Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
009 - Data Driven Disruption: Carbon Underwriting's Modern Approach to Insurance
148 Plays1 year ago
In this episode on Stacked Data Podcast, we explored how the team at Carbon Underwriting use the modern data stack to innovate and revolutionized the way they operate in a the traditional industry of Insurance, allowing them to swiftly deliver top-tier data products that can exponentially improve work in the insurance space.
We dive into there leading data product “Graphene”, taking an old school Excel reporting, consolidating this all in a common data model and surfacing an analytics layer on top! This allows there underwriters to deliver work at a much faster rate.
This episode is all about how to build teams, infrastructure and processes to create data products that are truly impactful.
Key Takeaways:
Start with your customer: Work closely with your stakeholders to understand their pain points and mission, from here you can start to design a data product that can enable them.
The power of leveraging the modern data stack: The use of cutting-edge technology has transformed there data infrastructure, enabling faster, more agile operations.
Accelerating product delivery: How this modern approach has supercharged our ability to bring data products to the market at an unprecedented pace.
Overcoming challenges with efficiency: We discussed the hurdles faced and the strategies employed by the team to harness the potential of the modern data stack effectively.
Recommended
Transcript
Introduction to Stacked Podcast
00:00:02
Speaker
Hello and welcome to the Stacked podcast brought to you by Cognify, the recruitment partner for modern data teams hosted by me, Harry Golop. Stacked with incredible content from the most influential and successful data teams, interviewing industry experts who share their invaluable journeys, groundbreaking projects, and most importantly, their key learnings. So get ready to join us as we uncover the dynamic world of modern data.
Meet Ferg and Nav from Carbon Underwriting
00:00:34
Speaker
Hi guys, welcome to the Data Stack podcast. Today I'm joined by Ferg and Nav from Carbon Underwriting. It's a pleasure to have you on guys. How are you doing? All good. Yeah, all good. Cheers. Cheers for having us on. Brilliant, brilliant. First, for the audience guys, it'd be great if you could just give us a quick intro to both yourselves and Carbon Underwriting and what you guys are all about.
Fergus's Journey to Carbon
00:00:56
Speaker
So I'm Fergus. I'm the lead analytics engineer here at Carbon. I've been at Carbon for two and a half years now. I was one of the first sort of 10 employees and the first sort of non-senior hire to the tech team. Previously I worked in payments fintech and various bits like that.
00:01:12
Speaker
Yeah, I probably started pretty much around about the same time as Fergus and been here like it's almost the same time and I've had a sort of exciting background in like data consulting before that, like, you know, developments, software development, built startup products, data products for various different global enterprises before moving into carbon and this exciting new insurance proposition.
Founding of Carbon and the Vision
00:01:38
Speaker
Oh folks, interesting journeys and give us a bit of an overview of carbon underwriting and what you guys really are all about and deliver. Yeah, so carbon essentially stemmed from our, we've got our three founders of Nick, Ben and Jackie. They worked for a pretty large insurance entity beforehand and they sort of realized that the data that they had wasn't being sort of used to its full potential.
00:02:03
Speaker
And so this led to them sort of leaving, start up their own sort of syndicate. They started on the syndicate in a box scheme, which is sort of the Lloyd's startup scheme for insured syndicates. You go through quite a rigorous sort of three year sort of process. But what you really need to start a syndicate in a box is you need a sort of unique value proposition. You need something that's innovative in the market.
00:02:24
Speaker
the idea was there was sort of these large incumbents that have been going for sort of upwards of 300 years and they're sort of very happy to sort of trundle along and say, well, we're still making money.
Innovation with Lloyd's Syndicate and Graphene
00:02:32
Speaker
There's no need for us to innovate. Lloyd starts up the scheme so that we have innovation in the market, whether that's sort of a new tech product, which is in our case, it might be there's a venture capital sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of, a sort of,
00:02:46
Speaker
So we've been on the sort of three year process through the Syndicate and Box Scheme with Graphene, which is our tech product, which is our tech product being our sort of, we've recently graduated being a full Syndicate, which means the sort of the training wheels are off and we're allowed to go sort of in the underwriters terms, gangbusters, arms of new territories, new classes, without these sort of like sort of restrictions that Lloyds puts in.
00:03:09
Speaker
Amazing, so your business was born out of innovation and I think that's what we're going to dive into today. You've already mentioned it, your tech product, your beta product Graphene and that's what today's episode is all about, the creation of data products at scale that are going to essentially drive business value.
