← Marketecture: Get Smart. Fast.
Marketecture: Get Smart. Fast.
How Criteo's Todd Parsons Is Reinventing Performance Marketing with AI
00:20:15
Loading summary
Transcript67 lines
- [00:00]
A
Oh hello. I'm Jeremy Bloom, co founder and CEO of marketexture Media and boy do I have news for you. Market is back and if you've been watching from a distance thinking that looks phenomenal. But a trip to New York is just a bridge or several too far. We've got great news for you. We're bringing Market live to the beating heart of Adland. That's right. On September 23, 2026, market coming to Chicago. We're going to be bringing the same sold out energy, sharp insights and industry defining conversations to the center of advertising's biggest transformations. From media and commerce to AI tech and modern marketing. This is where the people shaping what's next for the ad industry will be, want to be and need to be. Early registration is live now, so lock in your ticket@chicago.architecturelive.com Again, lock in your ticket at chicago.architecture live.com Disclaimer I live in Chicago. It's not Chicago, it's Chicago.
- [01:18]
B
This podcast is brought to you by the Build, a new podcast from the guys behind Sincera, Michael Sullivan and Ian Myers. They built their company by figuring out clever solutions to a few important ad tech problems in our industry. And that's exact what the show is about. Mike and Ian interview some of the smartest tech minds in the biz to hear about how they identified opportunities, solved their hardest challenges, and grew their businesses in the process. Listen to the Build with Mike o' Sullivan wherever you get your podcast. Welcome to marketexture where you can get smart fast with in depth interviews of leading tech executives. I'm Ari Paparo. I'm joined today by Todd Parsons, the CPO of Criteo. Is that correct? You're the cpo? Got the title right and the pronunciation correct.
- [02:16]
C
I'm the president of Performance Media and the cpo.
- [02:19]
B
Well, thanks for joining me. So we recently had your boss, the CEO of Criteo, on the show. I interviewed him at CAN on the yacht and everyone's complaining it was hot. But now I'm back in New York and it's much, much hotter.
- [02:36]
C
It is. Great.
- [02:36]
B
So what have we learned?
- [02:38]
C
Yeah, we definitely learned that we're better off on a boat, right?
- [02:44]
B
Yeah, I would strongly prefer to be on a boat than doing this interview remotely. So I want to talk about the product stuff that's going on at Criteo because you're investing. You've always been leading edge investing in product. So about earlier this year you introduced go, which is your AI powered performance platform. Take us through what GO is and how it's going so far. Yeah.
- [03:11]
C
Okay, we're excited about Go. It's basically centering around solving the problem that performance marketing was originally built around channels. So you got search over here, you got social over there, display and Programmatic somewhere, each with its own tools, workflows, optimization, logic and console and so on and so forth. So that model's breaking down. Marketers that we're serving, they just don't need another dashboard to manage the fragmentation. So Go is really a system that is reducing that complexity and it's our answer to the shift. It's a lot of AI assisted performance in a platform that uses our commerce intelligence to make execution simpler and a lot smarter. And we hope people enjoy it.
- [04:01]
B
All right. And enjoy it, that's an interesting choice of verb. Whether they enjoy it or not. Who are the customers? Is it retailers, brands or everybody?
- [04:15]
C
It's a little bit of everybody. The customer profile that we're most excited about serving are users of Google Performance Max and pmax as it's called, and also Meta's Advantage Plus. So that's a wide variety of folks, but they're basically users that are excited about getting outcome based performance and are willing to accept a non transparent model for doing that, for having their budgets reallocated across the channels towards a specific outcome. So it does vary, but you're talking about more small to medium sized businesses at the core.
- [04:55]
B
And is it basically just give us your money, we give you performance. How do they differentiate versus the other platforms?
- [05:02]
C
I mean effectively what's differentiated is that it is cross channel. So it's the only platform that I know of where you can get to all of the goodness of the open web still, whether it's online video or whether it's app or whether it's just simply Display and also Meta and TikTok and more recently OpenAI. So that's really the distinguishing proposition and in terms of that, the outcomes, it's really simple. It's like new to brand and new to product are calling in the industry discovery these days. How am I going to get discovered? How do I get either a lead or a net new customer and then what remarketing am I going to do with those? It's just that those three different tactics cross channel.
- [05:49]
B
And are the customers largely tied into the retail media world? Are they, are they uploading, you know, product feeds and things like that as the dominant use case, less tied into
- [06:00]
C
the retail media world in terms of how the product is executing from an inventory perspective, but definitely from a merchant perspective, it's much the same. Same product feeds, same site side tagging or server side data collection that allows us to merge two data sets and get to the outcomes that we're talking about.
- [06:23]
B
So tell us how data fits in. So Criteo has enormous amount of shopper data. I think the number you throw around, 740 million daily shoppers. So how valuable is that data and how does it work in like the deterministic way where it's data about that particular customer versus more generally identifying shoppers writ large? It's both.
