Episode 29: Making Privacy Your Marketing Superpower

In this Digital Clinic episode, Justin Smyth, Director of Business Solutions at Wimmer Solutions, draws from his experience at major brands like Microsoft and Nordstrom to explore how privacy regulations are fundamentally reshaping digital marketing. As traditional tracking methods are increasingly less viable, Justin explains why first-party data has become the foundation for effective marketing. The conversation covers practical AI implementation in marketing workflows, the difference between simple AI tools and sophisticated agentic systems, and Justin’s vision for the future where marketers become orchestrators of AI agents rather than direct task executors. With insights on building privacy-compliant data infrastructure, navigating the evolving regulatory landscape, and preparing for AI-powered search and shopping experiences, this episode is essential for marketers looking to build sustainable, customer-centric strategies in an increasingly complex digital environment.

Listen & Subscribe:

The Current State of Privacy Regulations

Aaron Burnett: There are so many things that I’d love to talk with you about. You have such a fascinating background, and you also have very interesting and well-informed opinions on a number of topics. You know that we focus on privacy-first industries, and we have a concentration in healthcare and MedTech where privacy regulations are really a principal constraint on digital marketing in these industries. They’re increasingly a constraint on digital marketing in most industries. I’m curious to know your perspective on the current state of privacy regulations. I’m also interested to know what you think people should be paying attention to and what the implications are for the future where you think we’re headed. 

Justin Smyth: It is a unique time. I’d say drawing back to my earlier career, I’ve done a lot around leveraging the data to provide these interactive experiences or really personalize the marketing and advertising campaigns that I’ve created. In those earlier days of understanding the importance of the data, I really took this perspective of I have to be careful with how I’m using this data and taking that privacy-first approach. I think in this era that we’re at now, it’s getting ever more complicated. We don’t have mechanisms like cookies that we could really depend on anymore, or even IP tracking that we leveraged quite a bit in the earlier days. We really have to think about all the different regulations that are coming out. GDPR was the first, but here in the US you have so many different variations from Colorado to California and it’s just going to get even more complex until the federal government has more of an overreaching policy that comes about. For marketers it’s not going to be easy. You really have to think about the data that you’re leveraging, but really how are you activating that data? It’s not about just collecting more, it’s really earning that access and really thinking in a secure manner. 

Aaron: So in this rapidly changing environment, given everything that’s happening now and we think will happen in the future, how has the character, skills, aptitude, disposition of an effective digital marketer changed? And what does the digital marketer of the future look like? Who should we be hiring? Who should we become? 

What Makes an Effective Digital Marketer in the Privacy Era

Justin: Those are all great questions. I think the marketer themselves really thinking of the customer, of the service and experience. Not just saying, I want to go target more people, or I need more data, but what is the experience? How do you want to best service your customer? It really starts there. We have to leverage data. So being a bit privacy mindset too, how do you best manage that data? How do you stay compliant? And then with the ever-changing landscape, there’s always something new, really digging into what’s current, what could be coming out, and thinking more forward too, I think is important. 

Aaron: I think I have noticed that there are people who have been very effective digital marketers who have really struggled with AI, struggled to wrap their heads around it, been resistant to it, not wanting to integrate it into their own workflows, not wanting to adopt it. And sometimes it’s been surprising the type of people, the individuals who have been resistant versus the type of people who have learned very quickly, had enthusiasm, been intrinsically motivated, embraced it. Have you had a similar experience and are there common characteristics among those who are thriving in a more AI driven marketing ecosystem versus those who aren’t? 

Justin: I think that’s two-pronged. It first starts with customer obsession. The marketers that I’ve seen that really excelled is really getting into how can I best service that customer? What kind of experience can I create? Which helps develop the technology, whether it’s AI or just analytic platforms at times too, but it’s digging into the customer experience and that experience you want to drive. And then the other part is the innovation and really looking at what’s new, what’s innovative, what can I leverage now, because the market’s always going to change. If you just stay with what’s the normal and what everybody else is doing, at some point you’re going to get left behind. 

