15 mins read

3 Secrets to Convex’s Success: Culture, Data and Operational Excellence

by Juan de Castro, COO, Cytora

This a shortened version of Making Risk Flow podcast, episode: “3 Secrets to Convex’s Success: Culture, Data and Operational Excellence”. This time on Making Risk Flow, host Juan de Castro is joined by Paul Brand, the CEO of Convex Insurance. In the episode, Juan and Paul discuss the three pillars of Convex’s success in the insurance industry: creating the perfect company culture, optimising technology to make data-driven decisions, and how operational excellence can give businesses a competitive advantage.

Listen to the full episode here

Juan de Castro: Hello. My name is Juan De Castro, and you’re listening to Making Risk Flow. Every episode, I sit down with my industry leading guests to demystify digital risk flows, share practical knowledge, and help you use them to unlock scalability in commercial insurance. Welcome to a new episode of Making Risk Flow. In some of our previous episodes, we spoke with many leaders of insurers that are looking at digitising insurance. Today we’ve got the opportunity to talk with an industry veteran with many years working for Catlin with one of the leading London market insurers. And now he’s using all those lessons learned from so many years to reinvent Convex. So it’s a pleasure to have today with me, Paul Brand, CEO of Convex. So welcome, Paul, and thank you so much for making the time to join me.

Paul Brand: Lovely to be here. Nice to have the opportunity of having a chat.

Juan de Castro: Great, let’s start with a brief overview of your background and role.

Paul Brand: I started in the industry back in 1982, with a major US company called the Insurance Company in North America that is now kind of locked away somewhere in what has become Chubb. The great thing about that business was I learned, so I was just given ridiculous amounts of authority at a very young age, way more than I would possibly grant. And that sort of led me to joining Stephen in the Catlin business. In 1987, I was employee number seven of that business top line of maybe £10 million turnover. And then between 1987 and 2015, when we sold the business to Excel, we grew it from that seven employees and £10 million top line to a $6 billion top line and about 2300 employees. I think one of the key factors that we had was just a bit more of an ability to use technology in slightly better ways than the competition. Now, sort of in the early 80s, technology was an ability to add up and people sort of laugh at you. As you said, I built a competitive advantage from adding things up. But you only need to sort of look back at sort of catastrophes like the excess of loss spiral and the various challenges that Lloyds had during that late 80s period and the early 90s to go control of your downside risk and how you think about it, how you monitor it, and how you create resilience in an organisation which starts with very simple just while we add the exposures up is a key factor to success. Because it’s not only the brilliant things that you do as an insurance business that really lead you to great outcomes. It’s actually a lot of it is not doing the silly things is almost probably more powerful.

Juan de Castro: Perhaps you want to summarize also your current role at Convex?

Paul Brand: So I’m CEO of Convex. Stephen and I founded the business back in, we got funded in April 2019. So very rapidly growing specialty insurer reinsurer based primarily in London and Bermuda. So our first capital raise was $1.7 billion. And then we post the further market acceleration due to COVID. We raised another one and a half billion dollars. We’ve got broadly 400 and 2450 people. Now, we’ve heavily relied on outsourcing and increasingly moving from outsourcing into automation to really build the business. And top line end of 2022 was a little over 3 billion and we’re still growing rapidly in 2023. So ambition for Convex, we want to be the best we can be in the market that we’re in. So as I think about that, London and Bermuda specialty insurance and reinsurance marketplace would like to be the best business in that market. And to do that, we were really focused on kind of three key strategies. The first one is really about culture and culture and behaviours and how we really bring a very different feel and reality of being in Convex to the business and use that culture essentially to make ourselves very easy to engage with, very easy for clients and brokers to do business with us. And a lot of that comes from you can smell how organisations feel and companies actually where people are happy and they’re enjoying working there. And there’s that positive sense and there’s that positive energy. Those businesses are much more attractive to do business with than the companies where you go in and you’re met with somebody who’s clearly just had the most awful day they’ve ever had in their lives and probably writing their CV’s up and thinking about how they get on. Why do you want to do business with people who are halfway out the door? Second pillar. And again, there’s a client focus to this in that the insurance industry wastes a huge amount of our clients’ premium by having really inefficient and costly processes. So use an opportunity created by technology to have brilliant operational outcomes because anybody can be cheap, but you’ve got to have brilliant operational outcomes, but for a lower operational cost. And as a new business without the legacy, without essentially being entangled with all of the systems and processes that have grown sort of organically in a lot of the businesses that, well, certainly the businesses I’ve worked in and a lot of the other businesses would consider around the market. And then last but not least, and not least in terms of the complexity of actually delivering it is very simple, make better decisions using data and technology. So going back to when I started in the business, any analysis that I did, I essentially wrote out by hand but quickly moved on to having desktop computers that we could actually start to use spreadsheets. I started using a spreadsheet called Symphony, which I doubt that anybody has ever heard of before getting into Lotus123 and then moving on to Excel, but actually sort of assembling the technology and the subject matter expertise and the ability to actually deliver, to actually make that third pillar of all right, now we can say we’ve got a decision making advantage because of everything that we’ve done with data and technology. That’s not easy. I’m happy with the progress, but it’s not something you can ever really put down.

