On this episode of After Earnings, Zeta Global, the AI marketing cloud company that aims to leverage zero latency AI and trillions of customer signals to help enterprises acquire, grow, and retain customers. I'm joined by David Steinberg, co-founder and CEO of Zeta, to discuss how the company walks the fine line between pricing takes, customer relationships, and growth. And we dig into the launch of Athena, Zeta's voice-enabled conversational marketing agent, as well as David's response to the narrative that the LLMs will displace enterprise software. We go through with his passionate defense of why that sell-off doesn't make a ton of sense. And we also talk about Zeta's track record of acquiring 18 companies to date following five pillars of M&A strategy. Let's get into it. We're going to talk about Zeta Global Holdings. This is particularly exciting because you are founder and CEO. And so talking to a public company chief executive who's been there since day zero is particularly interesting for our audience, David. If you don't mind, just break down a little bit the core product offerings of the company. Because whenever we hear data-driven marketing or data-driven go-to-market solutions, there are lots of different ways that can actually manifest itself.
Zeta's primary goal is to help very large enterprises cost-efficiently manage their customer relationship management, customer acquisition, and customer monetization. It's a little bit of a differentiated strategy because our industry, like you talked about, has a ton of point solutions. There's a lot of different companies that do a lot of different things. The ZMP, or Zeta Marketing Platform, is everything a marketer needs in one user interface and one reporting infrastructure. So it allows you to literally create customers in a substantially more efficient way than you can without it. A recent Forrester study showed that for every $1 a marketer spends through the Zeta Marketing Platform, we return 600% in the form of business. So whereas most of our competitors are plugging in from a point solution perspective, by having everything, we have one of the world's largest data clouds, and in 2017, we pivoted our entire business to put artificial intelligence and data as native or foundational to the application layer. So that was a decision that really changed everything for us. It allows for effectively zero latency in using AI for decisioning. When you think about the way AI plugs into most platforms today, you'll have sort of a platform, and then you'll step out of that platform to an algorithm to do a query, and then that algorithm will go to a data ecosystem to draw data to go back to the algorithm to create an answer, and then it'll then inform the platform what to do. That latency in a world where we have to make a decision on behalf of our clients in a millisecond destroys return on investment.
And so here's a question for you, David. That means you were going into AI before it became expression du jour, right? Just help us understand a little bit what is it that you saw, given your background, in terms of AI and the opportunity to apply it here? That's number one. And number two, you roll forward, I call it 7 years or so later, how are you finding the talent in a period of time where there is a war for AI talent to continue building on what you've started?
Well, let me answer your second question first. We started hiring our talent 8 years ago. We started buying our GPUs 8 years ago. So if you look at our business model, last year, 2025, was our third year in a row of 30% compounded growth top line. We grew EBITDA by 44%, but I think one of the most sort of impressive things about our business is last year we grew free cash flow by 78%. In a world where a lot of enterprises are growing revenue, growing EBITDA, but free cash flow is consolidating down because they're spending all of their money with Nvidia and others, we made those investments years ago. So from a forethought perspective, you would not think I would be the AI pioneer in our industry from a background perspective. I'm an economist by training. I'm a serial entrepreneur. I've founded seven companies. I've sold four, taken two public. I'm chairman of another, which I don't operate. I'm just non-executive chairman there. And what I would tell you is we woke up in 2017, and we had been rolling up email service provisioning businesses. We had started sort of in customer acquisition, and we wanted to build what we were calling a lifecycle marketing platform. So it was sort of the germination of what we are today, where all of the companies that were in marketing, and still in marketing, either do CRM or customer acquisition or customer monetization. It's sort of very bifurcated. And we had this vision that was very data-centric at the time, that if you had the data from customer acquisition, you could be smarter about customer retention. And if you had the data about customer retention, you could be smarter about customer monetization, and so on. Started the proverbial flywheel, right? I'll try not to use too much jargon today. When you look at that, we were ingesting so much data in 2017 through the eight different platforms we were operating, which in and of itself was not efficient. So I read 5 to 7 hours a day, and I do what I call triangulation. And I started looking at artificial intelligence. At the time, it was really complex machine learning, but we started looking at it and its ability to process data and reporting in some ways that humans never could. And then we sort of were faced with a really interesting quandary. We had just raised money from a bunch of Canadian pension funds, and the goal was for us to get public rapidly. And the business was making real money. It was a very profitable business, but we made the decision to do two acquisitions. One that would become the core of our data cloud, and one that was a machine — it was effectively a recommendation engine on everything you would read. It would then recommend what articles you could read next, and they were selling it on a SaaS basis to publishers. Now, publishers didn't have the money to buy.
