Commerce

Podcast: Understanding Amazon’s advertising advantages (with Adam Epstein)

By Eric Benjamin Seufert

On this week’s episode of the podcast, I am joined by Adam Epstein, the co-founder and CEO of Gigi, to talk about Amazon’s strategic advertising advantages. Among other things, we discuss:

  • The value of Amazon’s deterministic identity spine, which spans ninety percent of US households
  • Whether Amazon can successfully transition its e-commerce data dominance into non-endemic categories like automotive and pharmaceutical services
  • The risks faced by independent demand side platforms like Trade Desk as tech giants internalize inventory
  • What role agencies will play in a future where AI agents manage cross-channel programmatic media buying autonomously
  • When shoppable TV transitions from a niche consumer behavior to a primary driver of performance marketing results
  • How the integration of Amazon’s retail data with AI chatbots might permanently displace traditional search-based product discovery

Thanks to the sponsors of this week’s episode of the Mobile Dev Memo podcast:

  • ⁠⁠INCRMNTAL⁠⁠⁠. True attribution measures incrementality, always on.
  • Xsolla⁠. With the Xsolla Web Shop, you can create a direct storefront, cut fees down to as low as 5%, and keep players engaged with bundles, rewards, and analytics.
  • ⁠Branch⁠. Branch is an AI-powered MMP, connecting every paid, owned, and organic touchpoint so growth teams can see exactly where to put their dollars to bring users in the door and keep them coming back

Interested in sponsoring the Mobile Dev Memo podcast? Contact Mobile Dev Memo advertising.

The Mobile Dev Memo podcast is available on:

Transcript

Eric Seufert: Welcome to the Mobile Dev Memo podcast. I am your host, Eric Seufert, and I am joined today by Adam Epstein, who is joining the podcast for I believe the third time. Is that right, Adam?

Adam Epstein: That is right, Eric. Pleasure to be here.

ES: Pleasure to have you. So, we spoke last year right after Cannes, and as we speak now, Cannes will be in a few weeks. Are you going?

AE: Not a linen and Rosé guy, Eric?

ES: Not a linen and Rosé, yacht parties, dock shoes, and brand strategy guy. But maybe one day if someone paid for me to go, I would go. I want to talk about Amazon. You are the designated Mobile Dev Memo Amazon expert. They just had their upfronts. Before we get to that, please reintroduce yourself to the audience.

AE: Adam Epstein, co-founder and CEO of GG. GG is an agentic operating system for programmatic media. Our first product is AI agents for the Amazon DSP. We launched this product in the summer of last year. Since then, we have grown significantly. We power hundreds of millions in Amazon DSP spend. In November last year, we won Amazon’s prestigious Global Technology Innovation Partner Award, and we are continuing to grow. Prior to GG, I was previously co-president of a large retail media company called Perpetua, which is now owned by Omnicom. I have been working in Amazon ads as a partner for almost eight years now and have powered billions in ad spend and 15 to 20 billion in GMV. I am excited to share some of those learnings today.

ES: Full disclosure, I am an investor in GG. Let’s dive right into the Amazon upfronts. Amazon announced that its authenticated graph now deterministically reaches 90 percent of US households with verified logged-in non-modeled data. Talk me through the implication of that for advertisers. What is the power of that data set?

AE: To understand the power of that data set, we need to understand where the growth of Amazon’s advertising business is going to come from. Amazon’s advertising business in 2024 was just shy of 50 billion dollars, still growing 20 percent year on year. If you forecast that out, it will likely grow significantly. The path to 100 billion dollars in advertising revenue is going to come primarily from Amazon’s DSP. Amazon’s advertising business is split in two: on-site sponsored ads, which are pay-per-click ads that occur on search or product detail pages, or the Amazon DSP, which uses Amazon’s first-party shopping data as well as a suite of other Amazon ad tech tools to buy display ads, video ads, primarily connected TV ads, and audio ads on Amazon’s owned and operated properties like Prime Video, as well as other third-party properties like Disney and Netflix.

A lot of Amazon’s advertising growth, if not all, is going to come from the Amazon DSP, primarily because the advertising revenue on sponsored ads is capped by inventory and price. Everything on Amazon is already an ad, and CPCs have more or less flatlined because sellers and brands on Amazon.com are unable to pay more for ads based on their profitability thresholds. A lot of this incremental revenue is going to come from the DSP and specifically from CTV buying. If you look at all the channels that Amazon DSP offers, display advertising is not a growing market. There is less time occurring on traditional websites as people spend more time in LLMs. However, CTV is a growing market. It is a significant piece of Amazon’s market share, specifically the stronghold that they have within Prime Video. The way in which media buyers buy TV ads, at least the largest media buyers, they do so with a notion of reach and frequency. In particular, having a deduplicated identity spine to manage to reach and frequency across all TV buys is a critical piece of a broader TV strategy.