Juan de Castro:

Definitely, that’s a brilliant framework. It’s that combination of great culture and behaviours with ensuring that you get brilliant outcomes at a low operating cost and better decisions with data, and obviously, you’re doing something well. I mean, you started the introduction with it took you 28 years to grow Catlin to a 6 billion business, and in four years you are almost there with Convex. So perhaps let’s go one by one through those three elements you just described and get your thoughts on each of them. So it’s starting with the right culture and the right behaviours. Where do you start? How do you instil the right culture?

Paul Brand: The first step is, well, who do you actually employ and what type of people are they? And then there’s a bit around how many people you’ve got. So I think it’s a lot easier to have a consistent culture when you’re dealing with an organisation which is smaller than when it’s larger. So if you’ve got thousands of people in an organisation, you can’t know them and they can’t know you. So an awful lot of behaviours really come from how people behave at the top and the style that they’re bringing to the interactions. Insurance is about having bumps, isn’t it? That’s what our clients are paying us for. They give us money to transfer risks on world balance sheets and then bad things happen in the world and we go, right, we’ll pay for that. Well, if every time that happens, leadership runs around going, I didn’t know you were going to have losses, it gets a bit nuts, doesn’t it? So having that, being able to separate out whether things have happened for a bad reason or they’ve just happened is a key part of it there’s, then I think a really interesting choice that companies make as to whether they’re fundamentally going to be driven by permissions. So before you do anything, you have to ask permission to do it. You need a clear, I want to do X, can I do it? Yes. Or whether you’re going to be run by constraints, which is you don’t have to ask anything unless you’ve been explicitly told there’s a constraint on it. So as I think about how we’ve set up our underwriting units, they’re very driven with that constraint model. So we put an awful lot of effort into strategically thinking about we want to be in this line of business because of these reasons, because of these characteristics. We believe are in that marketplace and we think these are the best clients that are in that marketplace and then we let them go off and do it. Obviously there’s peer review and obviously there’s challenge, but what we don’t want to do is say, well now every single decision needs to be agreed by committee X and so this just creates that trust within the organisation. So people have got the right to essentially drive a lot of the outcomes for themselves and I just think that gives a real sense of belonging and importance to the work that you’re doing. And then just the behaviours part is a lot of that is almost as simple as manners. If you say you’re going to do something, do it. If something goes wrong, don’t necessarily blow up and blame the person. Go is this a mistake or is this something deliberate that has happened separately between those two things? When clients have claims, treat them with trust. The last thing most clients want to do is have a claim. And if you go around as their insurer saying well, you’ve had a claim so you must have done something wrong. It’s a bad model and it gets right up our clients’ noses. A long time ago I had a motorcycle stolen from outside my house. The first thing the loss adjuster did was come around and say, I want to see both keys. Which is essentially saying I want you to prove to me that you’ve not sold your motorcycle to your friend and are now claiming the money back from you. So the first thing that my insurance company said to me is, are you a fraudster? It doesn’t engage people with the industry when you have those types of interactions.

Juan de Castro: Then we complain when insurance is one of the least trusted industries right for that type of behaviour exactly, right?

Paul Brand: Yeah. I think it’s one of the areas where technology can really bring potentially some solution because the reason why insurance companies do that is because some of our clients, and in certain business lines, quite a few of the clients might actually be fraudsters. So how do you separate the wheat from the chaff so you’re not treating everybody as though they’re a fraudster? Because some of them are. I think there’s a skill there, isn’t there?

