Is the age of the moat over? Criteo’s Marc Fischli on OpenAI, agentic buying and life after retargeting
Criteo has rebuilt itself twice in seven years. Marc Fischli tells The Media Stack why the next act rests on the belief that nobody keeps a moat anymore.
Marc Fischli opens with what Criteo is not. “People still think we’re a re-targeter, which obviously is the birth of Criteo,” he says when we sit down in the Criteo yacht at the Cannes Lions Festival “You arguably could say that we were one of the inventors of the retargeting piece.”
That business made Criteo’s name and then nearly dated it. When Megan Clarken arrived as Chief Executive some seven years ago, the third-party cookie was already under threat, and she pivoted the company towards retail media, largely through acquisition. That business remains one of its two pillars. The second rebuild came about 18 months ago, when Google finally decided against killing the cookie. “That freed up a significant number of engineers, over 100 engineers, pretty much overnight,” Fischli says. The result was a strategic shift back into performance media as a growth engine, coinciding with the arrival of current Chief Executive Michael Komasinski. The products now reaching the market are what that cohort of engineers has been building since.
Fischli runs Criteo’s international markets, covering EMEA and Asia Pacific with a focus on performance media. He joined three and a half years ago after a long spell at dunnhumby, latterly as Chief Client Officer, and Diageo before that, a mix he says now runs through Criteo’s senior leadership: martech specialists alongside retail, CPG and agency backgrounds.
“The battle for the direct contact with the consumer, or the traveller, or the user, is on,” he says. “At the beginning, the LLMs wanted to go into that. I think they’ve taken a bit of a step back, because the retailers are, as they should be, fiercely defensive around that. But the battle is on. Where that transaction happens is where the best data is. Who owns that, and how exclusive it is, is going to be quite important. If everyone knows everything, the value of that goes.”
Strip the company back, and Fischli sees two assets. The first is its commerce graph, which he says knows what people do on the open web and across apps well because everything gets tagged. The second is access to the product catalogues of its retailers and e-tailers at SKU level, alongside 235 retailers on the retail media side, some of which give Criteo visibility into what happens offline in their stores.
“That is probably something we have underutilised as a company for a while, but we are not a data sales house,” he says. “It’s not our data at the end of the day. But what we can do is use this data to have advantaged services for our clients.” In practice that means targeting, creative and, increasingly, measurement and incrementality.
The product architecture built on those assets runs in three directions on the performance side. Criteo GO is a self-serve, cross-channel platform aimed at small and mid-sized advertisers and at the tail brands of large ones. “If you’re a Unilever, you have 3,000 brands. The last 1,000 never really got a proper outlet,” Fischli says. An enterprise version, co-built with agency partners and in beta, is due in 2027, with curation through Criteo’s own and third-party SSPs as the third leg. On retail media, the company is moving everything to auction-based buying and positions Commerce Max, its multi-retailer buying interface, as its answer to fragmentation.
Everyone talks nice at conferences. Then Monday morning happens
At a panel earlier in the week, Fischli was asked why Europe is ahead of the US in retail media. “The answer to that is, of course, in-store,” he says. The American industry started with sponsored products and stayed focused on them. In-store is messy, and the budgets sit with retail commercial teams inside yearly joint business planning negotiations rather than with marketing. The Europeans started in-store. “They’re used to messy stuff. When brands wanted to activate all of that universe, from sponsored product to a screen that happens to be in an aisle somewhere, the Europeans were actually quite well prepared.”
Both sides share a deeper structural problem. Retail media inverts a relationship retailers have spent decades enforcing. “In a retail brand relationship, the retailer normally calls the shots,” Fischli says. “Now retail media is different, because the client is inverse. The brand’s the client of the retailer, and that is a mindset change that people still need to go through.” Walmart understood it early with Walmart Connect. Many retailers in the US and Europe have built retail media networks that sit in marketing, “and it’s not particularly powerful.”
Advertisers have their own version of the gap between intention and behaviour. “The true ‘I get fantastic value out of retail media’ is still more the exception rather than the rule,” he says. “Everyone knows it, and everyone sits at the conferences and talks nice. And then we go back to our offices, and everything goes back to normal come Monday morning.”