00:03:32
Speaker
You've already mentioned insurance, a very traditional business, been around for a long time as an industry.
Tech Adoption Challenges in Insurance
00:03:40
Speaker
How are you guys using tech to innovate in your industry?
00:03:44
Speaker
I think one thing we all can agree on that with an insurance market here, from a technical point of view, we're not building rockets. We're not building tech that is as advanced as launching a rocket. Very simple problems, very simple data problems, but faced by a lot of cultural change. What we found in insurance specifically is people being working
00:04:10
Speaker
in non-data tech culture for a long time. And like any other industry, tech comes in and promises a whole new world. But as we humans don't often like the change, then there is this cultural aspect which comes into the place. Because to adopt new tech, you get an ultimately
00:04:31
Speaker
become the tech and embed that culture internally, and then go through that change process. And I think one of the biggest challenges that we still try to tackle with, and any other intro tech or any tech startup within insurance try to deal with, is how do they
00:04:51
Speaker
apply these technical solutions with respect to the culture of the industry, and then they change the culture in a positive way. So using tech and data just becomes unknown. And then from there on, you just copy-paste solutions, models that have worked across all of the other industries, whether it's marketing, retail, or manufacturing.
00:05:23
Speaker
But once you've got that cultural grip, that understanding of what are the key triggers, how you can seamlessly transform this industry or a business with the use of tech, but also not disrupting the cultural flow of the way people work. And I think that you can easily copy paste a lot of models out of the industry. And that's what we've sort of done and are doing and probably will be doing for some time in the future as well, because copy paste those simple models of technology
00:05:36
Speaker
These industries have been innovated with data for quite a long time.
00:05:53
Speaker
and apply that in here, see the value, move forward, if you don't see the value, go back to the sketchboard again, rewrite it, come back again.
Graphene's Impact on Data Use
00:06:04
Speaker
Just to give maybe a brief overview of what graphing is and does, the problem that it's solving is not being able to use your data in the correct way or an efficient way. Previously, in our founders last company, it was essentially run on Excel.
00:06:20
Speaker
with so every single sort of business area had their own sort of quite large excel files that made up their database for their individual business unit. But combining that up to these sort of higher levels or combining sort of across classes of business was something that just wasn't done or took months to do because everyone had their individual data model essentially that they'd built
00:06:41
Speaker
in order to service that business area and so we're doing is combining everything together into a common data model and that is really the core of our product is a common data model with an analytics suite built on top of it so what we do is get all of the data from essentially insurances premium claims and risk data.
00:07:00
Speaker
We get all of that data in from our different sort of cover holders and MJ's. We standardize it all into our common data model and display that back to the underwriters in a variety of different formats. Obviously the most basic being sort of your standard BI of sort of dashboards. Then we've also got a lot of other solutions that we've got plugins for Excel.
00:07:18
Speaker
So anyone that's still a little bit scared of using dashboards can download the data exactly the same. It's exactly the same format, but straight into Excel. Similarly, we've got a sort of a natural language based search solution where you can search for the insight that you want. That would generate SQL, which will then sort of provide you with the insight and sort of a variety of other tools aside from that.
00:07:38
Speaker
the fact that folks had earlier about the bringing like people got a lot of Excel sort of stuff legacy various different ways of combining them but what we focused on and you know extend on a lot is our common data model that plug-and-play data model can you just bring in any cover all the data into it and then see the performance out just plug that data in and see what comes out that is sort of I think the real sort of
00:08:06
Speaker
value that we get out of graphene and helps us making it into a full data ecosystem. And now plug it into an app, whether Excel, mobile app, a website, a dashboard or whatever, but you see golden standard data everywhere.
00:08:23
Speaker
Brilliant. So you guys essentially are taking a very outdated system. Excel was obviously widely used, I think, still in every business. And I don't think Excel will ever die. But I suppose, in a way, creating that single source of truth in terms of a database, connecting all the businesses' data, and then being able to allow them to action and gain insights on it at a much quicker rate.
Enhancing Business with Graphene
00:08:49
Speaker
Amazing, so at Carbon Underwriting you're clearly leading the way for data transformation within the insurance world. It doesn't seem like this is being done in many other areas. Are you able to give us a bit more of an overview of your platform, your tool and graphene and how it's specifically exploiting business value?
