TransUnion Live – The Top Trends in Data and Analytics in 2023

Watch our on-demand TransUnion Live episode about current data and analytics trends. Comparing forecasts from the March 2023 Gartner® report, “Top Trends in Data and Analytics, 2023,” to our own observations, the discussion yields important insights to help leaders factor these trends into strategies and investments that drive new growth, efficiency, resiliency and innovation. 

Full Transcript Below

Andrew Goss/Host:

Well, welcome everyone to TransUnion live on LinkedIn, Andrew Goss, and I'm excited to be joined by my esteemed colleague, Nathaniel Loman. Welcome Nathaniel.

Nathaniel Loman/VP of Data and Analytics, TransUnion:

Thank you.

Andrew Goss/Host:

Well, we're here today to explore the trends in data and analytics that we witnessed this year by comparing them to what Gartner forecasted in its March, 2023 report top trends in data and analytics 2023 appropriately named. Now. Before we get started, I have a quick housekeeping note that I always mention at the beginning of these lives. We welcome and absolutely encourage your questions as long as they're related to today's conversation. Just put 'em in the comments area and if they're applicable and we don't get to them today, we'll do our best to circle back and respond to you afterwards. So let's get started with the conversation. Nathaniel, first off, I always like to start these conversations kind of getting the lay of the land, singing from the same song. Data and analytics may mean a lot of things to a lot of people. Can you start off explaining what we truly mean by data and analytics for the purposes of this specific conversation?

Nathaniel Loman/VP of Data and Analytics, TransUnion:

Sure, Andrew. So let's start with the data piece. It means to me the systematic, I would say robust and abundant gathering of data, pulling all that together and then curating it, disseminating it appropriately, and of course the governance of that data. The analytics piece to me at a very high level means taking what is essentially a raw asset and turning it into knowledge or insights. Now, that could mean insights in business policy, academia. In our world, we take it a step further than insights. So we are in the business of helping our customers make predictions. So for us, analytics means leveraging modern and powerful computer systems essentially to convert data, including really incomprehensibly large data sets into predictive or descriptive outcomes including predictive models and strategies for underwriting, marketing collections, et cetera. So we sometimes hear raw data described as a commodity. I think that's a little bit underselling it more accurately. It's an asset, it's an asset to be managed appropriately and in some cases a differentiating asset. But I take the point and the point is that when we lay our analytics on top of data, we're able to provide significant value add on top of the asset and further differentiate our business.

Andrew Goss/Host:

Great. Cool. Thanks for setting the stage here. Now let's get to the Gartner report. Talking about value optimization, it really focused in on that specific area. What did the report find and what has TransUnion observed in that area this year?

Nathaniel Loman/VP of Data and Analytics, TransUnion:

Great question. So Gartner had a couple of things that they highlighted that I found really interesting. So they mentioned value optimization as measuring outcomes using metrics that link the technical outcome to a business outcome. Now, this I wouldn't say has emerged this year for us in a narrow sense. We've been doing that for decades, particularly the arch typical example would be a swap in swap out analysis resulting from improved scores or strategies. In those cases, you look at how a model or strategy produces better rank ordering of risk, and you can see the revenue impact as well as the risk mitigation impact and how that flows through to value for a lender or a business. So in some sense we've been doing that for a long time and are very comfortable with the idea of connecting technical outcomes to business outcomes. But I think if we take it up a level, it's about the entire value of data and analytics to an enterprise, much more so than just a narrow result like that.

And this is a question that we have been working on for a long time, maybe forever, how to show the value of data analytics to the enterprise. So in our case, you could look at revenue allocation for let's say a contract or a sale, which is hard. It's hard enough because we have lots of parties involved in making our business successful and analytics is an important one, but it's not the only one. So how to allocate that is an ongoing challenge. We've gotten pretty good at it using modern CRM tools, but it will remain an ongoing challenge. But I think the challenge is even bigger than that. Our impact isn't really just narrowly focused on the sale of data or signing new contracts. It goes to customer relationships, it goes to competitiveness of our business when we lead with analytics services. We're able to really differentiate TU from our competitors by being there for our customers, by making sure that they're seeing the value in data, in that data that's just an asset until we layer on analytics and make it a product, make it a solution.