- [06:48]
C
It's a hybrid graph that we run that resolves to deterministic data and it uses a little bit of deterministic data inside of it, but it's mostly across device data setup that is, you know, creating a view of area of Todd and making it possible to look at where you might be landing across those channels and then, you know, target you in an environment where you're more likely to be responsive to an advertisement. So the data is. Is the primitive really. If we didn't have the data that we were operating, then we couldn't really look at those 740 million shoppers and, and know that, hey, it's pretty likely Todd's going to convert reach him on an Instagram reel versus an online video that when he gets to the point of purchase, he's going to buy whatever the product we recommend is.
- [07:42]
B
Right. And the model is similar to the big walled gardens in the sense that you can't buy that data, you can't rent that data. You're getting it as part of go.
- [07:52]
C
That's right. You're getting to it as part of go. We don't sell data. And I think you touch on a point that I'm always asked, which is like, okay, you've got all of this data. Why would I choose to work with you versus a Google or a meta? And the simple answer is, well, we're independent in terms of where we're landing your advertising. So we don't have a. And Amazon's the same. I don't have a good reason to direct you back to Prime Video. I don't have a good reason to direct you to YouTube. I'm going to go to where the outcome is most likely to be realized.
- [08:27]
B
So performance is a big word. People use it a lot of different ways. Are customers coming to you with, you know, a wide variety of goals and metrics? Are they looking for, mmm, mta, et cetera, or is it really just, you know, make the cash register ring?
- [08:44]
C
It's always make the Cash register ring. But the measurement varies. Right?
- [08:49]
B
Right.
- [08:49]
C
So we, we, I think a little known fact about the company is we, we are only getting paid on outcomes. We're not a media buying platform, not yet. So that may be in the cards later. For now we're getting paid on sales lift as you said. So depending on the customer, the source of truth for validating that what we're doing is actually selling product does vary. So really large customers that we service might be using a mix model, an attribution system of their own design or someone else's. More often than not though they're using, our customers are using GA4, sometimes Adobe analytics as their source of truth. And what we're doing is taking that setup and working our way backwards to, you know, the campaigns that we serve so that they're grading the homework and we're not.
- [09:42]
D
Heat up your fourth of July at the Home Depot with our wide variety of grills under $300 and make every gathering one to remember. Give your outdoor space a glow up whatever your budget is. With savings on seasonal plants starting at $5. With the grill fired up and your backyard set to perfection, you'll be able to invite friends and family over to kick off the party. Start celebrating with low prices guaranteed at the Home Depot. Prices may vary by store exclusions apply. See home depot.com pricematch for details.
- [10:12]
B
Right, and you said you only get paid on performance. What are the models available for risk taking on the customer side?
- [10:19]
C
I mean for now, you know, we're largely again looked at from a ROAS perspective. So you know, we're looking at sales lift. ROAS is the proof. Of course there are a bunch of tactics that lead to that. You know, more recently a good example is we opened up CPC and in OpenAI but fundamentally every customer is looking at more than post click. They're looking at sales lift. And so and that makes my job really easy because I get to come back to that every time from a product perspective and it's like sure, yeah, we're going to support these other tactics but if they don't sell, we're not going to support them.
- [10:59]
B
Yeah. So but sometimes that actual determination of what sales lift is, is, is complicated or not obvious. So do you still only get paid if the customer's definition of sales lift is achieved?
- [11:13]
C
I mean what happens if we, if we don't achieve that measure is that they move the budget. So.
- [11:19]
B
Right.
- [11:19]
C
It's pretty pure motivation. You know, like they'll more if we're hitting the numbers and they won't give us more if we're not.
- [11:26]
B
So let's talk about AI, which is obviously the hot topic at CAN as well. So first, I guess the question is how is AI tied into the GO product? How is it kind of the legacy of Critio's investment in machine learning over the last 20 years?
- [11:40]
C
Yeah, I mean, I would sort of separate it into a few different buckets. I mean, first of all, what you just said, and I'll start with that, we went from a fairly simple, you know, machine learning model or set of models to. To deep learning, so we could deal with signal loss a little bit better in all of the prediction that we make. So at the core of the product, there's a lot of AI decisioning that's going on. Which user, which recommendation, which bid, which optimization from a DCO perspective, and so on and so forth. So that's at the core in terms of product of the GO product. There's a lot more to it that the user experiences. One is like just simplifying onboarding. So the process of slurping up your product catalog and tagging your site, getting hooked in with your measurement source and so on and so forth is largely automated. And I wouldn't say it's agentic per se, but there's help along the way that makes that super easy to do once you're in the platform. And that onboarding thing used to take a long time for us as a managed company. Now it's compacted down to hours. So once you get onto the platform, it's. There's a lot more. There are agentic driven recommendations for campaign setup and where you might want to set a goal, for instance, and then going further, there's a lot of work to auto generate creatives, as you would expect, and go ahead and run those across the channels. So we're sort of picking off the key use cases one at a time. Ari, and just.
- [13:24]
B
Yeah, it makes sense.
- [13:25]
C
Easier and easier to use.
- [13:27]
B
Yeah. Critio, I mean, historically was known for having these very templated creatives where you just put the product in there and get people to click on it. So I imagine AI has really blown that out of the water.