00:09:12
Speaker
I feel like I answered quite a lot of that in the previous one. So maybe we jumped ahead a little bit there with that. But essentially what it provides is this ability to chop and change your data up to sort of a portfolio level and down to these individual risk levels.
00:09:29
Speaker
So when we're giving this tool to, we give this tool to sort of everyone in the value chain with Carbon. So we're trying to provide this value to our sort of cover holders, give them back data that they wouldn't otherwise see. A lot of them will just, again, they work primarily in Excel files. They have no way of checking the performance of their portfolio. They often don't have links with their sort of TPAs that they, the TPAs, those sort of claims handlers. So they won't be able to see exactly which areas are sort of, the areas of profit and loss.
00:09:58
Speaker
And we can see that not just at the individual sort of coverholder level, or we can aggregate that up to see sort of how a plumber's in Australia doing overall and provide that feedback to our cover holders to enable them to sort of write their business better. Similarly, we can also provide this data upwards to our sort of reinsurers.
00:10:15
Speaker
this whole portfolio level we can provide them access to see sort of like which countries are forming well or sort of which sort of general classes if they sort of we have re-insurers that will do sort of our entire portfolio. They're not interested in the sort of risk level data although they can drill down into that sort of level if they want to. Yeah and I think that's pretty much the value that provides. It's different to everyone within the chain. Everyone wants to see it in a different way but you've got to provide them the ability to go down and see.
00:10:40
Speaker
Arrangeo may want to see an individual risk that they think is causing a problem, but they generally want to see the other portfolio level. Amazing. So it depends on the use case for people that are using the products and how they want to slice and dice it and to gain the insights in which they're looking, but your product gives them that malleability in how they can use it. Yeah, exactly. Yeah.
00:11:01
Speaker
Perfect. So look, you've clearly identified an excellent use case for a data product and it's sort of how your business has started.
Technology Stack and Customer Focus
00:11:08
Speaker
So can you talk us through how this project started, how you've identified this area and how are you doing to ensure the long-term scalability and success of a data product and platform like this?
00:11:22
Speaker
That was a very deep question. I think we can keep talking about it for ages. But yeah, a few points I'd like to search on. Starting this famous saying, Jeff Bezos as well, start with your customer and work backwards. Often when you're building data products, what happens is data companies harvest a lot of data, bring about a lot of data, and then go upward from there that, hey, we've got this data, and we can offer you this and this.
00:11:51
Speaker
But then you often find that the consumers of those insights are often looking for something else. And what we've done here is, because we sit very closely to the business, the underwriters, is working backwards from there is like, we've got this. We've done exploratory data analysis. Our data can give us x, y, z. But you tell us what you need. And then we can work with you and work backwards, trying to, first of all, generate those insights.
00:12:21
Speaker
Once we have that, how are we going to surface to, like Asperger said earlier, we've got various different stakeholders. We have platforms used by co-holders, platforms used by read drawers, platforms used by TPA's, platforms used by times products as well, platforms used internally for various different sort of answering business questions. And how do you ensure that sort of technical scalability around it? So we've,
00:12:50
Speaker
We've started with a really hybrid slash monolith approach there, from technology quite a few. Building our stack on Google Cloud, first thing first, is because what we've seen is Google is, from the last few years, supporting the Insure Tech innovation.
00:13:13
Speaker
quite heavily and their auto ML solutions, data solutions, data processing solutions are probably one of the cheapest across the market. Like there's Azure, there's AWS, but as a startup, you look at also that speed to from getting your product from your code to deployment and what's how fast you can get it out. And they kind of, they sit sort of, you know, I think at the
00:13:41
Speaker
sort of top of the list of those cloud providers which help developers or tech companies to achieve that quite seamlessly. Once we've answered that sort of big question about this is going to be our house to build tech on, went on about because this platform is going to be consumed by various different parties,
00:14:04
Speaker
what technologies, what infrastructure do we support, but also being mindful of that it is a startup at the end of the day. We have a limited number of people intact to begin with. How do you avoid those people spending time on various different services? It's exciting to build everything and anything microservices, but then how do you support that? Because the more you write, the more you have to support, the more you have to maintain.