Strategy, in his diagnosis, is not the problem; boards are on side, and CFOs love the EBIT impact in thin-margin grocery. Structure and people are. “What is in it for me if I run the commercial team? As long as I do my number, I don’t need to worry about the other guy.” On his maths, if retail media captured even half the share of voice it should, the industry would be two or three times its current size.
Partnerships and the honesty problem
Criteo’s retail media pillar was built by acquisition at speed, an urgency Fischli says no longer applies. M&A will continue, but selectively and with a higher hurdle. What has changed is the default. “What is the absolute new normal is partnerships. Everyone in the industry has realised, maybe with the exception of the very large American tech companies, that they cannot do everything themselves. You need to be much clearer about what you’re really good at, and be honest with yourself, which is not something this industry is particularly good at.”
Partnerships stall, he adds, in keeping revenue-targeted teams comfortable with overlap rather than at the top, where sign-off is usually easy. Zero overlap is so rare that if it existed, the deal would already have been done.
2,000 advertisers on ChatGPT, with no guarantees
The OpenAI partnership is the test case. Generative and agentic AI have pushed up the value of high-quality data, and when OpenAI decided to move into advertising, there was a live question over whether it would bring the transaction in-house. It stayed open to purchases remaining with the retailer or e-tailer, and that was when the two companies started talking. “They’re not specialists on advertising,” Fischli says. “So we bring that data set, which makes the targeting better, helps with creative, and the measurement will be moving towards much, much better. It helps them grow and kick-start their advertising business. They bring a lot of the tech, and we bring a lot of the data set.”
Criteo is not OpenAI’s only partner, but it was the first advertising technology partner in the ChatGPT ads pilot in March and, according to Fischli, remains comfortably the largest. “There’s a first-mover advantage, and we have by far the biggest. We’ve just announced we have 2,000 advertisers live on OpenAI. This is significant in just a few weeks.”
He describes the relationship as close, particularly with OpenAI’s advertising team, and insists the commercial story has barely started. “Our joint clients haven’t seen anything yet on the true value that this will have. It’s very early days, because the measurement piece, we’re only starting to implement.”
Could OpenAI eventually compete directly with Criteo? “We don’t know is the short answer. Is that a possible scenario? Absolutely. If you’re OpenAI, the one thing you cannot lose is being at the forefront of having the best model in the industry. Doing the advertising side with a partner makes an awful lot of sense in today’s environment.”
The pace of change makes longer bets pointless. On a travel panel that week, he and fellow panellists compared their predictions from a year earlier. “What was the delta? Oh my god, 75%. Scrap everything.” Criteo’s insurance is integration: an advertiser can compare the value of the OpenAI part of a campaign against the Facebook and retargeting parts in one place. “I won’t quite call it a moat, but it’s definitely still a bit of protection in the short term.”
“One of the things I’m usually talking about is, is the age of the moat over? When digital advertising first started, everyone wanted to create a moat. Some people managed to do that, and it worked well for a number of years. Given the speed of innovation, I don’t know whether the age of the moat is still here. What will win, in my opinion, is not necessarily the people who create the best moat, but the people who remain the most adaptive.”
Six months to democratised MCP
His most testable claim concerns agentic buying. In May, Criteo and Dentsu announced what they call the industry’s first MCP campaign: a Dentsu trading agent orchestrated it end-to-end for a French client through Criteo’s Model Context Protocol, with no one logging into a Criteo interface. “The results are fantastic,” Fischli says. “I think we’re going to see in the second half of this year a positive explosion of the usage of this. Over the next six months, we will see the democratisation of MCP.”
He is dismissive of the wider industry’s claims on this front. “A lot of people are telling everyone they’re investing in this. The amount of people that have done something that truly works is very, very small.” Once democratisation arrives, he expects the industry to catch up within a further six months. “Agent-to-agent will be a true reality when we talk next year.”
Criteo’s protection rests on data exclusivity, and its newest partner sits closer to the consumer conversation every month. On Fischli’s own telling, nobody in this market can answer that question for more than six months at a time.