I want to start by thanking all of you for your attention today. I'm glad to have had the spot right after the coffee break, so I expect eyes to be wide and everybody focused, if not maybe making a few bathroom breaks during the presentation. I also want to thank Magnolia for the opportunity to present to you guys today on my thoughts and my perspective on digital experience customization. So obviously I'm going to get started soon. 20 minutes is not a lot of time, but Queen's set at 1985 Live Aid was 21 minutes, so set your expectations right about there. All right, so you may be asking yourself, as I do at every conference that I attend, who's this guy? Why is he talking? My name is Gage Talbert. I currently serve as the Vice President of Digital Marketing at Frost Bank. Over the last 12 years, my daily mission has been to work across business and technical teams to deliver stellar digital experiences for our users. So I've had the opportunity to hone my skills in both business-to-business and business-to-consumer environments. I have been responsible for nearly 1,000-page redesigns, which honestly felt good coming here, but then I've heard some of the other speakers who have 160 sites, some with 600 pages. And I'm like, okay, maybe I'm small potatoes. But I digress. I've also had the opportunity to manage multimillion-dollar annual media budgets, as well as helped hundreds of team members apply performance marketing methodologies to improve campaign outcomes, and built and run multiple personalization, optimization, and experimentation programs. But this presentation is not about me. So today I'd like us to dig deeper into the question, how granular should your experience customization be? We've all probably been part of persona-building workshops that start with some prompt along the lines of, what does target audience XYZ feel, or what keeps them up at night? But the reality is, most of us are still trying to access data that is in silos. We're still trying to define, score, and attribute our leads. And we're still trying to understand user behavior and keep up with the laws and regulations that allow us to use it. So my pitch to you today is to thoughtfully consider optimizing toward personas instead of individuals. So focusing on the groups that are centralized around a specific need instead of necessarily trying to go for that nirvana of one-to-one, one-to-one. Just because with unlimited resources, that would be great. Most of us don't have unlimited resources, which is one of the things I'll get into just a little bit later. So allow me to elaborate just a little bit. But I encourage you to take a moment to take in and interpret the graph in front of you here. The graph on the left is essentially your audience size and the axis on the left. And then the axis along the bottom is the number of personalization characteristics. And basically, this is an illustrative example. But what it's meant to show is personalization is effectively targeting, targeting, targeting. So as you try to personalize on more characteristics, in effect, what you're usually doing is shrinking the size of that addressable audience so you can deliver a more relevant experience to them. But over the entire population that you started with, your net efficiency may end up returning about to where it was because you've reached such a small audience that it doesn't have the scale to make the overall impact that you're trying to make. So, smaller audience with customer experience, less relative impact. In addition to this, there are also some other potential pitfalls to hyper-personalization to be aware of. Brand reputation is a really big one. If you use too many data points to characterize or it's something that the audience was not exclusively aware that you were picking up on about them, your brand could be interpreted as creepy. So, that's never a position we want to put ourselves in. Attentive's 2026 study on personalization showed that 46% of shoppers are concerned about their data privacy. I personally think that number is probably a little bit conservative. Whatever the number is, I am personally within those ranks. So, I'll give you an example as to why. A few years ago, right before COVID happened, my wife and I were training for a half marathon. And so, we went out to a friend's house and we were just having normal conversation. But if you've ever met anyone who has participated in a half marathon, basically while they're training, that's all they can kind of talk about. So, that's where the conversation for the evening went. So, we were talking about training distance runs, energy gels, protein bars, things like that. And we stumbled onto the topic of one particular brand of protein bar that my wife and I had never heard of. We'd never done any searches for it, had not been in our consideration set. It was discussed. We went on about it. And the next morning, one of those protein bars in the exact flavor we had talked about the night before showed up on our doorstep as a free sample. So, very relevant, very useful, creepy. So, they lost our business because I was like, I don't know who gave them that data. I know it wasn't me. They were probably listening. It might have been Facebook. I don't know. But, yeah, they were listening and I don't remember giving them explicit consent for that. So, you know, sort of swore them off. Also, the more data that you customize on, the more potential regulatory and compliance risk. So, each data point may have its own unique set of regulations and laws governing how it can be used, when it can be used, things like that. So, it just gives you more to keep an eye on. Also, if you'll indulge me to return to the data availability and accuracy point, I do have one more example there. A few months ago, I received a text unsolicited that said, Hi, Dave. I see that your 2016 Honda Accord insurance has lapsed. Please reach out to us. We'd love to help you out with that. A few problems with that. One, if you remember from my introduction slide, my name is not Dave. So, there's issue one. Secondly, I've never driven a Honda Accord. My little brother did on the same insurance policy that I was on in high school. But I never did. And the Honda Accord that we did have was a 2009, not a 2016. So, across three different dimensions, they had inaccurate and outdated data. And because of that, it had the opposite of the intended fact. Instead of