Zeta Global Holdings Corp. (ZETA) Earnings Call Transcript & Summary

Good afternoon. Thank you, everyone, for joining us. My name is Elizabeth Porter. I'm an analyst on the U.S. Software Equity Research team. And we are very excited to have the whole slate of Chief Executives from Zeta. So we have CEO, David; CFO, Chris; and Chief Data Officer, Neej. We are going to take audience Q&A. So mics will go around at the end. And before we start, for important disclosures, please see the Morgan Stanley research disclosure website at www.morganstanley.com/researchdisclosures. And with that, thank you all for joining us today.

Thank you, Elizabeth. We love being here.

Thank you.

Great. So at first, just to kick it off, Zeta sits in a very unique position in the nexus of marketing and advertising ecosystems. And so for those that are less familiar with the story, it'd be great to just get an overview of the business. What's the problem that you guys are trying to solve for clients when you founded the company? And what are some of the primary products in the platform today? And where do they really fit within the competitive landscape?

Well, that's a big question to start with, so thank you. Our main goal is to help very large enterprises to create, maintain and monetize customers at a substantially lower cost by using our software and our data. It's pretty simple. So we have 3 main use cases helping enterprises to create, maintain and monetize customers. We do it through using one fully integrated platform. So the original vision was to disintermediate multiple point solutions as it related to how marketers can operate. So we have one user interface, one reporting infrastructure by using our data and our software around marketing automation, where we can do everything from CRM to programmatic, connected television, social, mobile, display, we build CDPs, everything sits in a centralized hub with one user interface and one reporting infrastructure for the enterprise.

Great. And the company reported earnings last week, growing 30% organically, and that just excludes spend from some of the political cycles, from M&A. And you hit $1 billion in annual revenue about a year ahead of your initial plan that you had laid out in 2022. Would be great to get your key observations from the print and what you're hearing in conversations with shareholders coming out of earnings?

Well, I'll start, and then I'll let it go to you because you're hearing more from shareholders than I am. I just want to say how incredibly proud I am of the team hitting the revenue component of our 2025 plan a year early. We actually rolled the 2025 plan out in 2020.

Oh 2020.

No, no, no. We were private at the time. So you're exactly right. It was an internal plan, and then Chris joined us as our CFO and made it an external plan. But incredibly proud of the team. I think it really speaks to the quality of our products and the quality of our services and capabilities for our clients that we were able to scale as fast as we did.

Yes. I think investor feedback on the print was stellar. As you'd expect, focus on the 2025 guide, which from our first long-term model to the current newly announced long-term model, we've been consistent. We see ourselves as an at least 20% grower. So walking folks through what that 20% looks like, what assumptions we built into it, where we think there's conservatism. We've had feedback on, call it, the nonoperational sides of the business, but areas that we also acknowledge we want to address and improve, namely around dilution and stock-based compensation. It's our view that we can significantly step down dilution from 2024 to 2025, getting to our long-term target of 3% to 4% dilution in 2026. In the course of doing so, we think we can evolve how senior management is compensated in terms of tying significant share price appreciation to even when equity awards begin to be issued, so think somewhere in the 30s and then opportunities to ladder up into the 40s and the 50s. And speaking of stock, we look at today's share price and the assets Dave and I were talking to one another about Zeta's -- assets of Zeta's scale, so call it $1 billion in revenue, growing at a 4-year CAGR of 30% or a forward guide of at least 20% with $130 million of free cash flow, there simply is no better investment in using our existing $100 million share repurchase program very aggressively right now.

So we're going to step up the share buyback. We -- when we look at the stock in the marketplace, as Chris said, there's no better use of our cash -- balance sheet. And I think for the first time, we're going to go into our balance sheet to start buying. We had always said we would use half of free cash flow. We used 100% of free cash flow in the fourth quarter. We're going to step that up just because I don't see any better way of using our capital right now than buying back stock.