Juan de Castro: Creating the right culture is quite an art and science at the same time. And then a couple of things when you were discussing like keep the team small, the other one is avoid bureaucracy and empower your people. Those are things that often when you talk to employees that have been in the same company through different stages of growth, that is often what they complain about. It’s like, oh yeah, ten years ago when we were small, the culture felt different. So by keeping it small and really showcasing those behaviours, it sounds like a really powerful tool.

Paul Brand: So I think for companies to be successful, they sort of need to create divisions. If you look at I don’t know how many employees Google has, but it’s tens of thousands. Okay.

Juan de Castro: Hundreds of thousands, probably.

Paul Brand: Hundreds of thousands, yeah. I’ve got no idea. But essentially they have people working on individual projects which are contained. And I’d be surprised if you looked at an organisation like that, whether those project sizes are above three, four, 500 people. Because as soon as you start to get over that, then I do think you reach a tipping point where that sort of swap between productivity, control and bureaucracy starts to go wrong and you have to start shifting your culture more into the permissions bias as opposed to the constraints bias. And that’s very pernicious for creativity and it’s very pernicious for a sort of sense of belonging. It’s really interesting because some people actually prefer the permissions type of culture because it’s clearer, it’s simpler. And you’ve seen this with the return to work, haven’t you seen different companies take different measures and approaches to how they try and get employees back to work or don’t get employees back to work? I think a lot of that was done far too quickly and was far too rules based. Convex have been very plain that we think actually, if you’re dealing with clients, you’re dealing with brokers, and you’re actually trying to create things, you’re trying to work as teams, well, you are likely to be much more effective in face to face than you are over video. But that doesn’t stop you from occasionally doing things by video. We’ve been very clear that we had a flexible policy around presenteeism before COVID why wouldn’t we have a flexible policy afterwards? But at the same time, you’ve seen a lot of other companies go, well, we’ll be three, two or two three, or all the teams have to come in on this day, and it’s just make it work. It’s the leadership and the manager’s job to go right, I need to get this stuff done, I need to do it to a certain high standard. How do I achieve that and is it working and are the people working to make those outcomes?

Juan de Castro: I think we initially said how culture and behaviour fits into outcomes, the right business outcomes. So we’ve talked about culture, which was one of the three pillars of your vision for Convex. Another one you mentioned at the very beginning was data driven decisions, which is obviously very meaningful, but many people think, well, where do you start? When have you started thinking about this in Convex, what were the first steps?

Paul Brand: You have to start with the data, don’t you? And I think that you got a lot of effort in other companies thinking about the solution, but until you’ve actually worked out how you’re going to get the data, and particularly how you’re going to get the data at a low enough cost and the ability to grow the amount of data that you’re using. Because if you think about the amount of data we’re using to make decisions today versus the amount of decision data that we’re going to be using to make decisions in the future, well, clearly in the future it’s orders of magnitude more will be available. So I think the first step is how a company thinks about data? How does a company think about its data model? How does it bring consistency to those elements? And then how does it identify assets in the external environment that it can bring in to essentially digest and transform to being against the data model which they’ve created?

Juan de Castro: The first point you said about getting the data? And I think this is a challenge most of the insurance industry faces, which is the data captured about risk. More often than not, it’s only about the bound risks, not even the quoted ones, but even for the bound ones it’s incomplete. So where have you started capturing better and broader risk data?