determining relevance, it showed that they knew nothing about me at all. So, I was not about to continue on with them. Also, you just have to take into account journey complexity. You know, with operational constraints, even with AI to help streamline your operations, in effect, what you're doing with personalization is strategically fragmenting journeys. And if you go too far out there, what will eventually happen is it will be much more difficult to deliver consistently great experiences. So, that's not what I want for you. It's not what I want for me. So, we'll move on. So, the methodology that I recommend is starting with data-first personas. Building all of your personalization efforts on data-first elements, all characteristics that you can know, not just things that you've developed inferences based on. That'll be part of the overall persona-building process. But for personalization, especially in digital channels, it's very important that you have specific data definitions that are going to enable those experiences. That will make sure you stay relevant and also provide efficiencies of scale. So, as I said, humans are not just data points. But for digital channels, we have to start by knowing the data first. What data points will you use to identify and which characteristics we use, which dimensions? Which ones will give you clear context if there's an audit as to why you personalize based on that? And also, how will they actually help you deliver customized experiences? Broader descriptors that we typically think of when just starting out in audience segmentation activities, younger users, certain interests, how close they are to a brick-and-mortar location, they don't quite go deep enough. So, if someone says, I want to build this campaign around personalizing for younger users, my first question is going to be, how young are we talking here? Is it Gen Z, Millennial, Gen Alpha? Is this for current demand, or are you trying to build for future demand? Because your objective around those could be entirely different. The same is true for urban versus rural audiences. How are you going to define urban versus rural? Is it based on ZIP or postal code? Are you even regulatorily allowed to use those as dimensions for personalization? And also, you could use Radius from Cindy Center, but you have to ask all these questions first in order to get to a persona that you can actually activate against. So, step one in my seven-step framework is to explore your data sources. Before you can think outside the box, you have to understand that a box exists, right? So, has anybody here ever seen the movie The Truman Show? Yes? Okay, that makes me glad. I'm in a good company. For those of you who have not, it's effectively about a protagonist who lives his entire life in a giant product placement ad. Everything he sees is personally relevant. It's highly curated. But until he actually starts going a level deeper, trying to understand inconsistencies that he's found, he's not able to actually reach his true potential. And so, if you don't explore your data sources in depth, you will be Truman, which ends up being a good thing. He escapes, but I digress. If you're starting out using personalization, exploring data, you'll probably be doing it this way for the first time. If you feel like you have a pretty mature program already, I still recommend going back to this, because data that's available in your data sources, as you continue to work through your backlog of tech debt, it changes pretty frequently. And if you haven't done this step or gone far enough, you'll likely miss out on opportunities that you otherwise could be utilizing. Some of you may recognize the image on the slide here. All right, I used the pointer. Thank you for the technology. That is the Burj Khalifa in Dubai, an 828-meter structure from the ground, really just a modern feat of engineering. But it starts in much the way that any house, any building starts. You have to dig down and make sure you're set on a good foundation before you start building up. Otherwise, you'll crumple under your own weight. And actually, in doing research around that building for those statistics, because those were not just common knowledge that I had, the first article I found was, why isn't the Burj Khalifa sinking? I'm like, okay, yeah, clearly they set a good foundation. So you will spend most of your time in this phase of the framework, because if you don't, you won't understand the full possibilities. You want to start by identifying all of the systems that you'll need to audit data for. That could be your CRM, your web analytics, your third-party data sources, which you'd want to list them out individually, and then dig deeper into them as such. But effectively, what you want to do is get your major buckets in line, have them all identified, and then start going one by one into them. Also, as you go, note connective tissue between the systems. So email address, email address in both systems. The connectivity between these data systems will give you previously unaccessible combinations of data to personalize and target on. The end result should be a comprehensive catalog of the data that you have access to, both for personalization and also just overall marketing capabilities. Also, as you go, note what data you don't have. Just because you don't have it today doesn't mean that you shouldn't pursue it. These would be things, characteristics, identifiers, dimensions, that you theoretically or basically intuitively, based on prior experience, know should exist, but you can't find them in the system today. They definitely could still have some value for you, especially if they're present across multiple systems, which you'd have to determine based on each individual data point. Use this exercise to learn a new language. As a business-side stakeholder myself, but working mainly across business and technical teams over the last 12 years, understanding the vernacular that each department uses to describe the same thing, and understanding when a no is a no, or no is a not right now, or it's a maybe, but I'll have to look into it. All those things are very, very important. We can't all be data analysts, and that's okay. You wouldn't expect business teams to be inside your data warehouse running SQL queries or anything like that, but you will need to work with your technical teams in order to develop this data catalog that I'm