The Reality of Using AI in Marketing Workflows

Aaron: I assume you use AI quite a lot in your own work, a time or two. Yes, I would assume. Does the work feel different? Does the experience of completing something having used AI to augment the process or deliver more quickly or conduct research on your behalf, whatever it might be, does the experience of delivering that work feel different when you use AI versus not using AI? 

Justin: It is. I’d probably first start when I just started engaging with AI. I looked at what was coming out. I was like, this wasn’t right. It’d be a little bit off or not aligned to what reality is. So I did realize I need to put a bit more of a lens on it because if I just let AI do its own thing, it would just not be correct to what’s reality. So that was the first aspect to it. But then as I started to really put that lens on it, create some frameworks and really hone it in, that’s where I started to unlock some value where my research, it would be a bit quicker to do my research or to create some content or start mapping out processes that helped me do more with less, I guess you could say, which has been very helpful. 

Building First-Party Data Infrastructure at Scale

Aaron: So in healthcare and MedTech we’re dealing not just with state level regulations but also with things like HHS expansion of what could be a HIPAA violation. So the expansion of the definition of PHI, FTC enforcement of privacy violations and data breaches. And so in those contexts, in 2022, we started dealing with this reality where the new guidance from the Office of Civil Rights at HHS really rendered all third-party tracking a de facto HIPAA violation. If you were going to comply with those regulations at the time, you would need to not rely on any third-party cookies, data from any third-party platforms at all. Which leads to, I think, a necessary focus on and reliance on first-party data strategies. And you talk a little bit, I know you’ve worked with very large clients. You’ve worked at Microsoft for a number of years as a principal consultant. You previously worked at Nordstrom, so in very large ecosystems where data, the protection of data, data security are important, but so too is the ability to market in a sophisticated way. Share a little bit about the role of data strategy and in particular the rise of first-party data as central to digital marketing. 

Justin: The rise is especially important now. I look back into my Nordstrom days, which is now what, eight, nine years ago, and we were quite advanced in our approach and in our strategies. Where I look at it today, I am like, wow, it really has been eight years and we haven’t advanced much industry wise. But what we did there was really look into our data and what can we leverage. But we had to be very smart in how we did that because you can’t just stitch it all together and say it’s one and done. So through that we had to create the appropriate pipelines. We need to make sure we had these sandboxes that were clean data rooms and making sure it didn’t connect. That was the first step, is really going in and assessing what data do we have? What are the different processes, what are the pipelines we could establish? And working very closely with our privacy and security representatives to make sure that we’re constructing it appropriately. And once you lay that foundation, it gives you a lot of opportunity then to apply. Nowadays we call it AI, but back then it was machine learning to start building propensity models, creating these lookalike models to say, does this group have affinity to certain brands, or they most likely to be serviced in a certain manner? And that really unlocks some potential for us to better target and service our customer, which we know Nordstrom is big about their customer experience and service. But we also got to do some unique suppressions too, where we were able to reallocate dollar spent. So instead of spending millions of dollars on a user that at first we didn’t even know we were targeting them, but then we found out there’s no interest into buying that product. So why are we going to continue to spend millions of dollars there when we could suppress that user and then target other prospects that are more effective? 

Aaron: You mentioned needing to have the right policies, the data pipelines in place that are compliant, and that having those things in place gives you a resource that you can use for propensity modeling and then activation. What does that look like? What are the elements of ensuring that you are properly gathering, curating, and ensuring permission for first-party data and then what does that look like to activate on the backend of that? 

