Transcript
Speaker: Welcome to HSBC Global Viewpoint, the podcast series that brings together business leaders and industry experts to explore the latest global insights, trends, and opportunities.
Speaker: Make sure you're subscribed to stay up to date with new episodes.
Speaker: Thanks for listening.
Speaker: And now onto today's show.
Speaker: Welcome to the latest in our Digitalks podcast series.
Speaker: We will be featuring a variety of different topics and market developments that are currently trending globally.
Speaker: Managing data is paramount for today's insurers.
Speaker: We will focus on how insurers can better manage their data by looking at how data can be used and stored more effectively.
Speaker: Joining me to discuss this further is Sam Ray, Director in the Insurance Product Team at HSBC's Markets and Security Services, and Paul Clark, Head of Digital Data and Innovation, Europe and the Americas.
Speaker: Sam, over to you.
Speaker: Thanks, Gabriella.
Speaker: Paul, to kick us off, can you describe some of the different architectures that data can be stored in and why firms would look to use these?
Speaker: Thanks, Sam.
Speaker: Yes, of course.
Speaker: So, look, there's a number of different architectural approaches to storing and managing data.
Speaker: And actually, they all have different pros and cons.
Speaker: At the more traditional end, there are data warehouses where data is defined and categorized or structured up front at the point of load.
Speaker: This means the data in there is easy to query, has a structure to it which is helpful to navigate.
Speaker: It usually has an overall data model which describes how all the data fits together.
Speaker: On the flip side, warehouses are not so good at handling unstructured data like images,
Speaker: Because all the modeling work happens up front, they can be very slow to implement.
Speaker: Typically, you need a central data team to handle the structure of the data, and that causes handoffs and bottlenecks, which can be inefficient.
Speaker: Then a few years ago, data lakes started to become very popular.
Speaker: Data lakes operate differently to data warehouses in that you don't worry about structuring the data when you load it.
Speaker: Instead, you define the structure at the point you need to extract that data, and you do that on a kind of use case by use case basis.
Speaker: And that means you can make great progress really quickly with loading data, but it can also be very easy to fall into a trap of having poor governance or control over the data that's in there.
Speaker: Suddenly, you're 18 months in, you start to realize you're not sure exactly what data is in there or who owns that data or whether it's up to date.
Speaker: and how it's being used.
Speaker: So when people talk about their data lake becoming a data swamp, this is what they're referring to.
Speaker: Then more recently, data fabrics and data meshes have emerged, and they've got a different approach too.
Speaker: So these are designed to democratize the data in a firm by giving producers and consumers of data a set of common tools to load, define, and use that data.
Speaker: In HSBC, in security services business, for example, we've built a data mesh where producers of data can load their data sets in and consumers can use those to build their solutions all without needing a central data team to own those processes.
Speaker: We think this self-service approach with common infrastructure is a good solution for a large distributed business like ours, but
Speaker: As I said at the start, there's no real right or wrong here.
Speaker: Thanks, Paul.
Speaker: Looking beyond the structure that the data is stored in, can you run us through some of the important characteristics that data needs to have in order to be effectively utilized?
Speaker: Sure.
Speaker: Well, it's actually a huge question.
Speaker: We're talking here about metadata or data about data.
Speaker: And it's a whole error in itself.
Speaker: But in the interest of time, I'll pick out just four
Speaker: important areas as an overview.
Speaker: The first is data lineage.
Speaker: This is basically having visibility and an audit trail of where data has come from, who owns it and how it's changed over time.
Speaker: And this is critical, of course, because if you want to use data for anything, you have to know where it came from.
Speaker: You have to know whether the owner of that data will allow you to use it and you have to know whether that data changed and is up to date.
Speaker: The second is data discoverability.
Speaker: We talked before about data lakes becoming data swamps.
Speaker: One of the key causes of that is a lack of data discoverability.
Speaker: What we mean by that is an ability to make only certain approved datasets, if you like, visible for users of the data platform.
Speaker: So search tools will direct users to discoverable datasets that are permissioned to access.
Speaker: The third area I'll touch on is data observability.
Speaker: data observability is simply being able to see exactly how a data set is being used by who, when, and also for what purpose.
Speaker: So for example, if you're an owner of some NAV data and that data has been made available in the data platform, then you'll need to know who is wanting to use your data and for what purpose.
Speaker: And this is quite a change actually from the way things used to be done.
Speaker: Before, data would be deposited in a
Speaker: sort of big central data warehouse and how it is then used was honestly rarely tracked.