Amazon has this unparalleled advantage in this identity spine that they have created by having 90 percent of US households, which is built on the foundation of an announcement that they had last summer with Roku. Last summer at Cannes, Amazon and Roku had an announcement in which Amazon effectively said that they partnered with Roku to create a unified identity service, combined between Amazon’s notion of US identities as well as Roku’s. Roku is able to complement Amazon because Roku is the largest OEM in connected TV hardware. Effectively, they have been able to close gaps that Amazon was unable to identify deterministic households themselves to create this unified identity spine.

Now, with this unified identity service, Amazon can go to TV marketers who are ostensibly buying TV ads in other places, whether that be linear TV buys with publishers or other enterprise DSPs like DV360 and Trade Desk, and effectively say that you should centralize all of your TV media buying into Amazon because you will have the highest signal identity layer for unified reach and frequency out of any possible other mechanism of buying TV ads. That is very powerful to the largest TV media buyers, brands, and agencies in the world.

ES: I want to dig into that. How does an agency or an advertiser partner compare the utility of that identity spine to what is on offer from other DSPs? GG supports agencies, so talk to me about the value of that identity spine relative to what else they might be buying from in a programmatic setting.

AE: It is important to understand media planning across an agency and how an agency would choose to allocate media for a brand across other enterprise DSPs. If you think about the unique offering that Amazon has and the advertisers that Amazon has, Amazon has this incredible subset of e-commerce and CPG brands in which Amazon has incredible degrees of proprietary data on a significant amount of purchasing decisions done across their product offering. Those are often known as endemic advertisers to Amazon. Amazon effectively has done a very good job over the past few years of centralizing CTV spend across endemic advertisers to the Amazon DSP. A lot of the headwinds and a lot of the market share that Amazon has gained over the past few years has been creating their DSP to be not just a place for endemic advertisers to buy Prime Video inventory or display inventory to drive sales on Amazon, but a full-funnel solution across all CTV and all display and audio media buying for endemic advertisers.

Particularly, Amazon’s growth and the work that they are doing with agencies and brands right now is understanding and influencing media buying and planning decisions across non-endemic categories. Let’s talk about the biggest non-endemic categories for someone like Amazon: financial services, pharma, and auto. Amazon has a treasure trove of shopping data of when people buy cars, how people make household and family decisions around whether to buy an SUV or a minivan if they have a baby. Another interesting thing that Amazon built is Amazon Autos, which is effectively digital showrooms for any auto OEM to create for consumers to begin to compare and contrast various cars. Additionally, for some auto OEMs, they have actually enabled the ability to purchase those cars on Amazon. While that is a very small percentage of an auto OEM’s purchasing history in the US, I spoke to a CMO of a large auto OEM and he said more people are buying cars than you think, which is really compelling. Most interestingly, Amazon Autos and these digital showrooms have created a new touchpoint on the path to purchase for some of those auto OEMs. Rather than just going to the auto OEM’s website or a dealer website, Amazon has effectively created a new touchpoint to move people down the path to purchase that is unique and proprietary to them that no other DSP has to offer.

Effectively, they have gone to this massive category of auto media buying, which is one of the largest CTV categories in the world, and said not only do we have this unique data advantage in terms of our reach on US households, but we have these unique data advantages with the first-party data that we have on Amazon.com proper as well as Amazon Autos. You can extend that to pharma. Think about all the purchasing history that Amazon has about people on their health. Additionally, Amazon has made a number of acquisitions and has had a variety of initiatives to deepen their retail footprint on the pharma stack. There is a product called One Medical which allows you to get all of your prescriptions from Amazon. They bought PillPack not too long ago for over a billion dollars. They haven’t exposed those data sets to advertisers, but theoretically, if they did, then Amazon is effectively taking another non-endemic category and making it incredibly endemic and giving it an unfair data advantage in that category. When we think about the reach and this household penetration, it is about Amazon creating unfair data advantages relative to other ways that people would buy TV ads in particular and using those unfair data advantages to centralize as much of one’s budget as possible to the Amazon DSP.