Great. Okay. And before we get into more of the financials of the business, which I certainly want to dig into, I want to start off with a bigger kind of picture question of you've talked a lot about these replacement cycles happening at customers as centralized data just becomes more and more important. So where are customers coming from? And what are they going to when they replace? And where are we within this replacement cycle?

I think we're just getting started. So if you think about the -- I always joke, we compete with a few small companies, right, whether it's Salesforce, Oracle, Adobe, -- the Trade Desk. Most of those guys, if you look at ExactTarget, Responses and Neolane, which are the companies that were bought to create the marketing clouds, of Salesforce, Oracle and Adobe, they cannot move to a new AI platform. I know that's all they talk about. But the truth of the matter is they didn't throw out their old architecture. So if you've got a legacy Marketing Cloud, you've got to step out of the Marketing Cloud to an algorithm to ask a question. That algorithm has to do a data dip, go back to the algorithm to create intelligence and tell the Marketing Cloud what to do. We exist in a world where a millisecond matters from a return on investment perspective. So at Zeta, we made the decision in 2017 to start architecting our platform around artificial intelligence and data. In fact, as you know, Elizabeth, because you were on our IPO, the day we went public, our sign on the side of the New York Stock Exchange said Data plus AI equals Intent. Now nobody had any idea what we meant back then, but I'm glad we did it because we can prove we have the pictures. But we completely rearchitected our platform, putting Data and AI as foundational. So as those other assets hit 10 years since they were purchased by the big technology holding companies, we're starting to see the replacement cycle on marketing clouds accelerate. So we saw our pipeline up 60% in the fourth quarter. We saw RFPs up 40%. And we're seeing clients move at an accelerated pace.

And with that move, you've also talked about your opportunity to expand the addressable wallet. Right now, you're addressing about 1% of the available marketing and advertising.

For our existing customers.

For your existing customers and an opportunity to clearly kind of move that up. So what are the strategic advantages that you have right now to accelerate that $2 million average spend on the platform right now, isn't an insignificant amount. So how are you kind of moving that forward? And what gets customers to fundamentally shift more of their spend to Zeta?

Let me start by talking about One Zeta. So if you think about our platform, we've got 3 use cases: acquisition, retention, monetization and then you have multiple channels. Channel is probably 14, which is activating through all the different methodologies. It took us 16 years to get to our first $1 billion annual revenue, and we expect to get to our next $1 billion in the next 4 years. A lot of that is coming from what we call One Zeta, which is clients we've been able to get from 1% to over 5% of their existing wallet share by launching with all 3 use cases at once, customer acquisition, customer retention, customer monetization. And it's really beginning a very interesting flywheel where when you know how to acquire a customer, you know better how to save them. And when you know better how to save them, you know how to better acquire them and it gets smarter. So today, our existing 527 scaled customers spend $100 billion a year on marketing. We have 1% of wallet share. Our goal over the next 4 years is to get to 2%, but our long-term goal is to get to 5% to 10%. We brought in a gentleman named Ed See. Ed was running the Chief Marketing Officer practice at McKinsey to run the One Zeta platform, which getting Ed to join us was really a big accomplishment for us. He's a legend in the industry. But the flywheel is what we see is when clients use One Zeta, the efficiency rate goes up, they save more money and the revenue growth goes up, they generate more money. Chris, did you want to add to that?

What's interesting is you think about the One Zeta, but the magnitude of the opportunity, up to this point, most of Zeta's ARPU expansion has come from the hands of what we call upselling, which is adding channels over time. When we first announced our initial long-term model, we had about 1.2 channels [ per. ] Now we're knocking on the door of 3. And that will continue in the next 4-year model. But the bigger opportunity and where there's, call it, 3x to 5x more revenue leverage is when you add a use case versus just simply adding a number of channels. The opportunity set looks like we have 527 total scaled customers, less than 15% of those scaled customers use more than one use case today. Now each use case has hundreds of millions in scale. Each use case is growing at a rate equivalent to total Zeta, but we haven't yet cracked the code that we're now on to sell a platform sale with all 3 use cases. That is where the need of the market is anyway.

And Neej, you're talking to all these people. What are you seeing as it relates to that?