Speaker: Nowadays, it's becoming mandatory.
Speaker: So we have to give data owners all the right tools to control who can use their data and for what purpose.
Speaker: The fourth one I'll touch on today is data cataloging, which is essentially the tagging of data with descriptions,
Speaker: classifications, categorizations, et cetera.
Speaker: Creating an inventory of all the items stored, allowing them to be efficiently searched and filtered by users of the data platform.
Speaker: The discoverability piece, of course, is closely supported by good data cataloging.
Speaker: So having all this governance around the metadata is increasingly important as datasets become more complex, more wide ranging and closer to real time.
Speaker: Ideally,
Speaker: You build this governance into the core design of your data platform upfront, whereby the tools you give the users of the platform will themselves ensure adherence to the governance.
Speaker: We call this governance through code, and it's a good ethos to embed into everything you do when building a data platform.
Speaker: Thanks, Paul.
Speaker: That's really clear.
Speaker: We've talked so far about the structure of the data itself.
Speaker: Can we turn now to hear more about the physical storage of data?
Speaker: and main considerations for insurers when thinking about that physical location of data.
Speaker: Okay, yeah, well, this is another huge area, but probably the main area to highlight here is the concept of data residency.
Speaker: And what we mean by that is where geographically data is stored.
Speaker: This used to be obvious, you had a server and you knew where that server was.
Speaker: In the world of cloud storage, that isn't always as clear, of course,
Speaker: When we're using cloud providers, underneath all of that is a real server located in a country somewhere, not up in the sky.
Speaker: And where that server actually is can be important.
Speaker: So an example, let's say all your data is stored in a cloud provider in the UK.
Speaker: If you needed to store data from China, for example, or Saudi, then this setup is not OK.
Speaker: Increasingly, we're seeing countries mandate restrictions whereby domestic data
Speaker: cannot be stored offshore.
Speaker: And we call these data residency restrictions.
Speaker: So in our example, you would need an infrastructure that stores your China data in China, your Saudi data in Saudi, and the rest of your data in the UK, plus some way of joining those data sets up for your client for things like consolidated reporting.
Speaker: The solution is much easier if a cloud provider can support all the countries, but if they can't,
Speaker: then you need a data platform capable of connecting to different stores across multiple providers which can join data held on cloud with data held locally in countries with those residency restrictions.
Speaker: Thanks, Paul.
Speaker: And just before we close, can you tell us a bit more about data platforms?
Speaker: What problems do these new solutions solve for and what issues still remain?
Speaker: Sure.
Speaker: Well, the first thing to say is that cloud is transforming how data is not only stored, but also how it can be distributed.
Speaker: The rise of cloud has seen some big new entrants acting as sort of market-wide data platform providers capable of hosting and distributing data much more easily than ever before, with some players creating essentially a supermarket where producers such as data vendors can be connected to
Speaker: consumers over a common data platform.
Speaker: The benefits of being part of that ecosystem are clear that data vendors are able to sell their data to a much larger potential client base, integrating just that once to the data supermarket platform and allowing that to be their distribution channel.
Speaker: Meanwhile, for consumers, they can connect once to the supermarket platform and buy multiple products all from the same place.
Speaker: But more widely than this, we'll increasingly see the ability for parties to share the same data, not copies of data, but the same data.
Speaker: And that will reduce duplication of data across the industry, help reduce reconciliations, for example.
Speaker: And more excitingly, also allow co-development of innovative solutions between organizations, you know, on top of shared data sets.
Speaker: But, and to your point about what
Speaker: issues will still remain.
Speaker: If we really want to fully benefit from the potential interoperability that this technology development offers, then organisations do need to also be willing to work together to move towards industry-wide standards and data models.
Speaker: There's no platform and indeed technology alone that will fix that age-old problem.
Speaker: That's about all we've got time for today.
Speaker: Thank you very much, Paul, for your insights.
Speaker: And I'll hand back now to Gabriella.
Speaker: Thank you very much.
Speaker: Thanks so much, Sam and Paul.
Speaker: It's been really interesting to learn more about how important data is and how it can be used more effectively.
Speaker: I would like to thank you for listening to this edition in our series of DigiTalks podcasts.
Speaker: Stay tuned as we explore more trends in the coming weeks.
Speaker: Thank you for joining us at HSBC Global Viewpoint.
Speaker: We hope you enjoyed the discussion.
Speaker: Make sure you're subscribed to stay up to date with new episodes.
