Blueprints for Increasing Data Maturity Across the Organization

[MUSIC PLAYING] All right, good day, everyone, from Quest Empower. I'm here today with Kevin Smith, the CEO of Tokio Marine, who is a user of our erwin Data Intelligence platform. And my name is Susan Laine, and I am the thought leader here for erwin, of all things data. And I'm happy to bring you a presentation that Kevin and I did from Gartner, the Gartner Symposium in London in May of 2023. So, very recently, people really enjoyed the session. I really enjoyed doing the session with Kevin. And it's called "The Blueprint for Increasing Data Maturity Across the Organization". So, Kevin, I'd love for you to just introduce yourself a little bit more with some background. And then I'll let you kick it off. Great thanks, Sue. Good to be back here on the virtual stage again. So, yes, so Kevin smith. I've been CDO at Tokio Marine HCC, which is a global insurance company since July of '22, I guess. So it's now been about 15 months. And it has absolutely whizzed by. I have a background in data for many, many years now, particularly in the insurance company, some of the large insurance brokers. But, as anyone that knows me, I have a very big passion for data. So very happy to be here again and looking forward to walking through this again. Thank you. OK, so the vision, one of the things that was very important when we started on this journey was to get a clear vision of what we were trying to do for data governance and data in general within Tokio Marine HCC. So the vision statement I've got up on the screen there, the key words we've underlined-- and this is really what we think about almost every day now with our data. We want to make sure that we focus on data that is critical. We have a huge amount of data in insurance companies. So focusing on the things that are really important or critical to our business, building the trust. Probably the key word that you'll hear from a lot of CDOs, particularly focused on the governance side. Trusting our data is so important. So any way that we can prove that trust to our consumers is really important. Making it accessible, we're going to learn a bit more about that later. How to reduce the friction in people. Our consumers being able to get to the data is really important. And, of course, maintaining the security, agility. And something that's really important for Tokio Marine is that innovative spirit. So doing it in a way that's actually really sort of quite exciting. And sometimes bending it around a bit to make it, to make it work. One of the things that we try and do all the time is to improve the literacy of our community. But anyone that's worked in insurance know that they're generally very highly data literate anyway. Lots of actuaries, lots of data scientists that are working on models. But that's really how we were going about it. And how are we going to measure that? Well, the return will come from making sure that those really key users become even more efficient in the use of the data. And that's a relatively easy thing to measure, which is quite unusual for some of these things. So one of the other things that I've always done when I've joined a new company is I like to explain things in data models. I have a background in data modeling. And one of the great things about the insurance industry is that there is a model for a standard, an industry standard model. The organization is called ACORD. A-C-O-R-D. Often misspelled, but it's just an acronym that I can never remember the name of, but it's ACORD. And, when you boil it down, you really end up with these main objects of data. So, when you're describing anything within insurance, you can find one of these buckets that this will comfortably fit into. So, once that's been explained to the executive committee, it then starts to filter down, and you start to hear those execs talking in these types of language. So one of the things that-- early days, everyone often refers to policies within insurance companies. But, actually, it's really a contract. And that's often just a language barrier. But, once everyone starts to understand, oh, absolutely, we are essentially a contract business. And then the other thing to get going with really quickly is a data governance policy. This is a piece of work that I got going with very early and being able to get that policy in place and get the controls in place really helps with everyone to understand what it is that their roles are, their responsibilities, and why it's important. And those are the four main controls that we've now put in place. So we have a data modeling team that sits under me. And, therefore, we have a whole control around that. So anything that's being designed, whether it's our new warehouse, or whether it's interfaces, APIs, then it goes through our data control process. We have data quality, obviously. And, again, we have a separate data governance team that owns the data quality underneath me. And they work very closely with the business. We have a suite of tools that we use for monitoring data quality, but it's really almost a process, a business process that we work with the business to help them identify problems in their area. We can write data quality rules. We publish them, and then they monitor them. And they take charge of effectively their own destiny of improving data quality. Data management. Another key area where reference data and master data are huge areas