Details
Nothing to say, yet
Comment
Nothing to say, yet
Listen to keith datatalk final version by Destiny White MP3 song. keith datatalk final version song from Destiny White is available on Audio.com. The duration of song is 21:48. This high-quality MP3 track has 256 kbps bitrate and was uploaded on 21 May 2024. Stream and download keith datatalk final version by Destiny White for free on Audio.com – your ultimate destination for MP3 music.
Experian's Ascend technology platform aims to simplify the complex world of credit, fraud, and analytics by integrating multiple systems into a single interface. The platform helps businesses streamline their processes, improve productivity, and meet regulatory requirements. The development of Ascend focuses on providing a seamless user experience and allowing for easy integration with other vendors and in-house systems. Challenges in merging these systems include aligning technical interfaces, managing database schemas, and ensuring real-time processing. Experian has addressed these challenges by extensively integrating its products and providing automated security patches to enhance security and reliability. The platform offers a range of capabilities while allowing organizations to choose and integrate their own tools. The goal is to provide a smooth transition without the need for complex migrations. Hello, friends. Welcome to this episode of Experian's Data Talk podcast and video show featuring data science leaders and technologists from around the world. At Experian, we believe data has the power to transform lives and create a better tomorrow, and this show is dedicated to data and tech topics that matter to all of us. My name is Destiny White, and on Data Talk, we are talking with data science and tech leaders around the world. In this episode, we chat with Keith Little, our Managing Director of Analytics, Decisioning, and Platforms at Experian Software Solutions, about the recent enhancements to Experian's Ascend technology platform. The cloud-based solution integrates analytics, credit decisioning, and fraud tools into a single interface, which enables businesses to achieve greater efficiency with lower investment of time and money. Here's our conversation. Hello, Keith. Welcome to Data Talk. I'm so excited to chat with you today about the Ascend technology platform. So, what would you say are the main goals behind the development of Ascend? Well, thanks, Destiny. Great to be here. So, it's really about trying to help our clients deal with the complexity of what they're having to deal with on a day-to-day basis. So, if you look at the world across credit, fraud, and analytics, it's a world where you're having to balance so many things, so many priorities, for your clients, for your customers, for our clients' customers. And it's a complex world. So, when you look at everything from the technical integrations, through to the business integrations, through to how it, you know, the regulatory and compliance requirements around this, trying to hook together multiple systems around, you know, credit, fraud, and analytics, it gets very, very complex. So, the whole sort of ethos of the platform is around driving, you know, streamlining to make people more productive, trying to drive simplicity, although it's a very complex area, of course. So, you know, making some of the stuff that our teams do in terms of analytics can be always hard. But what can we do to simplify that? And really to try and build that platform in a way that, I guess, from a potential, because we have so many customers and clients, how do we do this in a way that, you know, requires them no migration? It's an evolution strategy. So, how can we bring all these sort of new capabilities and those ability to sort of simplify to our clients as quickly as possible and as simply as possible for them? That's a wonderful explanation. So, we know that the platform integrates multiple applications into one interface, one single interface. So, would you be able to share any challenges involved with merging these complex systems and how you overcame it? Yeah, I think the first thing probably to just outline is think of the business problem. So, Alita, when sort of the approaches and then the headaches that I'm sure listeners would see in terms of their sort of day to day life. So, if you look at any sort of, let's say, credit originations journey, it all contains all sorts of aspects of sort of capabilities that you need to stitch together and choreograph effectively. So, think of it, you have to identify the person who's applying. You probably put all some sort of data, bureau data, plus maybe alternative data sources. You have to do all sorts of things like affordability checks, risk models. You might be integrating some sort of machine learning models for some of the data that you've brought together. And you've got to make a decision and you've got to, you know, and, and, and, and. And you've got to do all of that within like a couple of seconds, typically, in response. Think about the complexity of that. And in most organisations, you know, it's not as simple as you've got one system that does all of that. You've got, you've got many things you have to drop into, sometimes multiple vendors, so third parties more often than not. In-house systems, so you're having to deal with legacy and you're trying to sort of choreograph all of these types of systems. So, if you think of it from a functional, technical perspective, you're trying to get alignment between vendors and the way they interface with you. You're trying to get your engineering teams, you know, priorities and are those priorities aligned to the in-house systems and how you bring it together? You've got all then the complexity of, you know, database schemas and making sure all the data