Tony Byrne, founder of Real Story Group, on CX Decoded
CX Decoded Podcast
February 22, 2022

CX Decoded Podcast Episode 7: Customer Data Platform Tech Realities With Tony Byrne

Traditional linear customer journeys are being transformed into endless decision loops. This creates big challenges for marketers, and Customer Data Platforms (CDPs) are one answer to that complex situation.

According to reports in the CMSWire CDP Market Guide, the CDP industry was projected to reach $1.55 billion in revenue by the end of 2021. It’s a 20% increase over 2020 revenue. CDP vendors based in the Americas account for 47% of companies in the industry, 59% of the industry workforce and receive 74% of funding.

In this episode of CX Decoded, Tony Byrne, founder of Real Story Group, takes us into the world of CDPs and shares his thoughts on CDPs and the challenges getting that customer data machine running smoothly. He also shares some real practical use cases for organizations.

Episode Transcript

Note: This transcript has been edited for space and clarity.

Rich Hein: Tony, we'd like to start off with a little background on you and your firm and maybe a fun fact about yourself outside of the professional world. You know, I think most of the people who are regular readers of CMSWire are already familiar with you, but for those who aren't, please, you know, share a little bit. 

Tony Byrne: Yeah, I'm the founder of Real Story Group, formerly CMS Watch. We're an analyst firm. Our mission in life is to evaluate martech vendors, put them under the microscope, squeeze them and see what comes out, and then tell people the real story on what the different vendor strengths and weaknesses are. Along the way, we've helped many large enterprises with their kind of broader stack decision-making. So we're a little bit unique in the analyst world that we only work on the so-called buy-side of the table, which means we only work with enterprise technology customers, and we never advise vendors.

Fun fact? I don't know how fun a fact it is, you know, the last five years or so I've made a little bit of an avocation around climate and climate science and global warming mitigation. It's been a really interesting area to research. I see a lot of commonalities, a lot of difficulties socially and organizationally with change management, a lot of multi-dimensional multivariate issues, a lot of complex decision making. And I sometimes find that my experiences in both worlds translate over into the woolly and fast-moving and sometimes deeply technical world of martech.

Dom Nicastro: So Tony, with that said, any commentary on me in the Boston area getting bombarded at the time of this recording with two feet of snow? Why did that happen?

Tony: Well, we're going to see, Dom, more extreme weather, there's going to be times of a lot greater precipitation and probably times of a lot more intense droughts. This is in our future, whether we like it or not. So it's just a question of how well we adapt, you know, and it's one thing to adapt to the death of third-party cookies, it's certainly going to be another thing to adapt to a changing climate.

Dom: Tony Byrne, he can do it all, Rich, he can just do it all, weather obedient. He's the best.

Rich: That is so true. You know, the climate market is apparently really heating up now. No, poor joke there. But, you know, I think one thing we can all agree upon is that it's hard to get to the real answer if everybody isn't working on the same set of facts and information, which I think, Dom, brings us up to your next question.

Dom: It does. Yeah, we just wanted to set the stage here, Tony. By the way, thanks for sharing that personal endeavor, it's definitely an inspiring one for sure. So let's get back to the business of CDPs, customer data platforms. What would be your definition? There are a lot of definitions floating out there. Do you have, like, an overall definition of one? Or is it just too many segments and categories to pigeonhole this into one definition? 

Tony: Yeah, it is difficult because the scope of CDPs' is expanding and what people call a CDP is expanding. I think, you know, if I was going to take a stab at it, it would be a business layer where marketers' DX and CX people can get access, easy access, for activating authoritative customer data, right? So, it's a place where I can go and get the data I need as a marketing CX or DX person to then make that data actionable in some customer engagement environment.

Can CDPs do a lot more than that? Yes. Do they kind of have to meet that bar? And do that? I would say 95% of the cases you do, occasionally you see a CDP deployed that isn't actually a business platform. So I would say that my definition is maybe 95% accurate.

Rich: As you mentioned, it seems like this is an area that just continues to have evolution throughout it. Where are you seeing the new growth in the CDP categories and usage? Like, where are you seeing as the newest categories?