Yes. So I think there's really 2 important motions that I see all the time. So the first is if you enter the Zeta door through the CRM lane, meaning that you want a CDP or a marketing automation stack, very early in the sale process, we're going to expose you to the media offerings we have. And this is a huge differentiator, right? Total -- lower total cost of ownership, higher impact of revenue. These are both things you get -- that you get from Zeta and you can't get otherwise. If you enter our ecosystem as a media customer, all of a sudden, you start spending money with us and you start realizing we can solve problems for you like churn or customer data management or how to take a customer that's buying once and convert them to a customer that's buying twice. So that leverage we can create upfront and over time is really what's fueling One Zeta, and that is also being compounded by the replacement cycle that Elizabeth mentioned.

And the 15% of customers that are only on one use case today, what has been the friction for that number not to be higher? Is it, "Hey, I already have a stack and so that's replacement or different buyer personas?"

Yes. I mean, once again, in the 16 years, we grew from 0 revenue to $1 billion a year. We're very compartmentalized as a business. We have our CRM division. We have our Activation division. We have our ZX division, so on and so forth. And we would sell in one way and then we would try to cross-sell. The re-architecture is with Ed as our Chief Growth Officer, it now we lead with it. And it's really interesting because in many ways, it's -- we're seeing a better take rate on the One Zeta. To put it in perspective, our goal over the next few years, not next few months, is to get 25 One Zeta customers spending $100 million each. which would give us material growth over our existing businesses $1 billion revenue. That's going to take us a number of years to get to that. And they scale, like they might start at $10 million, $15 million, $20 million and then scale from there. But the reality is that, yes, they all have an existing stack. But at the same time, what we found is when we can go in with all 3 at once, it's resonating. It's really been working.

Great. And another place for the business that [ has ] been working particularly well has been agency and holdcos. And so just beyond -- and that really exposes the customer base beyond just your traditional enterprise brands. So first, just why do the agency-holdcos want to work with Zeta? What use cases are they landing with? And how do they differ in terms of their spend versus the traditional customer?

Well, let me start with -- when we first decided to go into the agency business, I joke marketing has started with Mad Men, now it's Mad Science. And we made the decision to power the mad men, so to speak, versus competing with them. And it's worked very, very well for us. We started, as Sun Tzu says, fight where your enemy is not. We started with an automated social media platform that allowed us to help agencies to seamlessly buy social media marketing in a very, very cost efficient and very elegant way using our data and our technology and our software, which nobody else, none of the legacy DSPs were working inside of Meta. So our matching the Zeta ID to the Meta ID was a big differentiator for us. Second, as we looked at the legacy DSPs, it's funny because if you look at their business models, they say that they're charging 7% of spend to use their platform, yet their take rate is 19.5%. The reason for that is they're charging for every $100 the agency was running in marketing, $25 went to data purchasing. 50% of the data purchase would go to the legacy DSP, 50% of the data purchase would go to the data provider. Because we own 100% of our own data, we're bundling the data in for free to the agency-holdcos. Now they're also telling us our data is superior because it's deterministic first-party data versus probabilistic third-party data. But what most of our agency clients are doing is keeping a percentage of that for themselves and more than doubling their own margins, and they're giving a percentage of it back to their enterprise customer in the form of a meaningful discount. So not only are we able to drive accelerated growth with superior data and superior technology, but we've been told by most of the agency-holdcos we work with, we are by far the most profitable partner they work with as a company.

Great. And you also mentioned targeting this long tail of 1,000 independent agencies. How do you actually go after this? I would imagine this long tail of customers, potential customers. So what's your strategy for actually penetrating that market segment? And how do you see the mix between your traditional enterprise brands and agencies evolving over the next couple of years?

I think the agencies will continue to be a bigger and bigger piece of the pie. What the independent agencies don't have that the holdcos do is a vast amount of capital spend on infrastructure. So the independent agencies, and maybe Neej, you've got a lot of experience in selling to them. What they're looking for Zeta towards is how do you become my AI and data backbone. And in a trade for that, I'm willing to contract over a multiyear period for contractual minimum direct use. It doesn't mean we still can't do social, but it becomes a much more strategic relationship, longer-term relationship. Maybe I'll pass it to you as to how we close the deal.