ES: That is very complete and fascinating. I wrote a piece recently called Amazon’s Advertising Advantage, and the whole point was they have got this identity spine which is unmatched. If you look at Amazon’s revenue growth, it was decelerating from a high of 27 percent in Q4 2023, but now it is back up in the mid-20s and has been for the last four quarters. You tie that to all these partnerships that it has done where historically the walled gardens had exclusive access to their data sets. Amazon’s strategy seems to be the opposite. We have our own proprietary data across any number of verticals and we are broadcasting that out. We saw that with the Roku deal and the Disney deal, which were identity deals. They were not necessarily inventory expansion deals. That coincides with an increase of accelerated growth. It is a fascinating business model they have assembled here, and it seems like it is almost infinitely applicable to all the inventory everywhere, and CTV is obviously a big chunk of that.

AE: That is exactly it. It is not necessarily inventory everywhere because the one place that Amazon is unable to bring these data sets is to social ads as well as Google Search. But if we think of any place that is effectively in the open internet or pseudo-open internet, with CTV being the primary category for both scale and growth right now, Amazon has done an incredible job of enabling third-party partners to benefit from Amazon’s proprietary first-party data for both audience building as well as measurement. This is a really important piece to understand: Amazon views their competitors in CTV media buying as being The Trade Desk and DV360, Google’s DSP.

If we think about how Amazon can compete with these players, they compete with these unfair data advantages, and they also compete with the fact that one of the most scaled premium CTV ad placements is Prime Video ads, which you can only buy on the Amazon DSP. For many advertisers, you have to come to the Amazon DSP to buy Prime Video ads, and if you choose to do so, Amazon’s argument is that you should do so across all other premium CTV and centralize across Disney, Netflix, Paramount Plus, NBC. All of your TV media buys should be done through the Amazon DSP. Amazon across both their demand-side tech and supply-side tech and Amazon Publisher Services has done an incredible job of expanding the possibilities of supply across all premium CTV. 18 to 24 months ago, I think a lot of folks would have said that Amazon’s primary competition is The Trade Desk, and I think Amazon and Google have done a phenomenal job in the last two years of creating existential doubt on The Trade Desk’s very existence given that it doesn’t own any proprietary inventory or proprietary data.

Really, if you think about Amazon and Google competing against each other, it is YouTube as its lynchpin on Google’s side to bring media buyers into DV360 and it is Prime Video ads on Amazon’s side as its lynchpin to bring people into Amazon DSP. The positioning that Amazon has gone after relative to a Google is saying that we are the DSP for premium CTV inventory, juxtaposing that with Google and YouTube, which people will say are cat videos and creators and all of this other junk. If you want to have a unified CTV strategy across all premium inventory, you would do so on Amazon. Conversely, Google will say how YouTube is actually CTV, it’s premium, it’s all of these incredible engagements and touchpoints that are unique to Amazon, but Amazon’s strategy is very much positioning itself not only in the US but globally because of Prime Video as the premium CTV DSP in the world.

ES: I want to hover here for a little bit. Trade Desk is not a company that I follow super closely. Two or three years ago, I started taking a look at the company and asking what the moat is. Give me the bull case argument for this company being very successful. I talked to a lot of people in the web space and went through a list of what I thought were very systemic challenges the company faced. People would say that the CEO is a brilliant guy and he will navigate the company through this. Just yesterday, I saw from Adweek that Trade Desk got a new CFO, and he is the fourth CFO they have had in a year.

AE: Not usually a great sign.

ES: I was like, wait, hold on, that’s extraordinary. I have never heard of that. Is there a history of a company that was not extremely financially distressed or facing some sort of regulatory intervention that had four CFOs in a year? Now obviously one of those was interim, but still. The answer was no, that it was unprecedented. I just thought that was a detail that maybe was something that you would expect to be talked about more. It felt like a pretty big revelation there.

AE: I think the decline of Trade Desk has certainly been a topic of discussion across the broader ad tech community for the past year. There’s the Hemingway phrase of how did I go bankrupt? It was suddenly and then all at once. Amazon and Google have done an incredible job of arming their sales teams with ammunition as to why they should centralize their programmatic media budgets to both their platforms. Over the past three to four years, it was slowly chipping away at it, and I think we are now in a position in which it is suddenly and all at once.

We get exposure to CMOs and agency decision makers who are now just saying they don’t know why they are funneling any dollars through The Trade Desk. At the same time, Trade Desk has built a phenomenal business in which billions of advertising spend and hundreds of millions of dollars of revenue are flowing through it. It is growing, but it is not growing at the growth rate that it used to grow at, and it is unclear if it will ever achieve the growth rate and scale that it used to operate at. One of the things that I have been particularly disappointed in the CEO’s commentary towards the decline of their growth has been the shots that he has been firing towards Amazon, which are just categorically untrue.