So we're able to really facilitate business in ways that are not easy to quantify, but are nonetheless extremely important. So it's a big question. It's an important question and it's a difficult question, and I think if you were to try to, let's say, take a shortcut to understanding the value of analytics to a business, I think it's instructive to think about the counterfactual. So where would your organization be without data and analytics? Without data, obviously TTU doesn't have much of a business, but without analytics, we'd likewise be in a pretty bad spot. We'd have limited value add on top of our data assets, and we'd really struggle to differentiate ourselves. For other organizations, they're going to have their own assessment of the value of data and analytics, but just think about where you would be without it, and in most cases, you're going to be in a much worse spot.

Now, one other thing that Gartner highlighted that I thought was very relevant to my experience and to what we're seeing is that they emphasize that to focus investment on short-term ROI is going to overweight tactical investments relative to strategic or disruptive investments. This is something that I completely agree with, right? Tech enablement through significant strategic investments can allow for much more growth. It allows for experimentation, it allows for sort of a confident development of new solutions and disruptive strategies. It allows for r and d and it can also lead to significant internal efficiency and cost control sort of regimes by having a more complete platform. I think we'll come back to this as we talk about some of Gartner's other points about ible and composable ecosystems, but I think it is an important point here to emphasize.

Andrew Goss/Host:

Got it. So I may have buried the lead here a little bit just in terms of what everybody's talking about this year. So one of the hot topics this year across enterprises, consumers, et cetera, artificial intelligence, the report rightfully so, focuses extensively on ai. Let's talk briefly about that, how it addresses managing AI risk and emergent ai. What has TU seen in that area so far this year? Specifically what Gartner talks about? AI again, can be very broad, right?

Nathaniel Loman/VP of Data and Analytics, TransUnion:

Yeah. This is certainly a hot topic, so no surprise you see it addressed here. They're focused on governance as well taken. It is something that we take seriously and we have embedded risk management responsibilities in a formal and cross-functional committee to make sure that we're progressing on our AI journey in a responsible way. That being said, we view AI as an extension of machine learning capabilities that we have been cultivating for over a decade, and our intent is to exploit this in ways that produce value for us and our customers, while of course controlling those risks. So it's important to note that AI does bring a lot of opportunities, and it's not always going to be obvious where they're coming from, but there are opportunities. So I'll give you a couple examples amongst the risks that are cited. Fairness, privacy, human control. These are commonly cited risks of ai, but they're also all opportunities.

Let's focus on fairness. Fairness is something that's very ingrained in the consumer credit business. Just look at our governing regulation, FCRA, it's the Fair Credit Reporting Act. So we're very accustomed to fairness and to trying to build fair models. Ultimately, AI is going to introduce ways of building more fair models, and that's essentially already happened in many ways. So we have new capabilities that will lead to better models. So let me take this a little bit further and just sort of highlight how non-linear and unpredictable this can be. If you look at, for example, early generation chatbots that were released into the internet to public consumption, and I don't mean chat GPT three or four, I mean stuff from a decade ago. There were a couple of very public failures of those that they were released trolls and malignant people on the internet trained them with hate speech literally immediately, overnight.

And they got yanked very quickly after that happened. The tech companies, Google and others invested in ways to fix the problem. Some researchers at Google put out a paper on adversarial de-biasing, and the concept here is to build the predictive text generation AI model and then build a challenger model to it that challenges it to act in a desirable way and not in a undesirable way. That technique didn't stay there, it didn't stay with chatbots. It made its way out to our industry, and you can now build adversarial de-bias models for credit risk models that challenge the predictive model on the grounds of disparate impact or fairness. So the path from an AI advance to business impact in our industry or other industries, it's not going to be linear and it's probably not going to be easily predictable, but it's real. So in this case, we could now build models with less disparate impact. Who would've guessed 10 years ago as we're all laughing at a news report about a chatbot failure, that it's going to lead to us building better and more fair credit risk models 10 years later. It's not obvious, but it is important to realize that there are benefits to AI that come with the risks.

So generally we see this as an area that's full of opportunity. There are landmines. We are going to tread cautiously without losing sight of it, but we're devoting resources to it and trying to find that happy balance of risk and innovation.

Andrew Goss/Host:

Got it. Interesting. It's just to follow the path and evolution and relate it to what we do is really interesting. So another hot area, and I would argue this has been at the forefront of what everyone's been talking about enterprises has been data sharing. What does this mean when it comes to data and analytics? I mean data is in the word, but I think it's good to explain that and what does the report talk about and what has to observed in that area this year?