Yes. The independent agencies are looking for technology that looks a lot like what a brand would look for. So they want to power use cases across the customer life cycle, acquire, grow, retain. They want data, they want white label, and they want to look really smart in front of their customers. So for us, it is more emblematic of how we would sell to a brand. They want the platform. They want to create parent-child relationships to launch sub accounts for their brands. And we've been very successful in getting those types of agencies to not just adopt but also to scale, and they're in there every day using the platform to look smart, to generate intelligence, campaign ideas, and we see more and more of those actually going into CRM use cases as well.

And they're winning deals on partnering with us. And to call them small to midsized agencies, to be clear, these guys still have multibillion-dollar spend. One of the other things I think is important to note as you -- to answer the last part of your question is the agency-holdco business is at a slightly lower gross margin, but it's at a very, very high-quality contribution margin to our business. So in the same period of time in which that went from, call it, a de minimis percentage of our revenue to 20% of our revenue at the last -- the end of last year, we were able to meaningfully grow our operating margin in that same period of time because the cost of managing it is lower. All we're really paying out are sales commissions.

Great. And I want to switch a little bit to kind of the data aspect and how it was built. You guys have clearly talked about a lot of the advantages that people get when they use your data, that's why they're coming to you. But just take a step back and help us understand how the data was compiled? And then you've also done some technology tuck-ins for data, including LiveIntent. So it would be great just to get an update on that partnership as well and what you're hearing from customers.

So let me start by saying that one of the decisions we've made over the years was to buy companies that generated meaningful data every moment of every day. We've never bought companies that had very big databases because the truth of the matter is they get very old very quickly. So when we talk about having 245 million active Americans in our data cloud, that's the number of people who have interacted with us in the last 90 to 120 days. This is not a legacy number. It's consistently refreshing and consistently stepping up. And some of those assets were discussed, DISQUS, which is a very large commenting platform, our DSP, we own one of the largest ESPs. And then, of course, we just bought LiveIntent. We've been licensing data from LiveIntent. So buying them made a tremendous amount of sense for us. We also announced, as you know, Elizabeth, that we've already fully integrated LiveIntent into the Data Cloud and vice versa. If you think of our business, we had traditionally worked with the long tail, call it, 5 million publishers, but the long tail, whereas LiveIntent works with the ComScore 1000. So what they were doing was inserting an ad from a DSP perspective into messages from Tier 1 publishers and retailers to consumers. If you were getting a New York Times Wordle e-mail, you would get an ad placed in by LiveIntent. What we did was by merging our data cloud before we bought it, everybody would get the same ad based on the content they were consuming. Today, the ad is targeted based on deterministically who you are and what you're consuming. So it's already for the clients we've moved over meaningfully driven up revenue yield to our marketers and meaningfully driven up revenue to the publisher, we take a percentage on each side. So that business is growing quite quickly. Neej, do you want to add to that?

Yes. I think the main thing is that we think of data in 3 dimensions. We think of identity data, which represents the unit of a person. We think about signal data, what is a person trying to do next. And then we think about identifier data. If you wanted to reach them in an omnichannel way, how would you do so. So to be a really good data cloud, you need to have all 3 things happening persistently and being refreshed persistently. The companies that do this the best are the walled gardens, right? You think about Facebook and Google and Amazon. So we have those exact kind of dynamics, but we bring those benefits to the omnichannel world. And that's a one-of-one value proposition. There's no other company in the world that provides that kind of outcome and capability set to marketers. So it's a very powerful combination. LiveIntent was a great addition to the stack, and we'll continue to be opportunistic in the data space. But our 245 million Americans that we see, we see very regularly, and it's a very -- it's a mature data ecosystem at this point.