- [13:37]
C
Yeah, it has. And thank goodness not many people want to look at a templated creative. I know I don't. And so that ship has come in for us.
- [13:49]
B
What about agentic buying? Myth of reality or experimental area?
- [13:55]
C
Experimental, but increasingly seeing more of it. Not on the GO platform itself, but for the rest of the business. We opened up a lot of our backend wherever we have a public API through, through MCP to some of our bigger customers about a year ago. So I think we were pretty early in that. And honestly the first probably few months was we're just sitting around looking at people building reporting, which I think you would expect. Now of course those use cases are more orchestrated and so we have agents that are coming in, doing different things, planning and pulling audiences and structuring a forecast for a cross channel full funnel campaign and then go ahead and activate it. So it's definitely changed a lot in a year, but I would say it's still early days, you know.
- [14:47]
B
Right. It's automating the things that already need automation, but the, the idea of a kind of a brand new way of media buying is a little bit farther away.
- [14:56]
C
Yeah, I mean I would say just making the old version, you know, a little bit more autonomous and, and executable is what we're seeing right now.
- [15:05]
B
Right.
- [15:06]
C
But not a sea change. Right?
- [15:08]
B
Yeah. This is kind of a new thing for Critio in that it's kind of self service signup SMBs. It's not the enterprise sales that kind of help build the company. I mean Critio has always had thousands of advertisers, but it seems like this is categorically different. Can you speak to what you've learned and how different it is?
- [15:29]
C
Yeah, I mean it's a massive evolution for the company. Not to oversell it, but historically, like you said, We've been 100% managed services. So Go is opening up the platform to a much broader set of advertisers. You can imagine with SMBs in particular, you run out of people and attention very quickly and what happens is that just gums up the entire works of delivering a customer a good experience. So the speed is slowed down, there's complexity and the performance suffers. So what we're seeing is the opposite. With self service and Go, we're taking away some of the heavy operating overhead. But what's really exciting to me is we're seeing customers spend more, they churn less, they don't pick up the phone as much because they don't eat. So it's sort of everything looks good, you know, early on from that perspective.
- [16:28]
B
Right. Do you give any sense of the average size of the customer or any statistics you could share?
- [16:35]
C
Just massive amount of demand and we're onboarding them as fast as we can. I think the first three months have been like you have experience in building your own companies and products. Products. You're doing a lot of Working out of kinks in the workflow and we're seeing things like a lot of agent based fraud exploits and things that we didn't expect in the system. Just to give you a little bit of behind the scenes that we had to go back and address. So it's been more like making sure that we have all of our Rails correct and and we're onboarding many, many customers into the platform and the demand's really strong but it's still like a V1 in the classic sense. You're three months into it and everything looks good. It can always go faster, but it's going fast enough for us to learn and to continue modifying the experience, making it better.
- [17:34]
B
Sounds great. So let's do a quick lightning round. We'll just talk about go here because that's what we've been talking about. So what is the biggest competitive advantage?
- [17:42]
C
Cross channel.
- [17:44]
B
What is. Well, you do lightning round. Well, Short answer.
- [17:49]
C
Yeah, I try.
- [17:52]
B
What's the biggest challenge?
- [17:54]
C
Biggest challenge is getting recognized against these giant platforms and being credible.
- [18:02]
B
Anything you could say about the roadmap? What's coming next?
- [18:06]
C
Well, I was excited to mention earlier that we got OpenAI going. So we're just looking at other really ripe channels for getting performance. You can imagine app is something that we're pretty excited about sort of studying Applovin and incredible success there is not lost on us. So we're looking very much about how we can move channels. And then another really interesting thing that we're exploring and it's really early days on this, but is the sort of the confluence of what is called geo. You hear GEO talked about a lot of. I sort of have this perspective that GEO is going to be a lot more like paid marketing than SEO. And the reasoning for that is because if you remember, we're both old guys, right? Can I say that? And it used to be back in the day that everybody sort of hunkered down and waited for Google to announce its page rank updates. Remember that? And then it was a big, big mad scramble to figure out.
- [19:10]
B
Google Dance they called it.
- [19:11]
C
Yeah, exactly. So the thing about AI and especially the answer agents is that when the models update, everything changes much more rapidly. So every model update changes the way that brands are going to get discovered organically and they're still looking for the same outcome. Right. So my thesis there is that you're going to get paid for an outcome either way as this world drives towards outcomes. And I think GEO could play a role in that.
- [19:42]
B
Okay, that's interesting. That wasn't a short answer, but it was interesting. Last question.
- [19:46]
C
Lightning. Yeah, bad Last question.
- [19:49]
B
If GO was an animal, what animal would it be?
- [19:53]
A
Wow.
- [19:54]
C
Okay.
- [19:55]
A
I don't know.
- [19:56]
C
A leopard.
- [19:57]
B
Why is that?
- [19:58]
C
Because it's agile and it goes everywhere fast.
- [20:03]
B
All right, Good answer. All right, Todd Parsons, the CPO of Criteo. Thanks for joining us.
- [20:08]
C
It was great seeing you, Ari. Thanks. Have a good four.