Nathaniel Loman/VP of Data and Analytics, TransUnion:

Yeah, so I mean obviously we're an information services company. Sharing data, I would say is fundamental, but it's probably more it is our business, so we make trust possible. That's our motto. So we do that by sharing data in a way that preserves privacy and is secure. So that being the case, we are very focused today on privacy enabling technology, and we see that as an area of focus in our industry generally. So what that means as we engage with other data providers, with other customers is we're seeing a lot of focus these days on federated learning structures, most frequently executed within clean rooms as a way to bring together various different data sets from different parties and preserve privacy and mitigate risk. So we see that as an area of growth, area of focus right now. We also see a lot of privacy preserving AI technologies coming to the fore, and these are some that we've used extensively internally, including the tokenization of identities to mitigate risk as you're sharing data with customers. We also see synthetic data, which is sort of its own form of generative AI in a way where you can create records or data sets that look like real consumers, but it in fact are not.

And we use that for a lot of our testing and applications where we don't really need real data. It's much nicer to use synthetic data and safer. So we recognize these technologies as having a lot of transformational potential for sharing data safely and effectively. And I would say it's an important topic, and Gartner is right to call it out because if we can master sort of safe, effective, well-managed data sharing approaches, this will be an important hedge for potentially more onerous or restrictive privacy and data sharing regimes that we'd all prefer to. So I think it's good for us to stay out ahead of this just as it's good for us to stay out ahead of fairness, risk and AI and things like that. So beyond, I guess the narrow sort of way I've talked about it here, I would say that getting to a fuller view of the customer is driving a lot of investment in a large sort of data fabric structure.

Something Gartner also talks about that enables more analysts, more users to touch the data to make use of the data. So I think within the enterprise it's important to understand how you can move away from sort of siloed data structures as well, getting to a place where the full enterprise can sort of capitalize on the complete data that you have in a responsible way, making sure the right people have access to the right data sets of course, but making that access available in the appropriate way is going to lead to more developments, more creative juices flowing across the enterprise. So we do expect as we move from more siloed structures, and again, this is something we'll talk about with converged and composable ecosystems in a minute, but this process of bringing together data either internally or across enterprises will lead to significant value add. It will lead to more use of analytics and it will lead to more product development.

Andrew Goss/Host:

Interesting. So one of the other large focus areas getting away from, I'd say the flashy stuff is what Gartner calls observability. What exactly does that mean and what is to you seen in that area?

Nathaniel Loman/VP of Data and Analytics, TransUnion:

Yeah, I mean fundamentally it means bringing transparency to the data and analytics process. If you were to describe what we'd do as a black box, let's remove that black box and let everybody sort of see how this works. Now narrowly, we're not new again to the idea of observability. Obviously credit risk models have long had an observability requirement built in right adverse action reason codes are mandated by the FCRA. So we have a lot of experience explaining how our models work to consumers, to customers, and to regulators. The broader data and analytics system, however, is much more challenging to explain, especially to our business leaders who really need to understand it. So I think where Gartner is headed with this, and I completely agree, is that we need to bring observability to the system level. So what I mean by that is, for example, all your data pipelines observable and monitored.

Do you have alerts set up on your data pipelines in case something looks wrong in the data? Are you able to catch that and fix that before it becomes an issue? Are your data access and usage patterns observable? Are your costs and data processing and data storage costs observable? All of these factors are really important as you're trying to efficiently manage the business, manage the DNA ecosystem and manage the business. So the ability of business leaders to observe these is really important. And I think it raises the final point that as observability increases, you will have more business leaders viewing this in a non-expert way. But data and analytics professionals should be prepared, I think, to help educate and guide their leadership towards making the right decisions. I think it is important that we as professionals bring our expertise to the decision-making process for enterprise leadership.

Andrew Goss/Host:

Got it. So let's circle back to something that you've kind of been alluding to. And so I want to follow through on this. Let's get to converged and composable ecosystems might seem like a mouthful, but at its base level it's actually quite simplistic, at least I found it that way, and this is coming from a marketer. So take that with, it really is simple. Can you talk about exactly what converge and composable ecosystems is, what the report laid out and what is to seen in that area as well?

Nathaniel Loman/VP of Data and Analytics, TransUnion:

Yeah, it is a mouthful, but to me, I agree. It's fairly simple, right? To me, it refers to a modular, fully integrated, scalable end-to-end architecture for data and analytics is not a platform, but it's facilitated by a platform. It includes the knowledge, the people, the solutions that are built and enabled by the platform, but it's bigger. But if done right, again, we're bringing together that data fabric we talked about earlier. We're breaking down silos and getting everything onto one integrated end-to-end architecture, right, is going to massively accelerate product development. It's going to massively improve scalability and reusability of products. It's going to lead to significant efficiencies. Just imagine, I think this is another case where perhaps that the counterfactual is instructive. If we think about trying to build new solutions on siloed pieces of an ecosystem that the data isn't integrated, there's different applications that are not integrated and you have to tie those together.