Great. I'm going to ask another question, and then we will have the microphone go around. But I wanted to get on to the 2028 long-term targets that you guys announced just the other week. So can you walk us through the levers of your 2028 plan, where you're looking to make incremental investments? And what are the areas that you feel like you have the most confidence in?

Sure. I'll start off, and David will comment around the investment areas. So the first Zeta long-term plan, which was at least $1 billion in revenue, $200 million EBITDA, $110 million of free cash flow by 2025, we laid out a set of levers, which we've kept consistent from one model to the next, which is how fast we want to grow scaled customers and count, how big we want to grow their spend with us, the net revenue retention rate of our business, how many quota carriers we want to have and then the mix of our business. We've kept all of those consistent. So those levers being we think that the right way to set customer growth in the next model is anywhere between 4% and 8% -- that compares to 8% to 12% in the prior model, but the only distinction being that as we see this rapid growth opportunity with agencies, we'll sign one agency, whether it's a holdco or an independent agency. We only count it as one customer, even though there's probably dozens and dozens of brands that we're working with. So we've built in some give for that. But brand count should grow what our historical customer count has been, which has been mid-teens. We've increased the ARPU growth rate from one model to the next from 8% to 12% in the old model to 12% to 16%. And this is reflective of continued progress in adding channels, the unlock that we think -- One Zeta yields and then continue to sign customers with a bigger average starting price.

And when you think about investment long term, obviously, we would start with AI, right? And whereas it's interesting, most of our competitors who are trying to catch up are now paying, in some cases, 3,000% more for chipsets that we were buying a few years ago at a very low price. So it's been a competitive advantage for us that we bought a lot of this technology and built a lot of this infrastructure over the years instead of needing to catch up and do it all at once. In fact, our percentage of capital investment to revenue is down dramatically over the last few years, which has led to better free cash flow. Although I want to be clear, this year, we will spend more money on innovation on an absolute dollars perspective than in any year in our corporate history. So it's good that you're growing the business so fast that you can do those types of things. But we're going to continue to invest in our technology. We're going to continue to invest in people, bringing Ed See in as our Chief Growth Officer; bringing Pam in, who was running Oracle's Marketing Cloud to now run our CRM division. These are both major, major players in our industry. And we're feeling like the team is what we need right now to execute on the plan over the next few years.

Great. Do we have any questions in the audience? So I wanted to go back to some of the product initiatives that you guys have. And 2 key ones were mobile and AI with Intelligent Composer. As you mentioned, AI and automation have long been central to the Zeta business model. So how are you making AI actually more actionable with the Intelligent Agent Composer. And you mentioned you're on kind of version 3 of your AI versus many are still launching their products.

You are taking my lines here.

No, you gave them to me in the transcript. What are the key learnings from kind of your earlier versions? And how has that informed your current offering?

So we believe Agentic AI is really the future. I think a lot of people, as you know, are talking about that. We're on version 3 of our generative AI agent studio. To put it in perspective, if you look at last year, we grew our consumption revenue as a company by 40%. A lot of that can be tied back to the use of artificial intelligence. We had 127 of our clients adopt agents inside of their platform. They use the AI Agent Studio to build it. We saw AI usage go up 200% sequentially from Q3 into Q4, and we doubled our publicly available taxonomies using artificial intelligence in that same period. So rather than charging $19.99 a seat, what we've seen is the utilization, the consumption skyrockets when clients adopt this. And it's primarily today, we have a lot of use cases. But I'd say the most popular one is -- our platform is called ZOE, Zeta Operating Engine. You can say -- a marketer can say, ZOE, what are my most valuable underserved audiences? And what used to take 3 data scientists to $300,000 a year each, 3 weeks to do, ZOE can do in seconds. And then would you like to activate to those? It just becomes such an easy, would you like fries with that. And it's been a really powerful offering and really helped fuel our growth last year. Do you want to add to that, Chris?