Paul Brand: Yeah, I think the reason that there are problems with the data that you just identified is historically people have relied on manual processes to capture that data. So people are reading insurance contracts, extracting information, and then entering them into core systems. But clearly, you can now do that essentially using ingestion tools to read that data. So one of the big efforts that we’re putting time and effort into is something we’re calling our data acquisition mission. So that’s starting with slip data, so policy administration data and policy administration systems, and capturing a certain number of core fields, but doing that automatically. Well, you said people aren’t capturing all of the submission data. Well, as soon as you’ve actually got a machine doing it, the machine doesn’t care whether it’s done 100, a 1000, 10,000. It’s broadly all the same numbers. The investment is making the algorithm which is ingesting the data accurate enough and teaching that algorithm to make it more accurate over time. And so, very easy step to go, well, we’re not just going to look at bound, we’re going to look at everything that is submitted post that point. And you start with 30 fields. Well, it’s not difficult if you can do 30 to move that up to 100 and then move that from 100 to 300. And you start with four or five lines of business and then you go, well now we’ve got it working for those, we can move that on to our 30 or so lines of business. And then you can get to the interesting point of well, okay, so now we’ve done policy admin. Well, what about exposure, what about claims? And you’ve taught yourself the techniques of how do I create a data model which is relevant for essentially analysing and abstracting the real world into our data model view of the world? And then how do I identify assets which we can bring into that database. And then I think slightly beyond that, you start going, well, data storage is pretty cheap. Just why don’t we put everything that we’ve got into typically what people call a data lake, and then as we get interested in topics, then we can start to structure that data on query. But the first bit is, okay, how do we move the people out of the process and actually move the people into the intelligent part, which is not the actual capture, but the interpretation of what this means? What is it telling us? I think you’re seeing sort of a number of places where there’s more focus on the outcome than that kind of foundational piece. And I’m a big believer in the kind of the David Allen get things done kind of technology of you have an idea of a destination and then you break it up into little steps. The first step in making better decisions by using data and technology is really about data and getting hold of it. I think a second step is working out actually what you’re trying to optimise towards. So as you think about the insurance industry, there’s a huge amount of emphasis on expected loss costs within a twelve month period. You look at frameworks, catastrophe models and other pricing models that are available in the marketplace, and that’s important. But actually, also we need to recognize that on an individual risk basis is likely to be very wrong. So however much effort you put into a pricing model, we know that they’re not going to be that accurate. And you can sort of see that. You look at Ian and everybody’s running around the market after Ian going, is it 20 billion? Is it 30 billion, is it 40 billion? And lots of people engaging in punditry. It’s a bit more, it’s a bit less and trying to sound well, okay, so we’re using the same models that can’t tell us whether it’s 40 or 80 to make pricing decisions, which are 5%, and that loss had happened. Okay, so arguably you should know a bit more about it. So clearly the game is not really the one that we think we’re playing, which is brilliance in estimating expected loss costs. We’re never going to get much. It’s very difficult to see us getting a lot better than we currently are, however much more sophisticated we make the models. So it’s really about actually how you think about the relative attributes of individual risks. So how do you rank them, and then how do you think about how resilient is my portfolio versus other people?

Juan de Castro: And when you say the relative attributes of the risks, are you thinking, if you could, at a very granular level, correlate the attributes of that risk against the loss behaviour? That is how you can identify the better risks from the worse ones or no, it’s something different.

Paul Brand: Slightly, but slightly different as well. So using a very simple example, so there’s a limit to how much US windstorm Convex will put on its portfolio. There’s got to be a rational gross and net after reinsurance limit. So if we’ve got two risks, which very simple example. Let’s say we wrote all of our US windstorm through two risks. I mean, obviously it’s tons more, but we should be able to go this one is better than that one. And then you get into a rational decision as well, if this one is better, then I will take more of that one and less of that one. That’s much more likely to be accurate than saying the absolute price of risk A is X and the absolute price of risk B is Y. It’s a much simpler question to answer because the assumption errors that you’re taking in the modelling offset each other.

Juan de Castro: So you’re talking about really being able to call it prioritize or call it score risks. How attractive is a given risk compared to another? Right, so that you can make those decisions.

Paul Brand: Yeah, exactly. And so you rank them and some of that’s going to be around how does the model price compare to the actual price being paid? But a lot of it’s also going to be around softer factors, like say you’ve been with this client for a period of time, they’ve had losses before, have they paid you back? How did they deal with that? It’s clearly an important thing. All the clients are giving us data which is enabling us to make projections of if there is this type of loss, this is what happens. How good was that data when losses happened? What is the quality of the underlying portfolio that you might be taking? So how well risk managed is an insurance client? How good is the underwriting of a reinsurance client? Are the types of questions that you should be asking yourself because as you really think about how does Convex get better financial results than its competitors in the marketplace? Well, it’s really by having an outsized share of the best clients and building those portfolios to be more resilient and use of reinsurance, that drives the outcome over time. And it’s not particularly as you yes, we’ve been yeah, Convex have been very lucky because so far we’ve only operated a market where prices have gone up every year. Well, that’s not always going to be true. At some point in time, prices start to come down. We can’t say, oh, prices are coming down, Convex are going to go out of business. So there’s got to be how do you think about the lifetime value of clients through all of the phases of marketplace? How do you put that into your rating assessment and your individual client assessment? And I’ve not seen in a rating model yet really other than in the personal line space where that concept is very clearly understood.

Juan de Castro: It’s a lifetime value prediction over different cycles of pricing. Right?