Justin: We had to work very closely with our privacy and security personnel because they know the regulations and the laws best. And we would first identify what systems do we need to pull from? Was that an e-commerce system? Was it a point of sales? Was it maybe just loyalty programs? And what are the type of data attributes we want to pull from that? And once we identified it, that’s where we pulled in our privacy personnel, look through and say, can we stitch it in this manner? Or do we need to be mindful of how we’re aggregating that data? Once the framework has been set up to what data we’re pulling in, what is compliant to pull, then we started leveraging technologies. At that time was like AWS along with Snowflake and Snowplow that led us aggregate that data, apply Python scripts to then programmatically put them into different cohorts. And then we would leverage, at that time it was LiveRamp that helped us onboard that data compliantly through hash emails and such. It’s getting a little technical there, but leveraging platforms like LiveRamp really allowed us to be able to bring that data together in aggregation, but then start matching it up across the web to say we have 50% confidence, which is very high. Most of the time it’s about 30 or 40% confidence. That is the customers that we’re looking to target. And then you’d start connecting into other platforms like Google Analytics or Adobe Audience Manager that allowed you to activate that data. And that was about eight years ago or so. So I think technology has advanced quite greatly, especially in the composable CDPs that are out there nowadays, like Hightouch. Adobe themselves has advanced quite a bit too with their offerings. 

From Data De-identification to Targeted Activation

Aaron: For those who aren’t familiar, because I think this is maybe something that people talk about a lot, but is mysterious. It’s a mysterious step. So you’re working in de-identified data in a context like LiveRamp, and then you want to activate the data. Mechanically how did you go about activating that data without ending up back in a context in which you were targeting people at an individual and identified level? 

Justin: In the backend, we did a lot of formulas to aggregate that data based off of some key identifiers of the customer, like their email, their phone number, location. Once we’ve aggregated it from the back end, then we hashed it to de-identify who that customer was and a hashed identifier, and then upload that into LiveRamp, which then LiveRamp would do the matching to say, this is where we feel most confident that this is a customer matching into our population. 

Aaron: So more recently, you’ve been doing a lot of work around AI. The foundation of AI I think is data strategy. If you don’t have a good data strategy, if you don’t have a good data foundation, you certainly can use AI as a tool, but you can’t use it in a systematic way. You can’t use it in an advanced way. So I think the implication of what we’ve been talking about around privacy regulations, building a first-party data store, is that there is a data structure to support that. How does that data structure, how does that data foundation then support the type of work that you’ve been doing in AI? And maybe describe some of the more interesting work that you’ve been doing there. 

Justin: It really starts with modern data. So a bit of what I explained in Nordstrom back in the days of the different mechanisms and the way that we structured data, I’m seeing that now applied in the modern data architectures and some of the tools that you can leverage too to make sure that you have these clean rooms to bring your data into. And then applying machine learning to create those cohorts, aggregate your data. That’s, again, your foundation. And then from there you get to apply the AI because your data’s a bit more structured in those manners. The AI then can start making decisions off of it or leverage those algorithms that you created for your cohorts and propensity models. 

Interactive Marketing Experiences in the Zero-Trust Era

Aaron: We’ve communicated about a few of the types of projects that you’ve been doing. Agentic experiences, AI enabled consent management, personalization using AI. Can you describe a bit of that? Quite sophisticated work. And then I’d like to return to data strategy and how that can help or hinder that sort of work. 

Justin: Where we’ve been leaning more towards, and it draws into what I first mentioned of earning the access and thinking of that zero trust era. With that, we’re thinking of more of these interactive experiences and it’s really driven by mobile as well, or the augmented reality. If you think of Pokemon Go for instance. You go to event and then we have little Easter eggs throughout the event that then tie back to that brand. So then you could search for these Easter eggs while you’re at the event. That gives you more knowledge about the brand or helps you find these products. I think that’s a good way to get the access that you’re really trying to earn from the customer. But then it’s interactive for the customer. It’s not just you’re going to target me, it’s, I want to engage with you, I want to learn more, and I want to get more from your brand. 

Aaron: You mentioned in a recent post that you were working with a large client on an AI project and the project was hamstrung by data fragmentation. So can you talk a little bit about that and how data fragmentation can get in the way of these more sophisticated implementations? 