I think that the one thing I'd add is, so David is talking about prescience, which is the ability to take data interpretation and convert it through an agentic framework to be easier. You can look at a chart or a report, understand what you should do next. We've also seen a lot of gains in productivity. So I have these 1,000 e-mail templates in this other system, I need to convert them to a Zeta system. Great. This would have taken hundreds of hours of man hours to do before. Now the agents can do it virtually instantaneously. And then the promised land is going to be personalization, which is the last P here, which is where you can deliver truly one-to-one experiences, and we've already started to do that in some channels and some infrastructure technology across the web will need to catch up, but that's going to be where we're going to be heading next as well.

Great. And when we think about the guidance for revenue growth, it was about 21% when you back out some of the things like LiveIntent M&A and the political cycles. So within that kind of 21% growth rate, what does the guidance actually contemplate in terms of some of these newer initiatives, whether it's the AI or mobile opportunities?

A couple of different ways to think about it. And I say this is a bit tongue in cheek, but it's interesting. We start to have to explain why 20% isn't great. And I think there's only literally only 27 companies out of the 512 that are publicly traded this year that are expected to grow 20% or more and expand free cash flow. So I make that point. You would think about the 20% as follows. So as you point out, we ended last year at a 30% growth rate ex-Political Candidate and ex-LiveIntent, so on a pure organic basis. This year's guide of 21% makes a few simple assumptions. It could be different, could turn out to be better, and all likelihood will. Last year, insurance coming off a very weak for structural reasons, insurance industry grew 130%. We're not assuming 130% growth in our guide. I don't think anybody would think that would be prudent. Instead, we're assuming growth, call it, in the 20%s. That's about a 4-point change from the 30, so now you're down to 26%. Automotive, which is another very high-growing industry in 2024, grew about 40%, coming off again a secularly weak 2023. We're not going to assume 40% again. Could it? Maybe. We're going to assume something closer to mid-teens to 20s. So you got about a 3-point impact there. And then advocacy, which obviously is higher spend levels in election cycles than nonelection cycles, is in our model is likely going to go from a $36 million amount of spend in 2024 to somewhere around $20 million to $25 million. So we got about 1.5 points there. Now we build our guidance, and this is consistent with how we've approached guidance since we've been public, that we need the low end of our model to get to. So we -- our model is 110% to 115% net revenue retention rate. This would assume the low end of the range. The 4% to 8% customer count or 12% to 16% ARPU growth, the low end of the range. So we'd like to leave ourselves plenty of avenues to get to the number. We haven't had a quarter yet where it's been at the low end of everything. But I do think it's important that we've consistently said we're at least 20% grower; 30% is great. But 20%-25% is great, too.

And let's be clear, right? Elizabeth, you know us for a long time now. We've been public for 14 quarters. 14 quarters in a row, we have beaten our guidance and raised our guidance. But [ Imran ] had a question.

So in your presentation, you talked about 114% revenue retention, which is the high end of the business.

Well, we guide to 110% to 115%. We've always been between 111% and 114%, but we did do 114% last year, yes.

Yes. Last quarter it was 114%, which is best-in-class in the software industry. So can you talk a little bit about your visibility in the revenue, right? It's a software business, so that kind of revenue [ retention. ] Maybe talk a little bit about the visibility of the business.

And that's one of the things I think people have always not understood about our business. Yes, about half of our business is long-term subscription, half our business is utilization. But what people don't understand is we are deeply integrated into our customers. In many cases, we have people in their offices working with them, building forecasts. And most of our revenue is contracted out for at least a year. We can't include it in subscription because it does not have performance leftover indicators, and it's 1 year, not multiple years. But once again, there's a reason we've been able to beat and raise 14 quarters in a row. It's because we have an incredible visibility into our revenue.

Can you just talk about the scalability of the kind of performance-based component of the business or consumption-based part of the business as the business continues to grow off of an even larger and larger base, does it -- is there diminishing returns at all for advertisers if they're all leveraging the same data to advertise?

