be like, this is the client. This is what's going on for them. This is what the recent news articles are on them. But this is, you know, a really good picture of the client in a very holistic way and compares them to their peers. So the underwriters look at that screen. Then they say, OK, right, I've got a good feel for this client and good feel for how to compare their peers. And on the next screen, I want a really, really robust pricing tool. So the two screens should be branded, our cards are sort of yellow and black, right? So they're sort of yellow and black branded and it all looks like one interconnected system. The client insight on one side and the pricing tool on the other. Now you and I both know that actually what we've built, because we've obviously chatted about this very briefly, is that it isn’t one system. But my idea, my Nirvana is that ultimately the underwriter thinks it's one system because it feels like one system. So they move seamlessly from one screen to another. It all looks and feels the same. What's actually happening in the background is we've partnered with Hyperexponential, HX, which I know you guys are also partnering with separately. And what we do is we use HX to build the pricing tool and the data then flows seamlessly from the underwriting workbench, which we built in-house, which is on the sort of one screen where it's quite insight focused, the data sort of moves seamlessly between the pricing tool. But from an underwriting perspective, they jump back and forth between this sort of macro view of the client, their claims history, their experience, the culture score, what's going on for that client in the news, you know, in sort of what the recent press article is about them. And then they can jump and get into the absolute micro detail of how we actually want to analyse that risk. And they can start to leverage the analysis and the data captured in pillar one and pillar two. So this is where the analysis and the data really come to life for the underwriter, because until this point, really, in pillar one and pillar two, of great data and great analysis, the underwriter hasn't actually sort of seen that exposed and where they should be exposed to that day in, day out is actually in these two screens, which is where we bring the sort of great underwriter workbench and this pricing tool to life. So I know I've gone off on a bit of a tangent there, but hopefully it kind of brings it to life in terms of what we try to do. And, you know, when I described this, some of this stuff is happening and some of it is fiction, you know, some of this is where we want to kick it to and some of it is what we're doing today, right? Somewhere in between those two things.

Juan de Castro: I find that setup really fascinating. And actually, it aligns very much with my vision of what the setup of underwriting should be in the future. At the very tactical level, it's removed one of the biggest underwriters' pain points, which is this double keying of data, because even if you had the two of them branded Inigo, you had to be copying and pasting fields that would create a completely different experience for the underwriters.

Craig Knightley: 100%, so what typically has happened right at most places is the underwriter gets sent an email, they then try and manipulate the data, either it's in a PDF or in its Excel file, they take that loss run, they try and copy it as best they can into an Excel pricing tool that's somewhere sitting over here, right? You know, like a completely different screen. So they're copying that across. Hopefully there's not too many errors in there, but sometimes there are. Then they're going back to their sort of, their policy administration system, their underwriting workbench, they're double keying back the data there. There's no comparison to peers. There's no comparison to the policy they wrote the year before. The underwriting notes are in a completely different word document that's saved somewhere else, which they try to open up and access. And what we try to take away is all that friction, right? So what we try to say is almost as it hits us from the inbox, that the data is already flowing through, the underwriter doesn't need to touch it. They can just see these two screens with everything they need to know and make it sort of really easy for them. But also we make it, you know, I'd say intuitive and insightful. The two words I need. Is it intuitive to the underwriter? Does it sort of feel very easy to engage with? But also is it insightful? Looking at the losses for this company versus its peers, I think is insightful. And I think that gets back to the underwriter's gut. Can you see, is this risk better or worse than its peers? If it's better, we should charge a bit less potentially. If it's worse, we need to think through what the appropriate premium is and what our line size is and whether we're a bit more tactical on this one or not. But ultimately, try to really try to get that back to the underwriter in a really efficient way.

Juan de Castro: I really like your intuitive and insightful underwriter setup. Let's move on to the fourth pillar, which is the second macro monitoring. So I guess we're moving now from an individual risk analysis into stepping back and looking at the portfolio. So tell us a bit more about that.