He basically says that Amazon’s DSP is not a priority at Amazon. Absolutely not true. Amazon’s advertising business as we just said is going to get 30 billion in incremental revenue, and it’s going to come largely from the Amazon DSP. There are tens of thousands of people employed at Amazon Ads. Almost all of them, I would probably say 60 to 70 percent of them, are working on the Amazon DSP and not on-site sponsored ads. Another comment that he often says is Amazon favors their own inventory, in particular their sellers are always pushing Prime Video. This is categorically false. I don’t know if I should say this, but just in working very closely with Amazon sales teams, they have actually been incentivized to sell non-Amazon inventory over Amazon inventory.

The reason why gets back to the comment that I was making earlier in that Amazon’s opportunity is to centralize all of one’s TV buys under its DSP. It’s not going to do that if people are just coming for Prime Video. If I can sell you a couple media buys on Disney, then maybe I’ll bring you in for all of your Netflix buys and maybe I’ll bring you in for all of the future NBC and Paramount buys. Jeff Green has for whatever reason underplayed the competition between Amazon and has categorically expressed falsehoods in the strategic element and the go-to-market motion of Amazon Ads with respect to its DSP, which just doesn’t make any sense.

Trade Desk will continue to exist. They’ve done a great job of ingraining themselves across large agencies and brands. There are aspects of The Trade Desk that allow them to have unfair advantages, particularly compared to Google and Amazon. Trade Desk offers log-level data on impressions and people can match that back to outcomes. Getting back to pharma as an example, pharma is a two-sided marketplace to advertisers: they market to consumers and they market to healthcare professionals. A pharma brand needs to know the very specific impression that they bought to lead to a healthcare professional purchase, and at the moment you can’t do that on both Amazon and Google, but you can do that on The Trade Desk. The degree to which Trade Desk is open about their data and open with their data, particularly log-level data, I think you’re going to see continued momentum for them to entrench their market share. But if you talk to any seller of Google and any seller of Amazon and you ask them who they go after and what budgets they go after right now, it’s Trade Desk CTV budgets to be centralized to their own DSP.

ES: What is the right way to think about Amazon’s position in the digital economy? Because it’s not any one thing. It straddles a number of important roles. It is an RMN, it is a DSP, it owns its own CTV platform, it’s got a retail platform. How should advertisers view its role? What should they seek to get from it?

AE: Amazon’s advertising business and Amazon’s retail business sit at this incredible intersection between programmatic media and marketplace commerce. Amazon’s DSP is probably the fastest-growing at-scale DSP. Amazon’s commerce business is the largest online store in the world, and there are compounding benefits to both of those existing. If you think about all of the current trends within broader digital advertising: retail media, Amazon is by far the largest retail media business in the world. I think still to this day Amazon represents 70 to 80 percent of all retail media. CTV, Amazon’s DSP is the fastest-growing DSP for CTV buying as well. Prime Video ads, by turning advertising on by default, is the most at-scale ad-supported premium CTV service in the world.

Additionally, Amazon’s commerce capabilities are only continuing to proliferate. The reach that Amazon is getting across other categories and, again, turning non-endemic categories like auto into a pseudo-endemic category is just creating more proprietary signals for Amazon to broaden its scope across commerce more broadly. Amazon sits at this incredible intersection of everything commerce and digital advertising. Where are Amazon’s blind spots? Amazon does not touch social commerce to any meaningful degree. They have a partnership with Meta, but we haven’t really seen much scale with that. The inventory that Amazon has access to for its DSP is constrained relative to someone like a Meta and a TikTok, in which Meta just has this unique ability to create net-new inventory and monetize that inventory. Amazon’s DSP is constrained by broader CTV, which is still growing but is not growing to the rate that Meta is growing, and Amazon doesn’t have all of this organic content that a Meta and a YouTube have to create net-new ad inventory. Amazon is coming from a position of strength. The path to getting close to 100 billion in advertising revenue is there, the growth rate is continuing, but if we look at the market share of global digital advertising, Meta just passed Google. Amazon coming up to that second and competing with Google, I think Amazon is going to have some challenges with the current structure they operate in getting past 100 billion in revenue.