00:14:28
Speaker
And then the value you're producing is actually, you know, not matching the effort that you're putting. And then we've always tried to look at the innovation and the development from this angle that the value over effort lens is whatever effort it should be getting more value out of it. Because as I said, very beginning, we're not building rockets here. It's very simple tools, very simple technologies, copy-paste models.
00:14:54
Speaker
So we initially went with quite monolith slash hybrid approach, allowing Google Cloud giving us the stage to deploy Dockerized applications. We put Django as a backend for our application services. We put like, you know, a common front-end stack, React, Next.js.
00:15:14
Speaker
to build really atomic components that helps us also in the future saves us time for not pre-designing and rebuilding things from scratch a lot. And kind of designing that entire backend in such a way that the code base is not everywhere.
00:15:32
Speaker
Because if it's everywhere as a startup, most of the time you're spending is managing the code base, cleaning the code base. So how could we employ existing industrial expertise, industrial frameworks that has worked? For example, Django, can we just ask developers just to follow best practices with Django? And then you have.
00:15:50
Speaker
your whole new world of data ingestion, consumed data, various different types, and from really bad looking Excel spreadsheets to quite neat JSON feeds from APIs. And they come in our systems, various different times, probably some pipelines run every day, some run like, you know, weekly, some run monthly. And how do you get that data consistently into the platform?
00:16:15
Speaker
Because the shop front is one thing, which is what users are going to look at, but behind the scenes, all those pipelines. When it's a couple of developers, it's easy. You can just deploy a couple of containers in your cloud and let it do the job. Lift and shift, bring the data, do some processing, put it out there, and then some SQL on it. It's great.
00:16:37
Speaker
As you know that this platform is going to grow, your platform's going to need more and more users coming into it, which means you're going to have to put more team around it. So that's where you really start to think you need to put some practices around it where you're not really micromanaging the team.
00:16:52
Speaker
So allowing team to do full stack development naturally without you micromanaging them. So that's where we six or seven months into the development of the product, we start to look into solutions like DBT for data modeling, Airflow for general ETL and ETL pipeline orchestration. And then we also realized that as we hiring a lot of people, it sounds familiar to them.
00:17:19
Speaker
because if we hiring a developer and then the tech team, they've got their own motives. They want something back from a job as well. So as a culture, as a shared culture at Carbon is, you know, when as one, we try to sort of, you know, give tools to all the developers, every single team, the graphene to empower them to do things on their own, that sort of, you know, and then work together with other people without building those silos. And these frameworks have actually allowed us to
00:17:48
Speaker
do that cross collaboration across different teams and different people at different skill sets as well. They could be a graduate in our team and they could be someone with like 10 to 12 years of experience in the industry. How do we sort of just bridge the gap and make their work on the same platform without any limitations or, sorry, this is too advanced for me. I can't deploy an Airflow pipeline. Can you do it for me? But just making it seamless even for a graduate to work on that.
00:18:15
Speaker
I think that sort of touched upon various different pieces there, but just to come back to the scalability point there. Because it's not often one thing, it's not about how many resources you allow to your one application that, hey, this has got
00:18:30
Speaker
8 gigabytes of memory or 64 gigabytes of memory, and it's going to run. It's not always compute is cheap. It's not always about compute. It's about the operations. It's about the culture. And I think all of the things I said that collectively helped us achieve the situation where we are now and have an evolving situation. It's never going to be done. But that's helped us answer that scalability.
00:18:57
Speaker
Brilliant question. Brilliant. I love there's a few points in there which really stood out at the bit of the beginning about, you know, focusing on working with the business and that business problem at the source. It's such a key point which I think sometimes the data is forgotten and the business and the data team are too separate and you're really working with them to help them understand what their problems are and how you can alleviate their problems with data and then
00:19:27
Speaker
I think, as you said, the key points around scalability is the enablement of technology and the frameworks that you put in place, which can allow for anyone to be able to maintain and build on the systems that are already there, and then frameworks can govern and make sure everyone's at the same level and in the same pace.
00:19:51
Speaker
That's brilliant. I mean, you guys haven't been around long. As you've said, you're a startup and delivering data products at pace is not easy. Every data team, every business wants to, once they've made the decision, we're going to build this once to deliver it as quickly as possible. So you've already alluded to some of the technology that you guys have used, but maybe a quick overview of that technology and what processes
00:20:20
Speaker
that you've put in place and others can learn from to be able to achieve the delivery of some of the data products like you've delivered at the speed you have.