Tony: Yeah, it's interesting. I think we ought to just level check that many organizations, including those that have licensed a CDP for some time, are still in their kind of freshman or sophomore year with this stuff. And a lot of CDP implementations are still very much incomplete. And so the first thing to understand is that the technology is maybe a little bit ahead of the marketplace.

But if I look at areas of growth, I think one of the things that we've seen is that decisioning wants to be close to data so that decisioning could be next best action, it could be applying rules or models, it could be journey orchestration, it could — certainly personalization is a key subset of decisioning. And we know that, first of all, all of those things need good data, or they're not gonna work. So that's one thing driving CDPs.

But also, we're seeing that, the best way I can describe it is that those decisioning services seem to want to be close to the data and very closely intertwined with the data. Many machine learning algorithms, their outputs are data elements. And so we're seeing some sort of, I won't say convergence, because they're different categories, but we're seeing particularly CDP vendors getting more involved in decisioning and orchestration and vice versa. And so that's been interesting to observe.

Dom: Tony, you know, do these organizations ultimately need a CDP to pull off any type of real personalization at scale? Because organizations did this before CDPs, too. So I guess what I'm getting at is, like, what wasn't there before the emergence of CDPs for brands? Where were they struggling? And how do CDPs solve the problem?

Tony: Well, I think for a long time, many organizations had one, or in some cases many, of what I might describe as a marketing data store. And so your personalization subsystem, which had been around and predates CDPs, probably had its own marketing data store. And then your email platform had its own marketing data store. In some cases, your website, or certainly your ecommerce platform, had its own marketing data store.

So each one of these narrow channel applications had their own view of the customer. What they didn't always have, unless you built it yourself, and some companies did, was this unified view of the customer. And one of the things that you can perform is much more powerful and more accurate and maybe more sensitive personalization if I've access to the full panoply of attributes about this person, as opposed to just a narrow view into their profile.

Rich: Tony, I wanted to go back to what you said a moment ago when you talked about data wanting to be close to decisioning. We're gearing up for own DX Summit, and I met with RJ Agarwal, who talks about AI and prediction machines. I'm curious to know how important AI and machine learning are to this specific CDP platform as we talk about scaling and millions of potential customers and all this.

Tony: Yeah, so CDP vendors will tell you that it's very important. I tend to think in our experience that that's really demo candy for them. Predictive analytics is one of the 10 business use case categories that we evaluate in a CDP. Fewer and fewer of our large enterprise clients are looking at predictive analytics at that tier because, for predictive analytics to be really successful, or any kind of ML and AI, you really need a very large data set, you need, you know, data lake type data here. And typically, that's not what your CDP is. Typically, your CDP is a subset of data. And it may not even have the full view of the customer because not all that data is needed for the use cases that CDP is trying to fulfill.

So we tend to see the more sophisticated enterprises building their AI and ML operations, some people just shorthand to MLOps, kind of lower within their infrastructure in their data organizations. And then the outputs of those, whether it's models or data or rules, then could be reflected in the CDP or directly within the different applications. So, the AI that some CDP vendors bundle can look really sexy in a demo, but you have to ask yourself, is that the layer that I really want to be doing this? And I think the larger the enterprise, the more the answer to that is no.

Rich: And what layer would you push that down into, then?

Tony: I think that it's a complicated topic. There are many stakeholders for this data, and these outputs, and this analysis and decisioning, more than just marketing, and so we typically see this as more of an enterprise-wide function and not just a marketing function. And the best organizations have an MLOps organization with endpoints and a set of services that they can bring forward, including to a marketing team.

Now, having said that, that's an idealized view of the world. Many CDP vendors have benefited from the fact that sometimes marketers don't have access to these enterprise services. So the only way they're going to get them is if they're baked into the CDP. And that's a legitimate use case, too. It just can lead to a little bit of architectural messiness, and potentially, the AI working on a narrower set of data, which we know is not always ideal.