within insurance and probably the same across all industries. And being able to manage those reference data, again, it fits nicely into my world and the data. The data design team, for example, they have to design all of the APIs and the underlying data that goes into those data management reference sets. And, finally, I also own data strategy. And that's something we review every year. I'm now coming up to publish my first sort of major release of that in probably the first quarter of next year. I've done a data office strategy, which is sort of a lighter-weight version of that. But now the big data strategy for HCC is what we'll be publishing soon. Always a good question, what does actually good look like so you know where you're heading? And this is the way I like to think of data. If I know it's well-defined, and I know it's got some ownership, I know where it comes from, I know what quality it is, and I know how I can get hold of it, that, to me, is a really good piece of data that if I was sitting here with a consumer hat on, those are the things that I'd like to know. So if we can actually produce that, then that's where trust really gets built. And one of the really important things I always like to iterate on these is that it's of known quality. So sometimes you just don't want surprises. You want to know going into it that it's not that particularly-- it's not particularly great. But you don't necessarily have to have the highest level of quality to get the job done. You just need to know what it is that you're working with. Now, one of the other things that we've now started to do-- and my whole world seems to have revolved around this. And some of this is something that me and Sue have been talking a lot about since Gartner, probably post-Gartner in the world of data products. And, yeah, I struggle to think of any other way of actually thinking about things now. But products, we products in our everyday lives-- maybe you open the fridge, and you'll find products that you want to consume. But data products is similar in that sort of style in that they're there just to meet what-- meet the need of the customer. Now, it could take many different flavors. It could be a data set. Or it could be a dashboard. Or it could be a report. Or it could be a whole warehouse. You can have a-- you can describe products in any way you like. But they're there really for the consumers. And I also sometimes like to think of them as a bit like icebergs. Sometimes, for a user, they want just a bit at the very top. But good products are all made up from other products as well. So a lot of work goes on in building a fantastic dashboard. But what you don't see is all the products that have been built underneath the water with those data sets. So there's usually sort of four, four boxes that make those up-- the inputs, the technology, the life cycle, and the people. And, sometimes, people focus on-- too much on the technology. The people bit is a really key-- is really key. Those subject matter experts, those consumers, the teams that are going to support it because the other real key factor about data products is they're effectively alive. They're going to always evolve. You don't want a product that's just going to be stale after a year. You want to have a product owner that's passionate about that. And that's part of my role now is really to work and find those other passionate people in our community that want to own these and want to see it evolve. And I help build this whole ecosystem that allows that to happen. And putting in the teams that can help continue to evolve these products. Yeah, I like to say that it makes it a lot more tangible, especially to the business if they can come in and request a product and not have to know exactly what asset it is, but they can request a product. And then we can deliver what type of data, what type of AI model, what type of files, reports, et cetera for their product, for their mission that they're trying to accomplish. It just makes everything so much more tangible and hands on. Yeah, absolutely. And we kicked this off with an all-day workshop outside of the office, had an away day organized. And we got the right people in the room. And we educated them on what data products were. And it was a great day. And, from that point onwards. Probably about three months ago, four months ago now, everyone talks in that language now. And it's made life a lot easier. Absolutely. So you can think of three types of products. Sometimes you want to create a product that's just aligned to a source system, so source-aligned products. Your data scientists are typically going to want to get hold of just the raw source data. So we can create products that we've just landed straight. And they're very quick for us to do, put it in the lake and give access to the data sciences. And, suddenly, they're working on those things. And then, as you work your way up with different consumers, you want some that maybe have got those joined these-- going back to those different domains of data I outlined, you want to join your claim data with your policy data with your finance data. Those are the enterprise-type products. They're the harder ones to build, obviously. But it's a lot easier to build them when you've been through the whole data modeling. Exercise, you've got your glossaries defined. So everyone knows what's what. And you're building products upon products, really. So that's what those