points sort of flow through the system in a seamless way. And then you sort of overlay, you have to do that all in near real time. You then overlay to that all the regulatory and compliance requirements around that, all the reporting, you know, dealing with models, model risk governance, dealing with decisioning, affordability and fairness and all of these types of things. And when you look at that, you think, wow, you know, that's sort of complex. So, so what the platform is trying to do is, A, Experian is unique in terms of the range of capabilities it brings. So, it's got some amazing world class data. It's got some amazing world class software to be able to align to that data across, obviously, decisioning and fraud and marketing. And we've got all the sort of analytical smarts where we can bring that together in our product. So, the sort of the core concepts of what we're trying to bring by bringing those capabilities together in, you know, from a user experience perspective. So, it's sort of tough to tell that story about how do we sort of now I can say I can start to visualize what my puzzles are trying to be. I can start to drive that sort of integration and how we sort of build those sort of tooling together. So, then when you look at the challenges, it's firstly from a user experience perspective, we might have capabilities that are very disparate. So, you know, single sign on. So, as a client, I want to be able to log on. I want access to all the things I have permission to access. So, you see the range of capabilities that we offer across the platform. I then, as well as using those capabilities, I also want to make sure all those capabilities talk to each other. So, we've done a huge amount of work under the covers to make sure all of these products integrate with each other. So, we've got products like Sandbox, which is where you get access to all of our rich data. You've got analytics capabilities in there so you can build your models. Those models you can deploy using tools like Ascend Ops, which is a whole model development life cycle and monitoring capability. And then from those, you can then start to pull those into your sort of decisioning tooling. So, your orchestration tools, all the things that do sort of the workflow choreography that I talked about at the start. And whether that's a credit use case or even fraud, you know, fraud again, bringing in disparate data sources and trying to make decisions on the basis of those decisions and how they sort of align. And that sort of making that smooth is very hard and difficult. And I would say the teams have done a huge amount of work under the covers. So, yes, you've got a life pretty user experience and you can see all the capabilities. But the actual magic is under the covers in terms of all of that integration that's gone on to make sure all of the elements and capabilities will talk to each other. And so that's less, you know, it's one of the many headaches that we can start to reduce from our clients. And it's a big headache. We've also thought about, you know, those capabilities. You know, we don't expect every organization we work with, of course, to have to take all of our capabilities. They will have their own as well. So we've also made the platform as open as possible. And, you know, timing of suppliers and other vendors or in-house systems, you know, we want to make sure we can integrate smoothly as well. So I guess that that's one big puzzle. The other big one, of course, is security. You know, security. We have a security first mantra of experience, hugely important for us. And, you know, one of the puzzles that organizations have to deal with is security, of course. And that's the velocity of security vulnerabilities and the machine, the machinations of the teams that are working on that, you know, particularly in the example I use where you're trying to stitch together multiple third parties or in-house systems. You know, we build security, patch automation throughout our platform. So you can always be safe in the knowledge that the platform is secure and it's just happening without you needing to do anything. You know, just another headache out the way and a major one, obviously, you know, in terms of security perspective and then, you know, resilience and reliability. We spent about not far off a year re-platforming the entire platform into a much, obviously using cloud technologies, but using a much more robust sort of resilience pattern. So with all of that integration to try and make stuff as simple as possible, and you can get all very technical on sort of APIs and schemas and structures and how we sort of link all of those tools right through how we think about security and the ability to sort of just make that. Problem go away for our clients by maintaining sort of auto patching and all of that stuff and attack those vulnerabilities that keep coming up. And of course, then in terms of how we we think about resilience on that always on infrastructure that clients more and more demands because their customers more and more demand as well. You know, those have been the big challenges on this journey and, you know, the user experience, you know, say the icing on the cake, it's important because it brings those capabilities to our clients. But they can be sort of safe in the knowledge that they don't have to go through some awful migrations, you know, to another system, you know, another system. We've brought that evolution together. So people will start to our clients will be able to start to take those benefits out of the box. Wow, absolutely. Just so many different hurdles, it seems like to go through to get that seamless customer experience. Would you be able to talk about the Ascend analytical sandbox and how our clients can use it? Yeah, I mean, that's one of