referencing. So one of the most common things that I've heard in business and technical team meetings upon follow-up is, the business team will say, I'm really not sure if they understood what I wanted. And the technical teams would say something along the lines of, I'm not sure they understood what I meant or what we're currently capable of. So a lot of the time, I'd say eight out of ten times that there's an issue between business and technical teams working together, it boils down to communication. And it's a fairly simple issue, but it can be crippling. Like, it just can't be surmounted in some capacity. So if you know the language, you'll be able to avoid a lot of those more tense conversations and team more efficiently. So, step two. Once you know how big the box you're trying to think outside of is and what's currently in it, throw out anything that's unusable or inactionable. And I don't mean that literally. Don't go and delete half a data source because it's not pertinent for this use case. But just put it in your sort of mental recycle bin for this particular use case. There are many reasons for data to be inactionable. The most pertinent here is the inability to scale. So what I mean by that is if you have an incredibly valuable data point, it speaks to the right audience, it will help you achieve relevance, but you only have it for 3% of your total audience, it's probably not all that useful because you're only going to be able to reach a very small percent of your population. But the data SMEs that you just learned the language of in step one will be able to help you identify those issues very early on. As important as the data that you'll need to customize the user's creative experience offer is also the data that you'll use to trigger it. So some data fields and technical infrastructure that don't at first glance seem especially valuable could be incredibly valuable as trigger criteria. So things like UTM parameters, first-party pixel data, JavaScript having run, a number of different things. Also returning visitor status, delivering a different experience to someone the second time they come to your website versus the first. So, step three. Now that we've done our research and we understand what I am coining the art of the actionable instead of the art of the possible, now we'll get into actually strategizing individual experiences and journeys. The starting point for that is choosing an objective that actually matters to your organization. This is honestly one of the major areas that I see a lot of teams go wrong, not just in their personalization, but in their marketing overall. Either they choose several measures of success, which in my mind changes them from KPIs to metrics, and it also leaves a lot of room for interpretation of whether or not something was truly successful on the back end. So if you don't sort of set that stake in the ground at the time of the ideation, you're really not going to be able to be in a good spot to say objectively, did this work or did this not? Also in that same vein, make sure you're collecting benchmark data if you don't already have it. Because that benchmark data is going to give you key context as to where did you start and how much improvement did you make via this individual personalization. So tie it to a business objective that could be an efficiency KPI, like reduction in Cost-Per's, engagement or conversion rate improvements, reduction in attrition rates, also efficiency KPIs, sorry, efficacy KPIs, like impressions, web visits, leads, awareness, reputation management, all those good things. And if you absolutely have to choose more than one KPI, choose at least one of each. So one efficiency and one effectiveness. One will make sure that you have an idea of relatively from start to current, how are you making improvements? And the other will give you that idea of, all right, but will I be able to drive considerable business impact overall based on my total objective? Now that we've understood our data universe and chosen our objectives, it's time to prioritize our minimum viable data. Now what do I mean by minimum? That could take a lot of different forms, but the way I look at it is at this point you should be trying to balance a number of factors. One, how many data points are required to truly achieve relevance with your audience? Because just personalization, like slapping someone's name on something if you have access to it, is not necessarily going to meet the mark of personalization. You have to make sure that whatever experience, offer, creative that you're delivering shows relevance to what their need is. Secondly, how many data points can you deliver uniquely relevant offers and experiences based on? I'll get into that one in a little bit more detail in a minute. Also, how will the accuracy and completeness of data impact your overall audience size? How many users and profiles do you actually have the scale of data to target with this personalization? Some data points, like email address, channel consent, or preference data, will be useful for just pure execution at the channel level, but they do still count towards your minimum viable data, even if you're not going to be, quote unquote, personalizing based on them other than in creative specifications. One of the things that you'll need to do, calling back to the graph on one of my earlier slides, is you'll need to work the numbers. All of this, you know, personalization always has to be contextual. So too does your strategy. So the graph you see before you is not a hard and fast, like, don't just say, oh, Gage said don't do more than five, you know, personalization characteristics. You have to run the numbers yourself, not just based on the number of characteristics, but on the individual characteristics themselves and how they affect your overall audience size. I'll give you an example. At one point in my career, my campaign manager recommended that we create three journeys across six stages and four touch points per channel for audiences of less than 100 users. So when you run the numbers, they're just not going to deliver the overall impact that we want to. So I'm going to skip ahead because I just got the note that I have one minute left, so I missed the five- and three-minute warnings. So I apologize to you guys based on that. But our last step pre-execution is to scope on available program resources. Take a look at this real