Dom: You know, Rich, if I'm a marketer listening to this, I'm thinking, "MLOps team, that sounds super cool, I'm gonna pitch that to the higher-ups." But on a serious note, Tony, I mean, it's a great point, you know, in terms of the differences between, "oh, we provide a..." versus "no, we have an actual internal enterprise MLOps team." Have you worked with some of those teams and, kind of, who composes them? Because I think organizations listening will be like, "Wow, we need to get this done."

Tony: You know, and in many cases that may already exist, it could be in its formative stages, and they may not be looking at marketing as their initial constituent. And that's the complicating factor, right, is that when marketing needs enterprise services, they have to elevate their turn in the queue, right? For everyone else who's demanding this sort of, you know — sales is a big consumer of predictive analytics, so is customer service, we know, even finance is interested in MLR.

So there's a lot of people knocking on their door. And if you're last in line, I can see where maybe you might get a simpler version of some algorithmic predictive function in your hot little hands from some other vendors. The problem is, as you know, that every vendor is bundling these sorts of capabilities. And I think it's very dangerous for companies to have a cacophony of these activities at a time when customers and regulators are really looking for transparency and coherence in the way that you're using these algorithms. So that's another reason I think you'll see these increasingly centralized.

Rich: In our own CDP Market Guide, some of the use cases that we looked at are gaining a 360-degree customer view, as you mentioned, gaining actionable insights was another popular answer, increasing customer acquisition and engagement via personalization, which I think is a priority for most organizations at this point. What do you see out there in terms of practical use cases for CDPs that marketers are finding success with today?

Tony: Yeah. So, you know, we look at 10 of them, I'm gonna isolate two or three in particular that seemed to be really easy and powerful wins. One is providing better support to your outbound marketing campaign teams, the people who are driving email marketing, SMS, integrated SMS telephony, you know, whatever. The outbound marketing team, many of whom are stuck working with the data and segmentation interfaces within their email or marketing automation platforms where they're notoriously bad, whereas a CDP can theoretically get access to a much wider set of attributes and have you build much more sophisticated and targeted interactions that are a little bit closer to one-to-one rather than these kinds of mass batch-and-blast type things that you could do in MailChimp and Marketo and others. So that's one win.

The other area that is increasingly coming to force around support for digital advertising and media investment, and with the deprecation of third-party cookies, the criticality of having a good first-party data set and leveraging that. And so, what's an accessible place to have that, and do that and activate that? And it turns out a CDP can be very handy in that regard. There are some emerging opportunities around loyalty and rewards management. You can certainly supplement some of the weaker capabilities that you may have around ecommerce recommendations and optimization. So, there's no lack of use cases. It's a question of which ones are really high priority for you.

Rich: Yeah, Tony, we've been talking a lot on CMSWire, and I'm sure you've seen it too. Google keeps flip-flopping on cookies and what's coming next. But I'm curious to know, you know, how the shift away from third-party data and cookies is going to impact CDP usage if it does at all?

Tony: Well, I think in general, we've done some research on this, what we call the new world for data. And we came out with nine different adaptation strategies, and one of those that is a little bit like "Mom, Pop and apple pie" is, you know, lean more on your first-party data. And that assumes that your first-party data house is actually organized and accessible, which a CDP can help you do that.

It does have some impact, obviously, around anonymous records, and most CDPs will store anonymous visitor data to your website, and then ultimately stitch that into a known user when you get that person to raise their hand. And there are some complications there around how you set up cookies or other forms of listening in that respect. But in general, I think CDP vendors quite properly see this development as yet another driver for their existence.

Dom: Absolutely. Tony, looking at some of the implementation realities now of CDPs, and how it sort of fits into the existing martech stack. Let's say if an organization is exploring this and goes in clean, they have so many problems with customer data, and they just have decided to venture into the CDP world. What are some of those, you know, implementation realities they're going to face?

Tony: Yeah, I mean, like a lot of martech tools, the CDPs are never ever done. People keep adding more data and more use cases, but presumably, they're adding value to it. So the first thing you need to do is take a real product management approach, as opposed to just project and program management. You will have projects and programs, but you need to have a real product orientation to your CDP to figure out just which use cases are you going to fulfill, and what order and what data you're going to need to bring that in.