Scalability and Learning with Python
00:20:29
Speaker
You're great to understand that. Yes, I think Nav touched on quite a lot of the tech that we use, but really the key bit that underlies that is that it's primarily a Python stack, and this, as Nav said, allows anyone in the team to basically get involved with another area, whether that is the back end of the application or Airflow,
00:20:48
Speaker
or even dbt you write python in as well allows anyone to get involved in any area and specifically within my team will hire a lot of graduates in unless you're doing a computer science degree the most common language across any sort of stem degree is python and so it makes it very easy with the graduates we get in for them to become comfortable with
00:21:07
Speaker
basically the entire tech stack within quite a short period of time. You already know the core fundamentals behind it, and it allows us to have a go at an airflow job, for example. There's no pressure behind it. We're trying to empower everyone to do a bit of everything so that if someone's away, anyone can jump onto it. And we're also trying to empower people to do their own projects as well.
00:21:31
Speaker
sort of anything that they might find interesting, we're giving time to just go and focus on their own projects. The fact that the whole tech stack is written in this sort of core language, in this common language, makes it sort of very easy for anyone to do that.
00:21:45
Speaker
I think before we talked about the most of the tech stack with like Python, DBT, Airflow, Google Cloud. Another great thing about Google Cloud is just it's got a lot of documentation for everything that's got in sort of a variety of different languages as well. Some of the other cloud providers will put a sort of a new product up with little to no documentation on it, which is great if you already sort of have an understanding of it from another platform, but if it's an entirely new product to you, you could be less sort of
00:22:11
Speaker
floundering on and stack overflow and stuff like that for some rogue answer that someone's put up five years ago about it. Or in fact, tragedy these days. Just make a search where it's like, sorry, I don't know, because documentation is not... That documentation part actually helped a lot of our graduates to pick things to really, really fast pace.
00:22:31
Speaker
And also to touch upon what people, any question, should actually, when deciding which stack and solutions to pick, what we've tried to do always is, yeah, we've got a stack which is very people-focused, which is very human-focused.
00:22:51
Speaker
but also how effectively you could deliver your value proposition. So you've got to prove that, in our case, you've got to prove that can you build a product, a data product, that could give a single source of truth across various intersections of the business? How effectively you could prove that?
00:23:11
Speaker
cheaply and quickly you could prove that and then how you could design a piece of tech like for example we picked Python obviously we realized that it's the speed of learning and the support around it they'll help us deliver that value proposition faster rather than sticking on like hey I'm a big fan of Rust I'm gonna go with Rust because it can it is faster it is efficient it can do XYZ yeah great but
00:23:42
Speaker
I'm not going to be the only person that's going to be coding here. I need to look at how this platform is going to be grown. I need to look at the value proposition side of it, how this tech is actually going to be helping business to deliver the value proposition business.
00:23:56
Speaker
doesn't care about whether it's rust, pie, ton of job. It cares about, is my number looking right that I'm looking at? Or is my data looking right? Or when I click at this button, is it taking me to the right place? And at that point, you'd already sort of look at. So it's some conscious choices, the very beginning.
00:24:17
Speaker
Everything is fine if you just put some healthy boundaries around it and then take it from there. Evolve, don't settle on one thing, but be open to new changes and then keep evolving and trading.
00:24:30
Speaker
Also, the best way to create a silo is to have that situation before I come in, I want to do this in Rust. And it's like, well, no one else. If you want to leave, then we're pretty screwed at that point. So you've got to spend a lot of time learning Rust or just rewriting from scratch. So very much the best way to create a silo is having an individual team. And once you've created that culture, you will naturally see people, let's say you hit a problem where
00:24:57
Speaker
Python's got a latency. Python adds up a lot of latency. No, obviously, certainly we don't have those problems in our industry. But let's say you hit a problem where Python's not being the right solution to serve that problem. But if you created that culture, that mindset around the team, around the tech team like that, they will naturally be open to finding new solutions and people will come back. Then they are not forced. Then they are coming internally from the team.
00:25:25
Speaker
That makes perfect sense. And I think it's the key points that you made, obviously, is the molding your technology, your stack, and your processes to your people and to your culture.
Aligning Tech with Organizational Culture
00:25:36
Speaker
That's such an important thing. It's great that there might be this tool which can do X, but if you haven't got the people to use it and the knowledge to be able to exploit it, then what's the point, right?