It's a good question. The answer quite simply is no, simply because the ecosystem is so big, right? You're talking about globally, and this is not what we call our TAM, but globally, it's a $1 trillion annual spend at this point when you look at marketing. So for us, we have been able to scale the utilization component of the business. As I said, last year, it grew at 40%. So I think it will continue to grow. I don't think it will continue to grow at 40%. But what we're finding, especially for enterprises that use the One Zeta platform, that operating the CRM business starts the flywheel for the acquisition business and vice versa. So we're really seeing it scale very, very rapidly. I think that, once again, our goal of getting from 1% of our client spend to 2% of our client spend over the next 4 years. We're not sort of forecasting going from 1% to 10% or 20%, right? It's sort of -- it's a very nice build, but the efficiency really continues to scale.

There's an interesting slide that we have in our earnings supplemental that cohorts out and kind of demonstrates to the question that you're asking around our customers that have been with us less than a year on the platform. Most of our deals start with a proof of concept or a pilot $50,000 to $150,000. Within the first 12 months, we get our customers spend to around $900,000. On that 1-year to 3-year category, they spend around $1.3 million. And then our 3-year plus customers are spending at $2.6 million per. And to Imran's question, 90% of our customer revenue are customers that have been with us over a year. So the ability for us to have very accurate models of how that spend is not just replenished, but grown becomes very reliable.

And by the way, we -- if you look at every dollar spent through the Zeta marketing platform, not only are we generally able to create a 50% cost savings efficiency, we're able to drive an average of $5 to $7 in revenue. There are not a lot of companies that can do that on both sides of the artificial intelligence trade, so to speak.

Another one over here.

I have a question. How do you view the opportunities in the Open Internet space? For example, what's the time share or revenue share of the Open Internet compared to other walled gardens like Meta and Google?

So first of all, we work with Meta and we work with Google. So just to be very, very clear, our platform can target into the walled gardens or it can operate in the open web. We do both. What I would say is the open web is growing exponentially. It's growing faster than even the walled gardens that are growing quickly. And the ability to monetize customers there is becoming more and more important to cost efficiency because a lot of the walled gardens are getting more and more expensive for a lot of these enterprises. There was a question over here for this gentleman in front. So I think the open web continues to be a very large opportunity, and it's one of Zeta's specialties, but we do continue to operate, and I want to be clear again, inside of the walled gardens in addition to the open web, and we can connect individuals from an attribution perspective in both.

I have a technical question, right? So traditionally, like the advertising engine model is basically -- is based on [ RAM ] algorithm right or a mixture of RAM and other algorithm. So have you added a transformer layer to your advertising engine? If so, how many GPUs are there? Have you leased to power that transformer layer?

The answer is yes. I don't know the exact number of GPUs, but I know we already own them. And I know we partner with a number of clouds as well around that. But it is definitely a component of the tech stack that we have added. Our Chief Technology Officer, Chris Monberg, would be certainly better than the 3 of us in answering.

Yes. He can answer a part of it.

Yes. And so -- and just to answer your question, so we have our own LLMs internally at Zeta. But we also enable all the other LLMs for your agent building. So if you think clot is wonderful for something or you think OpenAI is great for something and you want to build prompts within our system with specific LLM models or even Snowflake and Cortex, you can do that directly within the platform. So we give you kind of full flexibility to deploy agents the way you want to, and then they can be strung together as well.

What I mean is apart from the large language model, which you can build agents or whatever, right? So apart from that, like in your advertising engine, right? So do you have that transformer layer [ attention head to ] like improve the correlation?

The answer is yes. We do.

You do, right?

Yes. And if you'd like to follow up on that question, let us connect you with our Chief Technology Officer. But the answer is absolutely positively. Yes, we have bought the GPUs. We've imported them. We have the translator -- we have it all built in. It comes through the client's CDP. So the data exists in there with the models that we've developed on a custom basis with 0 data exhaust out. So that's there. When you get into the deeper level of technological questions, unfortunately, the 3 of us are just not the right people organizationally to answer that question. But if you shoot us an e-mail, we're happy to get you to our CTO.

Perfect.

Any other questions? I think we've run quite a bit over for our slot. So thank you so much for taking the time with us today, and we're looking forward to seeing everything in 2025.

Thank you, Elizabeth.

Thank you.

Thanks.