I think the other kind of dirty secret of the business is that most CDPs tend to assume that your data house is somewhat in order underneath it. And in many cases, it's not. And that can cause a significant delay. And there are some critical issues around, for example, identity resolution that many CDP vendors, I would say most CDP vendors, would tend to shy away from and have you deal with at a different place in your stack.

So this whole notion that a CDP is going to get your data house together — it really isn't. To a large extent, you typically want to layer a CDP on top of what we might call some sort of virtualized customer data hub where you've worked out a lot of issues around identity resolution, data stewardship, data cleanliness, data governance and these sorts of things. And the CDP is then just simply providing a window into that data that marketers and CX and DX people can actually get a handle on because you can't have them interact directly with a data lake, right? But they can work with a CDP.

Dom: Now, another promise that CDP vendors have made, Tony, for a long time is the term, and you know vendors love this term, marketer-friendly. Your thoughts on that? Is a CDP marketer-friendly? In other words, even a marketer can use this.

Tony: It's a really interesting question, you know, what does user-friendly mean in any context in the martech world. It's been one of the more challenging dimensions, certainly for us, in the last 20 years as vendor evaluators. I think that there are some CDP vendors that are definitely focused more on marketer personas and developing, kind of, simpler systems for them. And there are others that are more infrastructure-oriented, and their systems tend to be more complicated, maybe more powerful. And the persona who's using it might be more of a data person, a DataOps person, as opposed to a MarketingOps person. So you definitely need to test these things out.

I think one of the areas where the rubber really hits the road is around the segmentation interface. And one of the things we always do when we're leading an enterprise to CDP selection is we very early on, right away in the first demo, we have someone from the enterprise, not the vendor, actually building the segment. So he said, "Okay, here's the segment we want to build, we want to have a new promotion for people at their half birthday, okay? So we need this and this, or that and this," and we literally start getting hands-on, building the segment ourselves, and you find out just how difficult, or I'll use the segment-building interfaces, and it's hard to tell when someone else is driving the car, but when you're driving the car, it becomes very clear.

So there actually are a lot of key considerations there on marketer friendliness, and it starts with segmentation, but then it goes to activating segments across different channels, sending something to Facebook, is it something that I can do myself? Or do I need to go to a data team to do that? The interesting development though, Dom, is that we are seeing greater sophistication on the marketing side and more of these hybrid Marketing-DataOps people who can think like a marketer but act like a data person, and the most powerful CVPs are really built for them.

Dom: How come everyone assumes that marketers don't understand technology? When's the day when we're gonna say something like, "Hey, this tool is IT-friendly. You might need a marketer's help here and there, IT person, but it's IT-friendly, you guys can figure this out."

Rich: It has, yeah. Technology has permeated that job completely. I think we could have an entire show, Dom, and maybe Tony can come back for this, where we talk about all the buzzwords that we don't like, marketing suddenly being one of them.

Tony: Yeah. But the usability based on the personas that you anticipate, that is important. We actually do find that sometimes a CDP can be a bridge too far for some people in a marketing team, and so they designate like maybe four or five people who are going to work in the CDP. And they're the ones who handle the work for everybody. We have seen that for sure.

Dom: Alright, let's get to the goods now, Tony. Tell us, like, a horror story about CDP implementation. Is something top of mind that just went awry? They had good intentions, it just didn't work out? Well, on the other side, too, something like, "Wow, these guys crushed it." So, give me the low end and the high end. What do you got?

Tony: There's a couple of low ends. I mean, there are fewer now, but certainly, in the early years, there were CDPs that just never went live. And usually, it was a massive mismatch between the vendor and the customer in terms of roles or use cases. But there have been some major quiet failures or cases where a company has literally gone out and bought three different CDPs for three different use cases, which ideally you should never have to do.

One of the big ones is that many organizations have underestimated whether you can support more kind of group-tech objects, like household and business, and not all CDPs can do that. So we've definitely seen, like, day one, okay, we forgot to ask you, you know, we work primarily through households, and they say, "Ah yeah, we don't really support households natively, but there's this workaround." And, you know, it falls over. So that's happened.