So it was like a brilliant team, in fact, headed up by Chris Momburg and Neeraj Gore, who are still to this day our chief technology officer and president of our data cloud, or chief data officer, I should say. And we bought that company, we shut the business down, and put them full-time on building this new Zeta Marketing Platform, which would use natural language processing to effectively ingest trillions of signals about individuals that then synthesizes down to a deterministic data set. So today we have 552 million people globally who have opted into our data cloud, 242 million in America alone, where we have an average of 5 to 7,000 data points per individual, demographic, psychographic, transactional, and then massive web behaviors, which are ingested through a tracking pixel that now sits on trillions of pages across 5.2 million publishers that have embedded the Zeta Data Cloud tracking pixel into their first-party technology stack. So all of that data feeds into our algorithms. We never sell our data to anybody at any price at any time. The data is simply used for training our algorithms exclusively. So an example of a client we might help is, you know, 51% of the Fortune 100 largest companies in the United States use our platform today, 24% of the Fortune 500. We would take a client's first-party data, we ingest it into what's called a consumer data platform, which we build. It's sort of the next-generation data repository. We then merge our data into it. And we enrich it with thousands and thousands of data elements per person. Now, to protect the consumer, we remove the personally identifiable information, and we replace it with a Zeta ID number. Then the algorithms, which live inside of the consumer data platform, are getting smarter and smarter and smarter on everything a customer did before purchasing. How do we re-architect that? What did they read? What did they search? What transactions did they do? All of these different things create intent-based scores that we then match to the other 552 million people in the data cloud, and we then target them by activating through the Zeta Marketing Platform. So what we're able to do is we're able to get to this 600% return on marketing by doing some very simple things. First, and you remove every existing customer.
But before we go there, David, I just want to touch on something because I'm so glad that you laid out in that detail how Zeta works, and sort of describing the data clean room that you effectively set up so that you can really mine as much, you know, a huge amount of information on a person-by-person basis. I guess the question for you is just to take a big step back for a moment and just to talk about the zeitgeist. What you've described, at least to me, sounds a little bit differentiated because you've been able to articulate the scale of the repository of data you've got and then you've been able to articulate how you mine it. Yet your share price seems to have been suffering year-to-date along with lots of other companies who are seen as part of the software apocalypse. And I think at the moment the market's not particularly discerning. To say there's this little software AI narrative going on right now. Let's talk about it if you don't mind because again what you've described at least to me sounds different. How frustrating is this for you?
Well, you know, listen, I have five children so I'm used to frustration. But what I would tell you is that it is very frustrating. You know, you put up incredible numbers. We've been public for 18 quarters. We've reported 18 quarters as a public company. 18 quarters in a row, we have beaten our guidance and raised our guidance. I'm not sure that's happened before with another public company. We've increased operating margin by 500 basis points over the last few years while growing the business on a compounded growth rate of greater than 30%. Just to give you the math, it took us 16 years to get to a billion dollars a year in revenue. We hit a billion in 2024. Our initial guidance for 2026 is 1.755 billion. That means in the 12 months since we will have closed 2024 and opened 2026, we'll have grown the business top line by 75%.
Successful businesses don't spin their wheels. They evolve. And in order to keep up, you need a bank relationship that evolves with you. Every business is inherently different and that's why with KeyBank, you'll never receive one-size-fits-all solutions. Their expansive team of experts across industry verticals develop bespoke solutions to your ever-changing challenges and opportunities. Reimagine business as usual. KeyBank opens doors. Learn more at key.com/b2b. That's key.com/b2b.
Please touch on connected television because you've talked about data you've collected in the past that's very familiar to us and to many marketers. Talk about how you're now going into partnership to try and make sure that you're capturing data from what I'm watching on Netflix, on Amazon Prime, on Apple TV and the like.
Yeah, so connected TV is our fastest growing business and it's now, you know, well it's a nine-figure business a year for us. So it's no longer like a quote just getting into it at this point. We're able to use all of our data to target you for an ad in connected TV today. So we are looking at that same data in the CDP and saying Ann is perfect to get this particular credit card or she's actively buying this particular type of car or she's thinking about churning off her existing cell phone plan and we can hit you in any digital methodology which includes connected television. So it's a big differentiator where I sort of joke my wife a few years ago subjected me to watching The Handmaid's Tale, which was fine. I shouldn't be making fun of the show. But it was sort of I had to watch it. At the time every commercial was the exact same advertiser like for like hours. Remember you're like binging the show, right? And it was one of the mornings I woke up and said, 'Guys, can we do what we do on the internet and connected TV?' They're like, 'Well, we need a device graph.' I was like, 'Build one.' 2 weeks later we had probably the largest device graph in America and we were able to reverse engineer that into connected television and it's grown very very rapidly for us.
I want to go back to SaaS again just for 1 second because I don't want to get off that without being really clear about how incredibly silly the narrative that large language models are going to displace enterprise software is.