quick. Try to drink it in in the 15 seconds that we have left. So I'll burn through the last few slides. But just take this into account that as you try to deliver relevant, comprehensive, customized experiences, the creative sprawl in a truly integrated, omni-channel, multi-stage journey can be astronomical. So the numbers you see on the slide here are just to get started. And in high-frequency campaigns, creative sprawl can be incredibly high. So within a year, three personalized journeys could add up to 1,000-plus assets if you're refreshing three times in a given year. So make sure you've talked to your team about the time commitment and how much available bandwidth you actually have to dedicate toward this. Otherwise, that foundation will not be set. You'll start trying to build these massive programs, and they'll kind of fizzle out and burn out because you can't sustain them. Anyway, moving on, a few other points. Building the journey and executing an evolution are just sort of like the last steps. I feel like a lot of presentations during the conference have touched on those, so I won't go into those too, too much. My presentation is much more along the lines of the preparatory steps and making sure that you have the necessary infrastructure to execute. But I would like to just mention that this framework is repeatable. Steps 1 and 2 don't have to be done especially frequently, but I do recommend that you revisit them from time to time just to make sure you're not losing out on new data capabilities. And that is all I have for today. So thank you so much. I think this is yours. Ooh, a clicker. Thanks. Thank you very much. Absolutely. Do I need to vacate for the next presentation? No, it's questions. Okay. Yeah. I see. Any questions from the room? For Gage, is this a very insightful thing on how to build a foundation for personalization? No. Why was I the only presentation who went over their time? I bet that's the point. You weren't the only one. There was someone else, but I won't name them. Okay, please don't. I'll name them later. Thank you. A question from me then. You obviously just demonstrated how to build a very solid foundation going through the data. How do you see AI influencing, accelerating, or how do you move forward with that? Yeah. So I definitely see AI as an accelerator. I think it was one of Chuck's slides yesterday that said, and I apologize if I'm misattributing this, but it basically said AI doesn't fix data issues. It amplifies them. So basically what you want to do is use this framework as an opportunity to set your foundation and then find ways to automate the things that are repeatable while still leaving the human strategic decision-making as a major core part of the process. Because you'll never be able to automate that true strategic understanding. You'll be able to automate some of your checks for when your data warehouse is updated with a new field. You'll be able to find ways to better and more efficiently develop creative using creative templates and things like that. But you still can't sort of factor out the human element there. Excellent answer. Ooh, we have a question. And for lack of a microphone. There you go. That one's great. Hey, you mentioned you have KPIs per persona so that you make sure that you're doing a good job. Can you give us a few examples of KPIs that you've used in the past so that how do you measure that the personalization is actually working? Sure. So I've done a few different personalization exercises in my current role that sort of popped to mind. Obviously there's the general. You personalize a page experience and you look for increases in engagement rate or conversion rate. We set up a custom application through Magnolia to develop templatized pages for each of our lenders in a specific area of our business. And so just the nature of visits in total and also attributable leads to those individuals because otherwise they would have just been attributed to the general internet bucket. Those were effectiveness metrics that we did use. Also engagement on our personal fraud prevention page. We set up a variant for our consumer side of our business and then our commercial-facing side of the business. And understanding the engagement rate improvements between the default and then the business-focused variant is definitely a huge KPI that we've used so far. Sure. Sure. Thank you very much. All right. Oh, okay. Do we have time for, I think we've got time for one more. The next speaker's like, yeah, take it. Oh, God. Wait, I'll run down. Thank you. So one thing that came to mind while I was watching your presentation is that a personalization effort in marketing based on like digital activity is an effort that can be solely owned by marketing. Right. If you start to talk about like personas and segments that require you to unify business data sources that aren't necessarily currently unified, that becomes a whole like organization data transformation effort. Yes, it is. I'm curious. It sounds like maybe you've made that kind of business case before and how did you convince the business to do that? Honestly, I started with data sources that I didn't have to ask permission for. So Google Analytics was one of the first and we set up some custom tagging via Google Tag Manager to measure the conversions that we wanted to keep track of the conversion rates on. And then we did have to eventually bring in our data warehouse team because, like I said, I'm not going into our data warehouse. I'm not setting up new tables. I'm not adjusting their existing structure because that would be terrible. I know enough to be dangerous, but I'm dangerous. So I started with use cases that I didn't have to seek as much external approval on to sort of show the value rather than telling that there's value. And then once I was able to show we can already attribute the data today in this system, all I need you to do is like import that data from this system into this system. And we find ways to, again, get that connected tissue between systems. That's when it really made it very easy for them to see the value is if you just make sure that this data point is here and this data point is here. Here's what I can do with that data. And just have, honestly, just a few examples was how I approached. Amazing. Thank you very much. I think on that note, it's a final round of the horse for Gage. Thank you very much. Thank you so much.