are. And then, finally, you have your analytical data products, which, for many people, those are the areas that they consume from. It's a dashboard. It's an interface. But, as I said, there's actually other products below the waterline that other people are consuming. But it's the analytical data process that often gets all the glory. But, actually, there's a lot of work built on it underneath. How am I organized? Well, we call ourselves a data office. We sit in-- actually, we sit in the data analytics and innovation space now. And so, yes, we're starting to look at all the innovation. We're looking at the AI as everyone is. But how we organized directly is I have a data management and data governance pillar. And I have a have a head for there. And then my data modeling and design and data analytics all sit under me. So we all operate in an Agile-- in an Agile way. Our ways of working allow us to interact a lot between those pillars. And it works very well. Key topic always is around ownership of data. And I'm very, very pleased with the way this has gone. I've been in organizations where this can be quite a challenge. Now, fortunately, for me, before I joined-- and they created the CDO role here. They created a data executive committee. And I present to them every quarter now. Prior to that, we were doing-- we were meeting every month. And so we have a really engaged and highly switched-on data executive team that have just been-- have just made my life a much easier to get some key messages across. And start to build the platform and the team that we need to move this forward. So this is how we think about it, strategic level, operational level processing and then support [INAUDIBLE] again, we like pyramids in Tokio Marine. Now, one of the other key aspects of data ownership is-- sometimes for an individual owner, they need to have an executive oversight of it as well because it's very common that we have multiple lines-- obviously, we have multiple lines of business and individual data owners within those lines of business that might need to operate in a slightly different way. But, actually, what we're trying to do at a data level is we want to be consistent. And, to do that, we can use the combination of our data executive committee, along with the ACORD data model at that top level. And we can-- and we've done this. We've allocated our executive to those high-level domains. So they have an interest overseeing lots of lines of business, lots of ownership within those-- in that same area. And that's-- that works really well. And it's-- I've had lots of people talk to me about this particular slide because it's like, never thought of doing it that way. But, yeah, it's one of the only ways I think you can really tackle it. And how do we govern ourselves? So we report up to the risk and capital committee. And we have that data executive committee I mentioned sitting up there, a strategic level. And now we have our data governance council. So That's been sitting now for a little while. And this is where I'd say where the real work-- where all the decisions are really made. The data governance council is-- was a little bit of a challenge to form because what I wanted to do was make sure that we had, on the data governance council, people that were genuinely passionate about data and wanting to improve it and increase that trust. So we've got a real mix of seniority and folks that are really down in the weeds of some of this data. But what they can do is they can understand their particular domain of data. And they can take that upwards or sideways or downwards. They can communicate it effectively. So we've had-- I think we've had two full sessions now. And we're already getting into the meat of reference data, which is a great place to start. And that is supported by a number of data working-- data governance working groups. So, for example, we have a dedicated one tackling FX rates at the moment because that's a big topic for us to work out. And we align that alongside architecture because, quite often, through that working group, there's something that comes out that needs architectural input. And I missed that the data Governance council always also bounces off the compliance team and the business infosec. So working well. Nice, clean structure. And I say very well-initiated before I joined with the data executive committee. And where do we feel we sit in our maturity? We feel like we're doing OK. There's lots of work to do. We're in that industrialization stage, I would say, now. The aspiration we've done-- I'd say I've been here about 15 months. A lot of that work was initially done in setting up that data governance team. And we're going through that. We're definitely going through that maturity and definitely touching on the industrialization of it. And then it will just turn into, how do we really scale up and ramp up? And that's what we look forward to as the next phase. Another way of looking at our maturity is to think about the tooling that we use. And you'll see that we have a lot of erwin on this page. We've been working with erwin for several years, a little bit of time before I joined as well. And I've been doing data modeling for a long time. And I've used erwin for many, many, many years. And so, when I learned that one of the tools that we had in our suite was also this new product called erwin Data Intelligence, and I was very excited about that because I got to know it