our it's a fantastic product. You think about it and maybe it's worth for the audience. No, I've been on the other side of the fence. I was an experienced customer for many years. I was a CIO at a large, complex organization, a large, complex bank looking after credit and fraud and and sort of payments. And within all of that, I've seen what it's like on the sort of the other side for majority of actually my career over 30 years. And and when I see the capability we're bringing together in the platform, when I see what the sandbox has in it, which is this, you know, this data resource around credit and fraud is, you know, it's manna from heaven. If you're a data scientist, this is the great sort of playground where you can bring your data, your client data, and you can start to work and analyze and create insights where you can build models based on those insights of what you're seeing. And you can start to tune your system. So, you know, as I said, it always starts from world class data and experience has world class data. And, you know, like I talked about, it's all about, you know, the platform is all about accelerating time to value. So let's talk a bit about the sandbox. So, as I said, in there, depending on what you get from us, but it's got, you know, 20 plus years in North America and every region has a difference across the globe. But in North America, 20 plus years of data, of credit data to go in and analyze, you know, multiple years of fraud data, hundreds of millions and billions of transactions and all sorts of things that you can start to look at and analyze. When you look at, for me, the excitement is you get that cross-industry perspective. So when you look at the most popular insight model, it's always looked at as a whole inference modeling, which, you know, you can get your data, you can look at the experience of cross-industry data. So if you think about what a credit team obsess about, it's effectively two, three things, three things. Accept rates. So are they getting people accepting in their credit offers and are those people of quality? So is delinquency being managed? Are they driving straight through processing? So have they built automation so they can, you know, this system, the client experience is seamless, productivity for the organization is low. And the third one being really around resiliency and, you know, security and resiliency, it's always up. When you look at the sandbox, you can do data and, you know, reject inferencing is the most popular model used there. You can see for people who you declined or you didn't make an offer for, you can see what outcome actually happened. You can see where they went to, what offers were taken up. You can see what their delinquency rates for. And it's fascinating to understand that because actually, typically, you can find that many of your clients who you declined actually turned out not to be bad. Why is that? What is that? And you can get sorts of insights from that. So that's hugely, hugely powerful in sort of getting those accept rates back. So you can tune either your scorecards, you can build models, tune your rules, et cetera, to sort of adjust and bring that sort of flow in. And then alternatively, on the other side, you can start to see, you know, a cross-industry view of people who you actually offered credit to, but they declined you. They opted out from taking on yours, but they went to competitors. You can start to see where did they go? What, you know, T's and C's did they sign up to? Again, all the sort of, you know, the factors that sort of, you know, forced them to make that decision. Again, you can start to tune your models, your insights, your rules and all of that type of stuff. So you get this huge, huge, rich insight of data in there. And then similarly on fraud, you know, this massive, rich, you know, suspicious emails, you know, many, many transactions, sort of billions of events of fraud events. And you can start to create clusters of these and start to look at, again, what's going on in terms of, you know, the two things effectively a fraud strategist will be always looking at is genuine decline rates and value detection rates. So effectively, are we keeping the bad people out and letting the good people in? There's nothing more, there's nothing more frustrating than getting your car declined when you're not a fraudster, because that can really annoy your customers as well as obviously you want to keep the fraudsters out. Right. So you've always got to have that sort of balance. And it's always a continual tradeoff that fraud strategy teams around the world are always trying to balance. And it's really important. So, again, imagine having that cross-industry data so that you can start to analyze and see what's, you know, see what's an augment to your sort of data sets. And we've had some amazing results with where people are. And then, again, to bring it back to the platform story, the Sandbox has been around, you know, it launched a few years ago. It's had amazing traction across industry. As I said, people are using it for these sorts of tools that I just mentioned. But, you know, it's all about then, OK, so, OK, I can see that insight. I can see that I need to update my models or build new, better models. I need to change rules, scorecards, etc. How do I go from that as a data scientist in my data science team to actually getting that into production, seeing this then for real? And that time to value, that sort of that cycle time is something where, you know, I just go back to the previous sort of about the complexity then of how do you bring that into your flow? And, you know, there's all the governance and compliance that goes with it. So, again, this is where the platform helps. So, you know, with the user experience that we're now delivering