00:25:50
Speaker
And also, one point here from Fegenau a few days earlier is, why to code? The big question. When in the team you've got all the solutions, you've got to deliver certain things, you've got to achieve certain objectives as a tech team. Yes, you can write code from scratch to do that. The more you write, the more you have to maintain, the more you have to govern.
00:26:12
Speaker
So the question around, is there any off-the-shelf product? Is there anything open source out there which we can use? And there must be something. There must be something to do that if we look out. And can we repurpose it? For example, A of law, we want to write a pipeline to bring data from source A to B.
00:26:32
Speaker
Is there a pre-built community DAG that we can use to do that? So the big question around why to code, I think is really important every time if any developers say that, really want to code this. Okay, if you really want to do this, let's see if it's not already available there in our ecosystem, dbt module or at the Airflow module or any Django plugin. If it's not there, then let's code from scratch.
00:26:56
Speaker
I think that's such a great point and I think as organisations scale and we see more of these data meshes and siloed teams across an organisation, that's one of the big problems. You have people recreating the same code and I think sometimes it's right, can we find this elsewhere because we're making more work for ourselves and time is valuable.
00:27:18
Speaker
at the end of the day and that's a really nice sort of segue to the next section. We've spoken quite a bit about technology but I know we've previously spoke about this sort of blueprint and this copy and paste that you've been able to enact here at Carbon.
00:27:34
Speaker
Can you elaborate more on the structure of the team and how this blueprint of the team has been able to empower your work and your ability to deliver a data product? Yeah, so we've got three pillars of our tech team. We've got the data operations team,
00:27:52
Speaker
They're largely focused on data ingestion and data quality from the various different sources that we get in. We've got the engineering team, which now heads up, which is focused on the infrastructure for the platform, for the web app we've got, for all of the Google Cloud infrastructure, that sort of thing.
00:28:08
Speaker
And then we've got the analytics team, which I lead. And the analytics team is very much a sort of cross-functional role. If you were to buy a sort of a project slash product manager, analytics engineer, data engineer, data analyst, a bit of everything.
00:28:27
Speaker
And we're sitting amongst the business to try and ascertain what their needs are. When we start any project, we will have a business sponsor. I think that is absolutely crucial. We're very lucky in the fact that our tool is mostly internal
The Role of Business Sponsors in Tech Adoption
00:28:43
Speaker
at the moment. We give it out to customers. We give it out to people who we're affiliated with, but we're not selling it as a SaaS application. So to have an internal business sponsor is extremely important. You need someone that will
00:28:56
Speaker
stand up and basically shout out your corner, but when it comes to why have you done this, what's the purpose of this. Also, it proves there's a want for it within the business. I think when we started off, one of the problems that we had was that we wanted to make basically just cool stuff.
00:29:12
Speaker
And we made some sort of cool products, but they weren't really adopted by the general team by the underwriting side of the business because while they look cool and they look cool in demos, they just, no one wanted to use them essentially. And what we really should have started off with was a business sponsor for that in order to.
00:29:31
Speaker
in order to just make sure it's a useful product to build. And there was maybe some time wasted to start with that, but now we've got this sort of, we'll have a business sponsor, we'll have a member of the analytics team sort of working as this sort of like subject matter expert with the underwriting slash also have a knowledge of the text size. We'll have a member of the engineering team on every project as well, and when necessary, a sort of data operations member as well. So we try and get, it's a little bit like this,
00:29:56
Speaker
Like we, there's all these different models like this hub and spoke, there's the tribes and squads where you have like the sort of cross diet, the cross lines like that. It's a little bit sort of, you need to sort of decide what's best for you. I wouldn't say sticking to a strict like bottle like that. We have, we've had a look at implementing all of them, but really with the size of our team, it doesn't really make sense. Like there's only sort of four people, four or five people in each team. It doesn't really make sense to sort of just stick each one of those people onto sort of one project and have them work on that.
00:30:23
Speaker
And also like imposing it on the business side of things as well because underwriters are not really sort of, you know, they've never been working embedded with tech teams. They've never been around that country and all of a sudden when you impose that, hey, this is the new way the product development is going to happen.
00:30:41
Speaker
that might be sort of a restriction to their day-to-day how they do it. So we sort of try to be as flexible with various different approaches and see what we could sort of change and learn and adopt as we go with the active feedback from underwriters as well. Ultimately work was best for them because if the change comes from them first and then we apply, then we know that it's actually making them more efficient in doing their job. That's the ultimate goal.