because I'd never used it prior to here. Then, yeah, I immediately loved it. And I guess one of the ways of describing it, as I said, I just get it. I think working with Sue, and Yetkin, and the team. We seem to be able to communicate very freely about how the product works and what we're trying to do. And it's just allowed my team as well to roll this out very, very easily. So we use a product called DQPro at what I call the data quality management monitoring type level. It's a specialist tool for insurance. And it's very business friendly in terms of its-- allows the business to monitor themselves in, in the quality. And then if I go down the stack, ownership, glossary, lineage, modeling all sit in erwin. And, also, a different sort of type of quality, what we refer to it as the discovery and profiling sits in erwin as well. So one of the things I've been doing for quite the last three or four months now is our team are building a new data warehouse. And, as we're adding new sources, one of the first things we do is we profile it. We profile the source. And it's been absolutely fascinating the way we go about this. And we can just quickly present that back to what I now call [INAUDIBLE] and say, did you realize these are your inner and outer limits of data? And why have you-- it's just a fascinating way of suddenly getting a conversation going. And what often comes out on the back of that is we then write data quality rules that they sit into. We push up into DQPro. And it's completed that loop. But we've been able to do that really rapidly and with a lot of easy insight and overview of it. So we use a bit of-- internally, data management is the piece that I'm sort of working on next as to whether we need a tool, whether we have all the right tools internally and build it ourselves. But we use Microsoft, mainly Microsoft services at the moment on the data management and data management side. But that covers most of our maturity. And then I think this is probably where I'm going to hand over to you, Sue, to explain a bit more about the whole product. Yes, thank you so much, Kevin. Hey, not to put you on the spot, but tell us a little bit about-- you talked about where you are today and where you want to go. What are you most excited for in 2024? What are you looking forward to with the next stage of your implementation? I think it's that genuine scaling up now. Analytics have just joined my team, literally this week. And I know you're going to-- you're going to talk about marketplace. And you know how excited I am about marketplace. But, for me, it's going to be about scaling up, getting these data products into erwin marketplace and getting our user base, going there to actually find the products that they want. I want to be able to see that it's getting easier for everyone to find their data, to access their data, and then do something with that data. And all the building blocks are there. And it's now a case of scaling up. So if we do this again in a year's time, you can ask me the same question. And, hopefully, I can say, ah, 2024 was good or in '23 and '24. But, yeah, that's what it's going to be about, really scaling up and getting those genuine products more trusted, more visible, more access. And I think you're going to help us with that. Yeah, that's fantastic. Thank you. What I love about your implementation is that you're always looking for more and more ways to impact the power of data and getting it more accessible to the end users and just bringing them into it and making it real. So great efforts. Love chatting with you and chatting with the team at Tokio Marine too. So let's talk a little bit about the underpinnings of that program as Kevin just spoke about. So, with the solution itself, a lot of our clients come to us because they've had data modeler for some time. So data modeler is where you're actually going out, and you're modeling that to be state for your next new environment. Then we have our catalog that supports all that great inventory of the information, the lineage of the information, and just understanding, what is the underpinnings of my entire organization? Data quality comes with the solution as well. So you can use the portfolio as a way to graphically look at the data and say, here's where I have all my critical data elements. Here's where I have my privacy information. And I see a lot of recursive information happening here. And I need to use the graphical representation as a way to go and profile my data and understand what's been profiled and what hasn't been. Where do I have good data? Where do I have bad data? And there's a lot of advantages of having data quality embedded because you can see it in the graphical portion of your solution anywhere that you are. And then, finally, data literacy zone. That's where we're giving all that great business information and curation of the data itself. And another thing that I really like about Kevin's approach-- and I've seen other clients do this too-- is that they were very policy driven in the very beginning. So they were looking at the policies inside of the organization so that they could get their arms around the data quickly and make sure that they know how to use this data and understand the rules behind it from a policy perspective. So that is a great approach and one that is definitely doable with a solution like this. So next slide, Kevin. There we go. So another aspect of the solution isn't just the solution