Craig Knightley: At the moment, we're about 1.2 billion dollars of premium this year for Inigo. We're right at about 3,000 accounts. And so if you think about that in terms of what it actually looks like day in day out, that's probably like about 15 accounts per day, right? So one every half an hour almost, right? And so obviously each underwriter that's here working is obviously incentivized to make the best decision they can for each risk they write each day. But then when we add them all up, what we try to do is make sure that actually top down does it look sensible? Is there anything we're missing? And we try to do that almost two ways. We group up the 3,000 policies into 40 different cohorts. So picture like a segment. A segment might be, as an example, general liability rail companies. We write about 10 million of our $1.2 billion premium is for rail companies for their general liability risk. And so what we try to do is say we're writing, I don't know, about 30 to 40 rail risks. When we add them all up, is that segment performing as we thought it was? Are we having more or fewer claims than anticipated? But also is there any extra data that we can use that will help us assess general liability risk for rail companies in a way that is different to what appears? And what we find actually, as an example, for passenger rail in the US is that 80% of passengers are travelling on 20% of the different systems. And guess what? Those 20% of systems, because they've got 80% of the passengers, account for 80% of claims. So again, are you getting adequately remunerated for those given they account for our 80% exposure, even though they account for only about 20% of policy cap, as an example. So we try to step back from it and see which segments are running best, which cohorts are running best, and how do you look to optimise that at macro level in terms of the management team? The other thing we do is, again, when we look at those 3,000 risks, about 250 of them account for 50% of the premium. So all things aren't created equal in terms of our premium for the things that we write. So some risks are really, really important to us and making sure that we're making the right decisions on those 250 risks is really, really important because they will drive our result. There'll be 250 risks that might only count for 1% of premiums. And obviously we'll make sure we made the right decisions there. There's 250 that might account for 50% of our premium. And then that's really important that we differentiate by size as well. So what we try to do at sort of management team level is take a step back, all these different decisions that the teams are taking and say, when we add it all up, does it look like it's performing as it should do? And so it's quite the macro view because I think, yeah, my experience at the last place as well was it had a really good culture, lots of people that are really incentivized to do the right things, but one of the things that made Hiscox really successful was actually they were able to see when there was an index that was really profitable, they would lean in quite quickly and when there was an index that they thought was less profitable, they would lean away. And I don't know if you remember this, but they used to send around the management accounts for all the different business units every month. And it created a culture whereby people were very focused on short-term profitability, which was kind of culturally another way of doing this macro thing, right? If you can show how things are performing day in, day out, I think it also helps underwriters and management teams shift their focus to say, actually, we're making quite a lot of margin here, how do we lean in? And then there'll be other areas where actually we're making less margin, how do we be a bit more tactical in the short-term and are we pricing this as we should? I think that's really, really important getting that right.

Juan de Castro: Definitely. And do you also have the concept of the target portfolio shape and do you use that?

Craig Knightley: I want to say yes. We obviously look at the sort of risk-free reward for different segments and, particularly for our CAT exposure. For a natural CAT exposure, we're very mindful of, you know, we obviously want to outperform the Lloyd's market, it’s our target. That's what we'd love to do. It doesn't mean we're going to achieve that, but that's what we'd like to do. We're also mindful that if you outperform, but you are taking on way more risk than anyone else, that isn't sort of true outperforming. That's just betting big and hoping that your number doesn't come in. And what we try to do is, you know, as much as we can get one in 20 to a one in 10 result into a fairly modest range of outcomes. So again, part of it is then how we use data analytics to kind of help understand the risk profile at a macro level, right? So I think all of that sort of chat about how your portfolio looks, how much reassurance you buy, how you make the one in 10 to one in 20 result a bit more predictable, all of that stuff is sort of under the banner of pillar four. So it's trying to do a whole load of different things. But it's really when you sort of step back from all of this, how do you use analytics to assess that risk and that real term. We are kind of looking at risk reward for different segments. We try to look at the 40 different segments of how they perform just generally. And we also try to look at things by size. How do we make sure that on the biggest risks that we absolutely got it right? And if we haven't got it right, how do we quite quickly call spread?