00:31:11
Speaker
I think you made some great points there. I love the one about cutting that business stakeholder, shouting your corner, that advocate. I think it's something that we're still missing within the data industry, more advocates within the business or the data team recruiting them, then people to shout about what you're doing. Because as data people, you still need to be salespeople. You still need to get people to use the products that you're building. And I think that reverts nicely back to
00:31:37
Speaker
how you set out when you said you started this data product. It was all about working with the business to understand problems. And it's about taking them on that journey so that when you deliver that tool at the end, it's not a surprise to them. They knew this was coming and they were excited about it. And that enablement, I think, is something that is still often missed within the industry. So I think that's definitely one of the key takeaway points I took from
00:32:04
Speaker
from that. So what are the other biggest lessons that you guys have made at Carbon and other people that are on journeys of building data products for their business?
Building, Failing, and Pivoting in Tech
00:32:18
Speaker
What advice would you give to them?
00:32:21
Speaker
I think I will always say this, but there's this I think there's two main things. The first one is so at least when you're in your sort of startup era, I guess is that build fast, fail fast, as in just build everything, make a quick piracy of it, show it to the business sort of like I always talk about this sort of like
00:32:39
Speaker
just get it out there and just get it out there, show it and then think about productionizing it afterwards. If you spend three months getting some really pretty production level code and it all looks fancy and then someone goes, I don't care about this, then you've wasted so much time with that. Also the other one is just
00:32:56
Speaker
ready to pivot at any point with us specifically we thought that the analytics were going to be all about this sort of like underwriting performance which it is to a degree but what it's what we've really found to be useful with is the efficiency within the business
00:33:12
Speaker
So we've been able to scale very quickly in terms of revenue or premium without that much of an increase in headcount. Normally, a company of our size, in terms of the premiums that we're doing, would be double or triple the size in terms of headcount. We're still in the mid-40s terms of headcount.
00:33:29
Speaker
is because we just don't need as many team members to do the same tasks. Other insurance companies will have a whole team to do reinsurance, sort of manage the reinsurance relationships and statements and all that. We have one person within the tech team who just sort of hits the button, prints out statements every month because we've got this sort of common data model in place.
00:33:47
Speaker
Again, I think that's the sort of main thing that we found in terms of pivoting for us. And I'll go back to that big question about why to always keep asking that question. You know, I've spoken with quite a few startups around the market and it's around, yeah.
00:34:04
Speaker
We're not building rockets. Repeating myself. We're not building rockets here in the insurance tech market. Very simple solutions. Just ask yourself a question repetitively. I'm building this the new thing. Do I really need to go? Can I just go out and see other markets? What's already there?
00:34:19
Speaker
And then, you know, spend a bit of time on research, be mindful about what you're actually picking and then come back and then try and apply that solution. Get it out in the hands of the customer as soon as you can and as simple as format as you can to prove that there is value in that solution. And if you can't prove that value as quickly, as simply, then probably you need to rethink or move on to something else.
00:34:48
Speaker
I like that. Simple is sexy. People like stuff that is easy to understand. It's simple. And also it's easy for you guys to maintain in the long run as well. So guys, what's next for the team at Carbon Underwriting?
Private Equity Funding and Future Plans
00:35:03
Speaker
What's next for Graphene, your product?
00:35:06
Speaker
Yeah, so we've recently just had a very exciting amount of PE funding, which would just enable us to expand both the underwriting side, so as an insurance entity and also the tech side. We're in an exciting period of planning at the moment for what we can do with this extra money that we've been given. So there's lots and lots of areas for tech. I won't say anything quite yet, but lots of exciting areas for us to expand into tech-wise and also from an insurance product,
00:35:35
Speaker
And I think that this question is often asked like, you know, now, yeah, graphene's there, like, what, what do you want to do next with this tool? But as we all just before you guys arrived, we were chatting about this point is we should never like, you know, you should never stop.