Justin: Data fragmentation is always a struggle to provide not only appropriate experience, but then to be able to target your customer appropriately. You could take a few different approaches to that, where you could either structure your content to represent experience well and then connect into your data, or as I mentioned earlier, taking that approach first, really assess your data and how can I bring it together? And with modern data tools and AI, it really helps you stitch that data back together. 

Agentic AI: What’s Working Today

Aaron: You’ve mentioned agentic AI, and I’m curious here. I read a lot about agentic AI. I see people posting a lot about agentic AI and the promise of agentic AI. I don’t see people posting about real world implementations that are delivering value today. You are working in very large and sophisticated environments. Are you seeing agentic AI in production delivering real value in a sophisticated manner? 

Justin: I see it done in a B2B environment and more so in the back office where it’s really helped augment roles and help, particularly in the customer support area. Help them resolve cases much quicker and augment their expertise and knowledge. I have yet to really see it done on the customer level yet. There are some chatbots that are coming out that are more advanced, that allow you to have a natural conversation and talk with this AI agent, which I’ve seen done quite well with some, but I haven’t seen that scale into manner yet. 

Aaron: So on the B2B side, I’m intrigued by this. What are you seeing that works and I would assume this truly is chained agents, an agent who is expert at one aspect of service delivery, passing a work product to another that refines it. Is that what you’re seeing as well? 

Justin: It definitely is, and especially when you think of a tiered support in a support model where, first it’s self-serve, how can I find something? Search has been very core to that, a search engine of being able to find the documents and be able to serve it up quickly. But then you go that next step up to then say, all right, if you’re just an admin level representative and I just have an issue with my password, how can I help resolve that quickly? And it draws back into the knowledge base that it has. And then looks at cases from the past and be able to make the best recommendation. But where you start getting more advanced is really augmenting the experts. So when you have a senior level expert that’s really leveraging that tool to help them resolve cases, maybe talk through strategy a bit, that’s where I’ve really seen it unlock because it gives that strategist more time to really focus on the problem and really dig into it while the redundant type of task could be resolved by that agent. 

Aaron: I’m curious, how is agentic AI different from simply creating, let’s say, a custom project or a custom GPT and training it on a significant corpus of data or in an enterprise setting, training it on an extensive wiki? Why is agentic AI different from that? 

Justin: Agentic AI really gets into not only having that knowledge, but then being able to run functions for you or even chain to another agent. You could first have your research agent that goes through and does all the research, looks at the different sources, and then gives you a recommendation. But then that next step is then from that recommendation, build out a content post, a blog post or two, and then from that blog post, post it to Facebook, post it to our website and then maybe take it a step further to then say, let’s maybe do some personalization. So based off of that content that you created and some customer cohorts that you know, now when that customer comes, serve up that appropriate content to them. 

Preparing Clients for Exponential Change

Aaron: All of the things we’ve talked about so far, privacy regulations, centrality of first-party data, which is a fundamental shift, the meteoric rise of AI and the very rapid development of capabilities, increased capabilities in AI. How are you guiding your clients? It seems to me that the rate of change right now is exponentially faster than at any time in my career. From moment to moment, I either think I’m ahead, behind, or doing all right, and I can’t quite get my footing depending upon what I’m reading at the moment or what I’m exposed to. How are you guiding your clients and how are you enabling them to not just keep up, but maybe plan for the future? Create a foundation that enables them to react to what will be increasing change. 

Justin: It starts with that privacy mindset and data mindset. Privacy is going to be your performance. You need to really think of how you’re going to gain the user’s trust and then get them to engage more deeply so that way you’re going to drive more sustainable growth. But how do you get there? It first starts with a readiness assessment. Do you have the data, which most people do, but then from there, how can you best activate that data and do it in a privacy mindset too? And that really starts going through auditing the data flows. What type of consent mechanisms do they have in place and what exposures might come about? You really have to think about the risks that come when you start leveraging your own data too. And that gets you ahead of most of others. Once you know that readiness assessment’s in place, then you can start creating your first-party data strategies. So do we want to create cohorts? Are there certain suppression mechanisms that you want to put into place? Is there ways that you want to leverage your content further? And help them figure out what’s the most effective data strategy for them that’ll help with their business goals. And then from there, start putting in the different technologies and solutions that could help activate those data strategies, whether it’s an Adobe platform, Google, Snowplow, really identify what are those platforms or tools will best fit for their business needs. And then start working to build those out. 