and that integration, you can now take those those insights. You can build your models in the Sandbox. You can use tools like a SendOps where you can also register those models, go through all the testing and compliance. Once those models are there and available, you can then go into, if it's a decisioning, you know, use case, you can then start to consume those models immediately and run all your tests and regression testing and see those improvements for real and then go into productivity. Again, taking out many, many steps in that integration complexity that organizations have to deal with, particularly if every one of those hops is like a different vendor or a different, you know, a different tool that you stitch together. It's lots of complexity that often have lots of engineers involved tinkering as well. You can't do it self-service, which is what our platform is sort of starting to enable. Again, it's taking that complexity out and sort of focusing the teams on what they need to be focusing on, which is things like acceptance and delinquency and straight through processing and automation and things, as opposed to spending lots of time having to do engineering and integration, spending lots of time working out governance and compliance and all of that type of stuff, which you do need a level amount. But is it too much, you know, because you've got all that complexity of stitching all of this stuff together and reporting, et cetera. But the sandbox is where the heart of all of our data and it always starts. And typically where, you know, the insight that you can gain is extremely powerful. Wow. Wow. That's super fascinating and seems like a very powerful tool. Keith, would you be able to tell us exactly how was client feedback incorporated into Shaping a Send? Yeah. So I guess part of it was I had some observations as working years as an experienced customer. I also dealt with a lot of experienced competitors as well. So I knew all the good and the bad around the group. So it sort of came with a perception and I'm certainly not always right. But in terms of sort of setting the strategy, there was clearly opportunity in terms of experience, got this wealth of capability. And as a customer, some of the capabilities I didn't even realize Experian had sort of almost sort of hid it in a sense. So, you know, it was people if you knew, you knew, but sometimes you don't. So bringing all of those capabilities together, because again, we're unique. Nobody else has the data. We have all the software to own that sort of life cycle. So it was clearly an opportunity that we need to look at. So clearly, as I said, I'm certainly not always right. But myself and the team, we looked at, we actually went out to, we partnered with Forrester, the analyst organization. And we went out and asked over 600 financial services organizations around, you know, sort of questionnaire in terms of all sorts of, you know, looking at the hypothesis of everything from platform, a platform approach to sort of convergence in terms of vendor management. What sort of what sort of what sort of insights we're getting in terms of what are people's pain points. So we're trying to sort of really plug in pain points across the life cycle. So, you know, I articulated the complexity I had to deal with, you know, dealing with multiple vendors, trying to stitch all this together, trying to do all of that in a secure and reliable way, trying to do that, you know, with all the regulatory compliance governance data complexity that's sort of driven there. And it turns out I wasn't my I wasn't the only one. And we got very strong feedback from from doing that research where we saw stuff. And just, you know, just to share some sort of, you know, sort of headline, headline sort of numbers, we sort of got to see, you know, almost half of the half of the people replied. Sorry, the customers, clients, financial service organizations, not just our clients, replied, you know, they're looking to reduce their vendor set because of the complexity. They give us numbers which are very scary, but it takes them up to 12 to 15 months to get the models I talked about that you build in your sandbox into production. And that a number of somewhere in the region, about 60 percent of models actually never reach production because it just it just never got through the process. We sort of heard that we heard, you know, all the things around security and, you know, and obviously regulatory oversight is getting heavier, you know, across across the globe. And having to keep up with that sort of cycle time and speed, that something needed to change in terms of how we think about doing business in that sort of area. So so there was a huge amount of insight looked at. Obviously, we've got a load of expertise with an experience brought back together and sort of the best minds driven by those pain points, you know, come up with a sort of strategy to, you know, bring the capabilities we have together and then look to evolve that over the future. Absolutely wonderful insights, Keith. Thank you so much for just taking the time to unpack the technology platforms today. I really enjoyed speaking with you. It seems like such an exciting innovation and I'm really excited for it to get into the hands of the public here. So thank you so much again, Keith, for coming on Data Talk. Thanks a lot. Thank you for listening to this episode of Data Talk. As a quick reminder, our show notes and video clips from today's podcast are available on the Experian blog. Just go to Experian dot com slash Data Talk, where you can also find the full archive of past shows. Thank you so much for listening. If you enjoyed today's show, please consider sharing with a friend and giving the show a five star rating. Your vote can give this show more visibility.