On the success side, I'd love to tell you that there was some trick formula to this. But the reality is, it's a lot of hard work, and your hero Tom Brady-type behavior, where you're eating the right foods and exercising the right way and, you know, able to really sustain this for a long time. Because the organizations that have done this right have really fit the CDP into a broader customer data fabric, and they're very clear about the role it's going to play. They've done a multi-dimensional team-based selection process, and they're just ready to go. And what's great about those use cases is, sometimes, those companies can start seeing value within three to four months, like tangible value.

Rich: Anecdotally, from what I've heard from the sources that I've interviewed, is that a lot of organizations just head into this evolution, and you mentioned it, is that their data house isn't in order. And now they're just trying to add this other layer on top. And I imagine that that's just got to be an incredibly challenging and frustrating situation to be in.

Tony: Yeah, it's deeply frustrating, particularly if you made promises around what the CDP can do, and some of them, but not all of them, will help you with customer data management. Those tend to be the more powerful, less marketer-friendly, if I can use that term, you know, and can help you at least provide the tools to help you put your data house in order. But you know, you got to decide architecturally, is that really what I want from a CDP? In many cases, the answer is no.

Dom: All right, Tony, we're gonna put you to the test here, your wit, your ability to be quick. We're gonna play a little word and phrase association game, we're just going to throw something out there and you give us, like, a one or a couple word answer. Are you ready?

Tony: Yeah.

Dom: Okay. Digital experience platforms.

Tony: Oh. Bull**** term.

Dom: Ooh. Okay. Tell us how you really feel. Alright.

Tony: They don't exist. It doesn't exist. Digital experience platforms don't exist.

Rich: I told you, Dom. This should have been the podcast.

Dom: Well, we got to get someone, because now we have two authorities saying they're not real. Tom Wentworth's earlier podcast, now Tony Byrne. I think he actually mentioned Tony in that podcast. So there's a connection here.

Rich: I feel a panel coming on.

Dom: That's it, we need a DXP vendor, Rich, we need the DXP vendor to tell us how cool they are. We need some balance here. All right. All right. Thank you, Tony, for that. Moving on. Aaron Rodgers.

Tony: I guess I have to say something different this time. I think he's presumptuous. He thinks because he's a great thrower of the football — I won't call him a great quarterback, he's a great thrower of the football — that he also knows about science, and he doesn't. And so, we all see that in the professional world as well. And one of the things that I've learned as an analyst is that it's a very humbling job, you think you've got theories and frameworks and opinions about how the world works. And then you see it in reality, and it can be quite different. But I do respect science, and I'm disappointed that he doesn't.

Dom: Alright.

Rich: Yeah, I would say for me, that would be mixed emotions. Aaron Rodgers.

Tony: Thank you. That was a much easier way. I need to have you be my media relations guy.

Dom: Okay, couple more. Vendor press releases.

Tony: Forgettable.

Dom: Good one. Finally, old school here, web content management systems.

Tony: Still relevant, still need them. Yeah.

Dom: That's what you grew up with in your professional analyst role, right? Pretty much? Was that the main gig, analyzing the WCM space for you?

Tony: Twenty years ago when I started out, yeah, and we still do, that's one of our major research streams, and the world has changed, they have changed a little, they've changed a lot, and they're still germane. You just need to really be clear about where they fit in your martech stack. It's all about the stack and the rationality and the coherence and the modernity of your stack these days. And so you have to have a stack strategy first, and then pick individual vendors later.

Rich: That's interesting. Thank you for sharing that. Tony, that wraps up today's podcast, we really can't thank you enough for coming on to CX Decoded. We always like to give our guests the opportunity to share where the audience can find you and follow your leadership.

Tony: Sure, just or find me on LinkedIn.

Rich: All right. Thanks again, Tony. Thank you. And thanks to all the audience for joining us on CX Decoded. And we'll see you next time.

Dom: Later.

Tony: Super, really enjoyed it. Have a good day, guys.