Aaron: I’ve talked to a lot of people and I’ve had this experience myself that when I can accelerate my work by using AI and sometimes I can really accelerate my work. Like you, I’ve created lots of custom GPTs and we have Claude projects and we have workflows internally. We stop short because we focus on healthcare and MedTech space where absolute accuracy and authorship are really valuable commodities. We stop short of developing content with AI, but we use it to augment our thinking or processes or conduct research or that sort of thing. And my experience is that it can feel like cheating to be able to deliver so quickly and so efficiently. And that’s something that I find that I am still adapting to, getting comfortable with over time. I’ve talked with lots of other people who’ve had the same sort of experience. I think the fundamental nature of what it means to do work and deliver value is changing quite quickly in a way that maybe at a soulful level, it’s taking us time to adapt to. 

Justin: Yeah, I would definitely agree. And I mentioned in the start it would become the outputs I would get. I was just, just didn’t feel right or it did have that bit of maybe I’m cheating here because I got something and is this good enough? So I put that lens on it and at times I’ll do a few iterations to really get to where I think it feels right. But yeah, the way that we work is definitely changing. I see it similar to the earlier .com era when web, the website started coming out, you had search and such, and it made work a bit easier to be able to search. Instead of you would go to library and you go find your books. You read through your books, and then you figure out your synopsis and you build out your articles. But now we have that next level of not just search, you have something that completes your work for you, but you still need to make sure you go through a few iterations and hone it in. 

Aaron: In the past you might ask someone who’s working with really cutting edge technologies or in large enterprises a question about the future and the time horizon might be 20 years from now. I feel like a futurist question at this point is three years or five years from now. So I’ll ask it that way. If we project forward five years into the future, what do you, give us a sense for what you envision the world of digital marketing to be? How are we able to identify, find our audiences, optimize campaigns, develop and refine creative, facilitate exceptional customer experiences, and deliver services and products? What does that look like in a way that’s very different from today? 

Justin: I think I would draw onto that agentic AI era and where the marketer becomes more of the orchestrator of these agents. And you have your different agents that is doing the audience research and segmentation. Another agent doing your creative and maybe even another agent doing the process mapping and flows along with your data normalization. So with that, your marketer becomes more of the orchestrator of those agents to run your operations. And then from that you still have a human touch just to make sure that it’s refined because I don’t think in five years these agents could go and operate on their own, but they will allow to automate a lot of your operations and processes. 

Aaron: Clients who adopted a first-party data strategy were able to really significantly improve performance in that way. Are you able to cite any examples or performance stats associated with incorporating a first-party data strategy into their digital marketing? 

Justin: What I saw at Nordstrom was one example of reallocating millions of dollars to really focus on higher conversion and increasing a conversion rate from two to three to 4%. 

The Evolution of Search and Shopping Experiences

Aaron: What have we not talked about that we should, or that you would like to talk about? 

Justin: Yeah, maybe where do we see the future of marketing experience going. I think of the earlier days of SEO and search engine optimization. We were really focused on optimizing for search and I feel some of that is starting to come back a bit. You have search now available on ChatGPT, and you could, I hear a lot from my customers where they just use ChatGPT to search instead of Google now or even Microsoft Copilot. But I think it starts getting interesting when GPT just this past week announced their shopping functionality and making it easier for the customer to just shop where you say, I’m looking for a fruit juicer, and what recommendations can you provide for me? And not only does it provide the recommendations, but it gives some context to why it’s given those recommendations. So we’re seeing search advance a bit more. But what if we maybe go five years ahead and you have your own assistant, your AI assistant there that does a lot of this searching for you. You just need to talk to it. And I think companies and brands really have to think about what that experience is going to be and how do you best service in those experiences. 