Paul Brand: Yeah. And the people who think it’s all about expected loss costs in a twelve month period, it’s almost as though they think they’re playing rugby, but they should be playing football. What they’re trying to optimise towards the decisions, the important decisions that they’re taking kind of have the wrong aspect to them.

Juan de Castro: Very interesting. I want to make sure we touch on operational excellence because as part of the initial framework you talked about cultural behaviours, data-driven decisions and operational excellence. So deep diving into this operational excellence, you’ve touched on some angles of this already, but how do you think about operational excellence at Convex?

Paul Brand: So I think it comes back at a lot of the culture and those two things are interrelated because the more bureaucratic you make your culture, the less likely you are to have bureaucracy, costs quite a lot of money as well as doing other strange things to the organisation. But we also when we, when we started the organisation, we had a choice we could have invested in our own servers. So these servers we’ve got the email, do you go and buy a bunch of servers, switch on Outlook, have lots of people managing them, or do you go right, well, we know that we go to Google, they can essentially run those servers for us and it’s a web interface as opposed to a desktop interface. So we go right, we’re going to use something which is cloud-based and we’re going to use something which is just the operational costs of running that are so much lower. And yet the actual outcomes that you get in terms of the usability is arguably just tons better. As you think about how do organizations collaborate. We’ve used a tool that was kind of given free with Dropbox, which is Dropbox Paper. And so we’re having a strategy meeting and we’ve got some leaders, some underwriters. You can actually see the notes and minutes of the meeting appear in real time and be edited and changed and sort of maximise the use of the cloud software as a service, minimise the bits that you can’t pick up. We decided not to have desktops in Convex. We all operate with laptops and that was driven because I spent a huge amount of my life in previous jobs on the road. So I needed to be able to work from everywhere because I was everywhere. So how do you bring that to an organisation and what does that do for you? Well, it means that when you get COVID, everybody goes home. Well, it’s no different. Everything still works provided you’ve got bandwidth, you can do everything that you were able to do before, get things done, have projects and actually deliver on them. So think about the ways that organisations work their objectives, very accountable outcomes that you can actually measure all of those just absolutely create an environment that is easy to use.

Juan de Castro: Everything you said just makes total sense. But I think this goes back to the bureaucracy and the permissions point you made earlier. Many organisations just potentially don’t have the courage to make those decisions of, okay, we’re just going to use Dropbox notes as an example, right? I think this is where I think your framework really reinforces each other. By having the right culture where people can make very sensible but different decisions from the ones companies have traditionally made, you’re really empowering that operational excellence that you were talking about. I think the beauty of the whole framework is how every component reinforces the other ones.

Paul Brand: Yeah. I hope you also always need to be looking over your shoulder to go, well, how do we improve it? What do we change? What’s the balance? Because it’s in that dialogue between permissions cultures or constraints cultures. You got to have some constraints. You got to be quite thoughtful about what those are. Otherwise, you can get obviously poor outcomes. But it’s sort of fundamental to a lot of it is dealing with people is are you prepared to trust and are you prepared to actually accept that other people can have brilliant ideas and actually drive brilliant outcomes? You get too many organisations and it’s a big thought of my own, which is you get people who are always trying to be the cleverest in the room. You get people I’m very solution orientated. I love solving problems. I won’t always have the best solution, but unless I can actually give myself the shut up or listen, then you don’t give space for other opinions and other ways of doing things to come to the fall. And that point you said about it’s got to be different to be better. If your ambition is to create a company that is better than the competitors, if you’re doing everything in the same way, and at times plagiarism can be a brilliant thing, but if you’re doing everything in the same way, then you’re going to get the same outcomes. I’m not doing Convex to be part of the pack.

Juan de Castro: That is, I think, a beautiful summary of your vision, you need to be different to be better and not be part of the pack. And Paul, I thoroughly enjoyed deep diving into this framework that you described at the very beginning of the three pillars of great culture, operational excellence and data driven decisions. Thank you so much for joining me today in this podcast and I’m sure the audience will really appreciate your insights, though.

Paul Brand: I hope so. Love to speak to you, thanks a lot.

Juan de Castro: Making Risk Flow is brought to you by Cytora. If you enjoy this podcast, consider subscribing to Making Risk Flow in Apple Podcasts, Spotify or wherever you get your podcast so you never miss an episode. To find out more about Cytora, visit cytora.com. Thanks for joining me. See you next time.