Juan de Castro: Let's move to the fifth pillar. How do you party with clients to help them understand the exposure and then being something above and beyond just an insurance policy?

Craig Knightley: Yeah and I think that this is the one that's probably the hardest, but also the one that's most exciting. So I think you picture a $1 billion plus revenue company. It might be fiction that we can almost turn around some of these companies and say, this is how we look at you in terms of your risk profile. Here's something that we can tell you that's insightful, right? So this might be mission impossible, but I'm hoping that it's not mission impossible and that we can actually provide something insightful back to clients. And actually they will hopefully value us for that. And so we have, you know, started to get a few examples of this. Again, if you can actually engineer reports with different clients and look at, you know, the relative to their peers, obviously we anonymize that, but you know, we can start to say actually, how is your risk profile looking relative to your peers and are you more diligent than others out there? I think clients find that interesting. Looking at things like, given climate change, looking at which of your properties are most exposed to climate change, which ones are most exposed as obviously things like flood and the risk of flood increases, which of your properties are most exposed and giving our perspective on that, I think clients find useful. But there's also, one of the recent things that the guys have done is they've worked with companies, with pipelines in the US. So we managed to source a very extensive data set about the risk of pipelines in the US and again, we've been able to sit there with our clients and show them a dashboard and show them relative to their peers, again, anonymizing their peers, but show them how risky they are relative to their peers. And often the risk manager is the person that is buying the insurance. And from their perspective, they find it really interesting. And if we can tell them something different about their risk than what most other insurers are telling them, I think they really value that interaction. So we're starting to build quite a lot of traction there. And I think it just changes the conversation. If a client has flown all the way into London from the US, sometimes extreme Australia, from Canada, from any one of these places and they flow into London and they sit down with one of our underwriters, we want our underwriters to walk into that meeting, be really knowledgeable about that industry, but also able to articulate the risk in a way that actually the client finds interesting. And I think if we're trying this thing, as long as it doesn't come across as egotistical and it comes across as actually we try to understand your risk in a different way, and here's how we're thinking about it. Here's some interesting graphs. What we find is that the conversation we have with our clients is very different than if we just turn up and say, hi, you know, we're from Inigo, it's going to be very similar meeting to all the other nine that you just had and very similar to the nine that you're going to have later on today. It just immediately changes the conversation. So we started to find that actually clients and brokers really value us, at least trying to do some of these things. To me, this would be nirvana. If we can sit down with clients and actually help them understand their risk in a way that they value, I think we'd have really reached our goal, right? Because obviously, our goal is to enable our performance through unrivalled client insights, unrivalled client insights is a pretty ambitious goal, actually. I think pillar five is, if we can get there, we'll see it in how we interact with clients. And so I am really excited about us trying to make it happen. Doesn't mean we will, but I'd love for us to get there. And I think the team's really where we get a lot of-, you said earlier, one of the questions you had about what do you do with underwriters and how do you navigate that, I think when we can sort of bring this to life for underwriters and they can actually see that they're building their brand, the Inigo brand, from how they interact with clients. That's where we get them, right? Ultimately, if we feel like we can empower them, they love that sense of being able to turn up to a client meeting and talk about risk in a different way. That's where we feel like we've really done our job and actually the underwriters really value us at that point.

Juan de Castro: And it's got to be really fulfilling for the underwriters when you can outperform the market in terms of risk selection and provide value to your clients. But I think this goes back to creating a fantastic, almost like employment brand for your underwriters. You cannot get any better.