Aaron: As memory increases and context windows increase, you potentially have an agent who is your personal shopper, who remembers what you’ve purchased and what you have liked over the last “X” number of months and years. And yeah, you’re right. That can be very compelling. I will say in order to make shopping on ChatGPT work, there is a simple but critical update that needs to be made to what’s called the robots text file on a website. We’ll post that change in the show notes. Any web developer who has access to your website can make that change. It should take all of 30 seconds, but without it, you won’t end up with your product inventory in ChatGPT. 

Justin: Crazy to think that robots text is still quite important. 

Aaron: I know. So I think that an implication of what’s happening here. My belief is we’re going back to an era in which the communication that you can do, that you should do behind the scenes. Robots text being one, but also schema. So markup language is really important. It was important for other things historically. Sometimes for authorships, sometimes creating relationships or semantic search, but the schema is very much in the ascendancy because of the way that it enables AI applications, engines to understand, interpret the context of the content on a given webpage. So we are spending a lot of time doing that again. 

Justin: I’d agree that structured data really helps feed into those AI tools and helps them decision. And then I think the other part to this too is content is still king. You need the content on your site and the more content you have across the web just draws your brand forward in those searches. 

Aaron: Yeah, I think the nature of the content though, the nature of the content that performs has continued to evolve. I think for a long time Google would say that content is king and tell you that you need to have high quality content and have certain characteristics, and that was aspirational. That wasn’t actually true. You could produce a whole lot of medium quality content and you could perform just fine. And then more recently, it is more true that unique content, high value content really does perform well at this point with the rise of so much AI generated content and the ability to create volumes of mediocre content. I think again, really thoughtfully developed, really well researched, really well produced, structured content will outperform all of the garbage content that’s produced in mass by people with access to ChatGPT, who have automated ChatGPT, maybe even in an agentic fashion to generate blog posts for them. So I’m hoping that the byproduct, the ultimate byproduct of a lot of this is that really high-quality marketing will trump mediocre volume marketing that is produced either by mediocre marketers or produced through automation. That’s certainly the approach that we’re taking, and it seems to be working so far. 

Justin: Yeah, I would agree. And I think it shows importance of really understanding your customer, digging into that experience you want to serve, but then you still need the human in the loop. The bots aren’t going to run everything yet. You still need a human there to review and provide that specialized touch. 

Aaron: Yep. I agree. So this actually reminds me of one other thing I wanted to ask you about that is related to content. So again, if we go back several years, the really simplistic way that people would approach content marketing is that I want to rank for a set of terms. And so I’m going to create this content that enables me to be performant in search. The way that we have approached it is this should be a part of a user journey, and there’s a part of the user journey where organic content is a right point of entry or it’s a point of emphasis. It’s the next message that a prospect needs to see in their journey toward conversion or education. One of the things that you’ve mentioned in some of our correspondence is AI enabled customer frameworks, and so I think by extension, customer journeys. Can you talk a little bit about how AI can be used to more thoughtfully develop or to refine and optimize customer journeys? 

Justin: I’ve been working a lot within Adobe’s customer journey orchestration tools and customer journey analytics. And it first starts with mapping out that journey that you want to serve and what are your different profiles behind that. But once you apply AI, just as you know with programmatic advertising, where a decision on what is the best ad I should serve, now you start mapping this throughout the journey and integrating your different channels so that way you’re serving the content to the customer’s needs. So if the customer likes to engage more in a CRM basis or email, you’re connecting with them appropriately through that channel. But then they might go on to paid media or social. You connect that experience then into those different channels. So it stays consistent throughout instead of fragmented. 

Aaron: This has been a fun conversation. I really appreciate it. 

Justin: Yeah, thank you for the time and I appreciate it.