Maybe you can make something work once, but then can you scale it? Can you tie in other products and it quickly becomes a bit of a Frankenstein solution versus a single end-to-end architecture that you can scale. So I think most of these are going to be cloud-based, so easy scalability, and in our experience, focusing on the software layer more than the hardware makes it much easier to scale again. But think about how much simpler it is if you have this well-designed end-to-end architecture to build new solutions that can interact with other pieces of the ecosystem in a natural way that doesn't require call outs to different systems, doesn't require a lot of integration that can scale together without one piece breaking or turning out to be a bottleneck for the scalability of the platform. So generally we think of the ecosystem as a combination of capabilities that are built on a platform and in the platform. So in the platform is a fundamental functioning of the platform, whereas we built solutions and capabilities on top of the platform that sort of make up the ecosystem. So it's a little bit of conceptual thing, but overall I think it does come back to a simple concept of a modular end-to-end integrated and scalable architecture.

Andrew Goss/Host:

Got it. And I think part of the answer to this next question that I have for you, we got to, and this, I think it's a good segue from convergent composable ecosystems, now that we've gone through many of these top trends that Gartner laid out in its report, let's focus on implementation. How can businesses incorporate these initiatives and activities at a high level? Lots to digest. So feel free to take your time here.

Nathaniel Loman/VP of Data and Analytics, TransUnion:

Sure. Well, of course I'm going to be little bit biased here, but I do think you should push data and analytics to the forefront of decision making. You want to make informed decisions, so bring analytics and data to the floor. As an analytics professional, I think it's important that we inject our successes into the conversation and demonstrate our wins to get to that value-based optimization to help our leadership understand the value we're bringing and help them make the right decisions there. So again, longstanding challenge, but important one, I think it's important to socialize our capabilities, skills and insights. Again, to the observability piece, we're going to have a lot of stakeholders now able to dig into data when we have ecosystems that are accessible. Maybe it's just business insights, maybe it's something more complex like machine learning as a service capabilities where a business user can create models and deploy them quickly, right?

That's going to democratize data and analytics to a large extent. But it's important that we bring our expertise to ensure that people are doing that responsibly and getting the right conclusions, building the right solutions, and really accessing the new paradigm of machine learning and AI in a way that is done correctly. I think it is important to avoid silos in our architectures and make sure that we design our ecosystems in a way that they are modular and flexible and efficient in ways that they can scale and you can build on them going forward. That's a big upfront investment to be honest. And it's not easy because ultimately you have to design this end-to-end solution, and it's very hard to shoehorn a bunch of old solutions together. So it's a big upfront investment, but the end state is worth it. The end state is absolutely worth it because it's going to lead to sort of step change capabilities in terms of what you can develop on that kind of platform or ecosystem.

Data sharing will of course remain very important. Doing it responsibly will prevent bigger problems in the future. And then AI build the expertise for AI in-House. Be ready to be nimble as new developments in AI or new approaches sort of come to market. Understand how you can incorporate those. Make sure you have a responsible, in our case, we think a committee to responsibly guide our AI strategy makes sense. Maybe that makes sense for our listeners here and control the risk of it, but don't lose sight of the opportunity, right? I think it is important to recognize that there are going to be opportunities here and you shouldn't be sort of paralyzed with fear when it comes ai.

Andrew Goss/Host:

Got it. Well, that's a great overview and guide for folks when they're looking in this area. So thank you so much. I don't see any comments in here from any or questions from anyone, but definitely appreciate all of the emojis on there, everyone. And if any questions come up after the fact from our audience, please reach out to us in this chat or by any other methods. So thanks so much, Nathaniel for joining us. This is really interesting and insightful conversation and a special thanks to our audience for joining us. Again, feel free to drop us a line if you have any questions as a follow-up and just a reminder, and we'll put it up on the screen here in a second, but to download the Gartner report, top trends and data and analytics 2023, visit www.transunion.com/data and analytics 2023. We'll put that URL up here on the video screen in a moment. And it should also be on our LinkedIn live page. And until next time, we'll see you on TransUnion live.