Craig Knightley: Completely right. And that's where my best moments at Inigo have been where I can see the actuary, the data science, and the underwriter there,  partnering in these projects. And then I've had the examples where the underwriters then gone and presented this insight back to clients. And then afterwards I get copied on email where the underwriter says, got all this great feedback from the broker and from the client. They've CC'd in the actuary and the data scientists. And like, it's just a great feeling for everyone involved. The underwriter is really pleased. Data scientist is skipping into the office. And it's amazing to be part of it. And the reason I love that moment is because I've gone all the way from being that prize snatcher, right? That was my job. I used to sit in the corner of the room, it was literally the corner of the room, with other actuaries and 95% of our conversations with other actuaries in that corner of that room, and it's, for me, amazing when you see the data scientist, the actuary, and the underwriter collaborating and able to actually hopefully add value back to the client and hopefully for us understand risk better. It's also just amazing, the culture that you start to build from that is-, those are my best days at work. Right? I love it, it's not every week that it's happening but when they do happen, it's really fulfilling. And I’m just really keen that we replicate more of those successes across Inigo. The more that we can do that, the more successful we'll be and we'll build a great culture. I think there's a lot that we can get out of data analytics, which is also the culture we build, not just the numbers.

Juan de Castro: This is fantastic. And I think as you went through the goal and the five pillars around better data and great analysis, the underwriting workbench, the macro monitoring of what's going on and this being helpful to your clients it really brings it to life. That's my last question. What's the ultimate ambition of Inigo? Where will you be in a few years' time?

Craig Knightley: It's a good question. I think our goal is to try to absolutely deliver on these promises we made and the goals that we've got. So I think if we've got 200 people that really enjoy working here, where they’re building long-term careers. I've been fortunate that people are giving me opportunities over time. If we're creating that culture where people have got those opportunities and we've got high retention of people and we're able to bring lots of these things to life such that actually we can turn around and say, look, we do know these 10,000 match really, really well from great data, great analysis, fantastic underwriting work. We can look down over the top of it and see that it's all working as it should and then provide these insights back to clients and work with them to build further insights. If we can make that happen and people really enjoy working here, then I think everyone around the sort of exec table at Inigo will be really, really pleased that they've been part of that. And I think our investor base, if we do all that and then make sure we actually do outperform, our investor base will be really, really pleased. And now the challenge for us is that the market cycle is always doing what it does. It goes up and it goes down. So there'll be times where, as we build that company, there'll be times where we can absolutely maximise that opportunity and we'll grow like we've done in the last three years, like 1.2 billion or more. And there'll be times where actually the market is not in our favour and we won't grow as aggressively as that. But I think if we've got the core differentiation of what we actually build, then the market will eventually create opportunities for us to actually realise value from that model. So I think I'm ready to absolutely deliver that goal. And then I think the upside of that in terms of premium will be what it will be. And there'll be years where we write loads more and there'll be years where we write a little bit less, but it will play out and we'll actually make very informed decisions and hopefully partner with clients for the long-term in a way that is sustainable for them and us. So I'm just really excited about hopefully making that happen.

Juan de Castro: I think what you're building is a really sustainable business that can thrive both in hard markets and soft markets and is able to react quite quickly to changing market conditions.

Craig Knightley: Completely. And I think if you break down what we actually try to do, it’s to understand for each one of those 10,000 clients, what is their risk of claiming any one policy at any one point in time. And if you've got the most accurate assessment of that, then ultimately there's a business model with that, right? Because you can partner with companies in the long term because you know what you think the true price is. So if you just keep going for that goal of being able to answer that risk better than anyone else, then you've got a sustainable model, right? And so for us, that's the mission. And we just have to make sure that we absolutely deliver that mission and don't come on podcasts and talk about it as if we've done it all. But it's still a long road. And that's why I suppose, you know, obviously we've worked together many moons ago and we talked about that already, but ultimately, as time goes on, I always underestimate the hard graph that comes in execution. And I think you could leave the business strategy free to go on the street. The challenge in all these businesses, I think, is execution, right? Have you got a great culture? Can you execute? And that's the challenge for us now, right? Is how we now go out and execute.

Juan de Castro: Definitely. It's been amazing, as always, catching up and doing this episode together. Obviously, the conversation has been incredibly insightful, but even more importantly, it's been very energising to hear your clarity of vision, how excited you are about making progress in this area. So all I can say is thank you so much for joining me.

Craig Knightley: Thank you