but how to go about an implementation from a value perspective. And I think if you make another click, you'll see the seven steps themselves. There you go. So if you want help in understanding, how can I roll this out and bring-- increase the maturity and bring back really great business value use cases as I'm doing it? We can coach you through that with these seven steps. So, at each step, we're delivering business value, use cases, and creating that impact, from modeling all the way to marketplace. And, in the interest of time, I don't think I'm going to hit on every single one of these. I'm going to go to the next slide, Kevin. So it might seem like this is going to take a long time to get from your model all the way out to the marketplace. But we view it as very iterative. And Kevin talked a lot about products in the very beginning. And I would say that that is our most popular use case right now because it keeps the business in the driver's seat on requesting a product and then iteratively working from the model to the catalog, which produces the code for that data and then curating it, so putting those business policies around that product and then making it available inside of the marketplace so that you can shop and share and deliver it across the organization. And you don't have all your domains creating the same products and creating the same data sets across the organization. So it's a one-stop shop to share and collaborate on your data products. All right, and that's it. One-stop shop. You're not only putting it together, but you're making sure that you have governed. And you've put the guardrails around that data. And you can trust it. So there's a lot of marketplaces out there today, but are they truly governed marketplaces? Are these data sets something that you can freely release across the organization and use and trust that the data is what it is and that you're not breaking any rules or any privacy information as you're using it. Next slide. The marketplace was released in May of 2023. And it focused on these four aspects-- data sets, so both third party data sets as well as internal data sets; AI models, so building that transparency and a one-stop shop to say what users are using, what AI models, and making sure that you're creating a center of excellence, really, for those AI models, and you understand the intent of those models, et cetera. And then we also implemented or brought to light a new scoring mechanism for the data itself. And I'll talk about that in just a moment. But those were the four aspects. Next slide. From an AI governance perspective, today is November 3rd of 2023. And President Joe Biden just signed off on all of the regulatory AI guidelines. So how are you going to create that glass box for the organization? We have a way to help you with that. And if you hit enter, it gives you some more information about how we can help you solve for really understanding those AI models. What are the intent? What is the data that you're serving up through the back end of those AI models? And we go as far as monitoring the drift of those data sets behind the AI model and alerting to you when things have changed to make sure that it's always in track with the intent of that model as you mature and you're keeping the bias out of it while you're monitoring those AI models. Next slide. For those of you at Empower who have had other Quest tools, you might have something called Toad, a very popular data prep tool that goes out to 50 different source systems to deliver those data sets through that catalog. So the catalog's not only delivering metadata. We're also helping you deliver those trusted data sets from the databases themselves. And if you don't have a data prep tool or a way to capture that information, you can talk to us about Toad as an integrated solution with the marketplace. And, finally, that third-party data. Whether you're going to maybe a Snowflake data marketplace or other marketplaces to bring back that third-party data. We're helping you understand, what are all those entitlements around the third-party data? Allowing you to really understand also the processes and the policies around that third-party data before you're shipping it out for other domains to use. And I believe the last slide is next, which talks about that new scoring mechanisms. We love to talk about this because there's no other catalog that really has this today. But since we have data profiling within the solution itself, we're giving you a combination of the profiling results of the data, the end user rankings, so the five stars. How well-liked is this data set? And then coupling it with our governance. How well is this data set being governed? And then we classify for you gold, silver, and bronze data so that you always know how good that data set is. And I love what Kevin said about, it's not just understanding the final results of that profiling, that it's good data. But it's knowing if it's good or bad data or gold, silver, or bronze data. So that's how we alert you and classify your data for you on the back end. So, with that, I'm going to let you guys know. I want to thank Kevin. First of all, thank you so much for joining us again today. I really-- sorry. Your pleasure. Yeah. Thank you. Thank you. And we also have a demo online through Empower if you guys want to go see some of these products in action. So thank you all very much. Very nice to be here today. [MUSIC PLAYING]