00:35:51
Speaker
perfecting what you've already got. I think regardless the new solutions, the new ideas that will be coming around graphene and more tech is one thing that should never stop and will never stop as we grow is how we keep perfecting what we've already got, how we keep perfecting the modeling approaches that we've got, how we keep reiterating rethinking
00:36:13
Speaker
better ways of doing that because it's not done. It's never done. And once you settle, then that's it. That's where you start the legacy and then things become legacy. So I think the most important thing is like never stop perfecting what we've already got the data models we've already got. And then
00:36:33
Speaker
the new world, the new exciting stuff, everyone's talking about the new techs. They keep coming and once you've got this sort of culture that we've talked about where everyone can do everything, people are empowered to do tech and they're going to keep coming up with new ideas because they are naturally empowered to do that with the tools that they have.
00:36:53
Speaker
I think that's great. Don't move too far away from your core value driver and what you set out to do in the first place. It sounds like really exciting time ahead, fresh injection of cache to play with and new ideas to explore. It sounds like a great time to be at or to be joining carbon underwriting.
00:37:15
Speaker
Guys, that brings us almost to the end of the show. It's been great speaking, but before I let you go, we have a quick fire round of questions. We ask all the guests to help data professionals on their career and with tips from you guys in the industry. So the first question, how do you assess the job opportunity and what makes a good job opportunity to you?
Company Culture in Job Selection
00:37:37
Speaker
I mean, for me, personally, the culture of the firm is very important. They can have the fanciest tech, they can have the sort of like the best product, the greatest like vision and roadmap. But if you go into a company and the culture is not there for sort of just like, again, to empower people to
00:37:54
Speaker
their own learning and development and also just to have a good time like you spend five out of seven days of your week at your job you should enjoy going into work there are a lot of my friends who work in various other industries don't enjoy their jobs i can sort of pretty thoroughly say that i enjoy my job especially with the culture that we've got here at carbon
00:38:12
Speaker
That's great. I think it's so important, as you said, you spend so much of your life, you might as well enjoy what you're doing and that environment also should challenge you and should inspire you to get better. So, great point. What's your best advice for people interviewing? Just be honest. Always do an exciting project on the side.
00:38:32
Speaker
that just shows your drive, I think, to me when I interview people is, yeah, your job's great, but what's your vision? What are you doing now? What's your passion? What's your passion? Yeah, I think the honesty is such a good point. It links in nicely with the culture piece, doesn't it? If you're honest and true to yourself, I think that shows through with the interviewer. And I suppose the final one for both of you, what would you recommend if you could recommend one resource to help the professionals upskill?
Learning Resources for Tech Enthusiasts
00:39:00
Speaker
Yeah, I mean, now I've said W3 Schools, absolutely classic of I'm sure every single tech professional has been on there at some point in their career. Personally, just a shout out to a YouTube channel called Tech with Tim. Basically every time that I wanted to go and find out a new sort of framework or doing my own personal projects, he has everything there. He has the best way of explaining things. He's just super knowledgeable and relatable. So I would recommend him as a good resource.
00:39:27
Speaker
I think just pick a project. I often say, like, say I want to build the next Facebook. Just go build it. You'll learn a lot. And even though you're not really competing the right way, just learn a lot. Pick a project and say, I'm going to do that and then explore, figure it out. Learn by doing. And there are these LLMs to answer to make life a lot easier these days. So use that to your power, often save time and learning as well.
00:39:54
Speaker
Brilliant. I like that. And yeah, I think YouTube's always a great resource as well. And yeah, shout out to Tech with Tim. Tech with Tim. Yeah. Brilliant. Well, look, guys, it's been an absolute pleasure to hear about Club and Underwriting, the work that you're doing with Raffine and how you're transforming the industry. It's been a pleasure to have you both on the show. And thanks for joining me. Cheers. Thank you for having us. Nice. And we'll see everyone else next week. Bye.
00:40:24
Speaker
Well, that's it for this week. Thank you so, so much for tuning in. I really hope you've learned something. I know I have. The Stack podcast aims to share real journeys and lessons that empower you and the entire community. Together, we aim to unlock new perspectives and overcome challenges in the ever evolving landscape of modern data.
00:40:45
Speaker
Today's episode was brought to you by Cognify, the recruitment partner for modern data teams. If you've enjoyed today's episode, hit that follow button to stay updated with our latest releases. More importantly, if you believe this episode could benefit someone you know, please share it with them. We're always on the lookout for new guests who have inspiring stories and valuable lessons to share with our community.
00:41:07
Speaker
If you or someone you know fits that pill, please don't hesitate to reach out. I've been Harry Gollop from Cognify, your host and guide on this data-driven journey. Until next time, over and out.