ES: Interesting. Amazon’s outgrowing search. I made a comment on Twitter yesterday because some account had said they can’t believe Google’s falling behind Anthropic and OpenAI. I just responded they announced a billion MAU for AI mode at I/O and search is growing 19 percent. Are they losing here or do you not need necessarily the absolute number one model based on some benchmark to be winning? What is your definition of winning here? I’d say 19 percent year-over-year growth at search scale plus a billion MAU in AI mode is winning. Someone commented it’s search and other as the line item. You think Maps and Google Play Store inventory are dragging that growth number up? That’s so pedantic and unhelpful. Talk to me about chatbot inventory because that’s something people haven’t spoken about. Amazon invested in Anthropic. They did the deal with Meta. My hypothesis is they do account linking with ChatGPT. That gives them the next social. They lost social. Twitch is not that big. Could this be the next aggregation of attention and they get ads in ChatGPT? Could you imagine that happening?

AE: Partnerships need to be forged between both ChatGPT and Amazon. If we think about ChatGPT and their future advertising ambitions, e-commerce is largely going to be a significant part of it, and an e-commerce product just simply doesn’t work at scale without being able to have reliable e-commerce data, inventory, catalog, pricing, delivery, that you get from Amazon. Amazon still represents 50 percent of total US e-commerce. A Shopify-enabled ChatGPT commerce product is going to be simply niche. ChatGPT doesn’t win on scale in ads and e-commerce at niche; it needs Amazon. Conversely, Amazon’s been investing in its own chatbot, as you know, Rufus. The interesting piece with Amazon is Amazon has this disruption dilemma that Google is also facing in that the lion’s share of Amazon’s advertising revenue right now like Google is on search: people bidding on keywords or product SKUs on Amazon.com.

If people are spending less time on the path to purchase on Amazon.com, whether that be in a third-party chatbot or whether that be in their own chatbot, then Amazon is actually losing their base of existing advertising revenue because people aren’t clicking on as many keywords or viewing as many product SKUs as they normally would. What Amazon’s very astutely done is they’ve created sponsored prompts within Rufus, its AI chatbot, and they’re showing an immediate ability to monetize that the same way that Google is doing a great job of monetizing their AI answers within Gemini.

If you look at potential surface areas with which Amazon can dramatically expand its advertising revenue and advertising potential, OpenAI would definitely be the first place that I would look towards, and it’s just unclear how that will materialize. Obviously, product listing ads within Amazon’s ad tech stack could easily be extended to OpenAI. Amazon’s DSP for display inventory could very easily be expanded to ChatGPT, but Amazon’s going to want a very significant piece of that advertising revenue. Amazon partnering with all frontier models except for Google obviously, and OpenAI is partnering with every cloud service provider. There’s lots of partnerships, but the partnership that I obviously want to see is what is the future advertising partnership between ChatGPT and Amazon? I just believe that ChatGPT is going to need to open up some of that advertising inventory that they’re creating to Amazon purchases, which will invariably lead to a significant amount of increased advertising revenue from Amazon. We just don’t know when that will be or how that will materialize. It’s also gaining customer mindshare of building habits to start your purchasing journey on OpenAI and ChatGPT. Because if those customer journeys and those shopping habits continue to start and exist primarily in Amazon itself enabled by Rufus, then OpenAI and ChatGPT will just continue to be a niche behavior for e-commerce. A partnership definitely is something that we’d like to see, but TBD.

ES: I don’t want to revisit my agentic commerce is a mirage battle because I think it’s over. People say we’re so early with agentic commerce; no we’re not. It’s over. Rufus, Sparky, that’s agentic commerce. It’s on-platform discovery. What people think we’re just on the cusp of happened. Amazon won, Walmart won. Instant checkout was suboptimal on its face. The affiliate model was suboptimal for economic and incentive reasons. Independent agentic commerce was never going to work out. We’re not at the very beginning stages. We’re not writing the playbook. The playbook exists, it’s advertising. If Amazon never opened up its catalog, instant checkout was DOA. Amazon is too big. E-commerce is too concentrated, so is advertising. You need cooperation from the largest entity in order to make anything work.

Now you’re seeing why Amazon would do a deal with or why they would invest in OpenAI. On the one hand they get a customer for Trainium, but the other part has to be that they are going to execute against the playbook that they’ve been executing against, which is to broadcast this identity spine onto a new surface area. They missed social and they’re going to seize the opportunity with chatbots. I’m almost certain of it. That’s what this has to be. Then that is the death of independent agentic commerce. Why? Because it demonstrates that that model is suboptimal. OpenAI can shut down instant checkout after six months which they did. You still have non-believers. But the second they open up to advertising from Amazon, Amazon broadcasting its identity spine plus its own advertising infrastructure onto that surface area, everyone else is going to say okay obviously that’s a superior model and that’s where all the chatbots are going to go. They have to. You partner with the biggest ad platforms, you partner with the biggest retail platforms, and you accept their ads and you build your own ad infrastructure because that’s the dominant economic model.

AE: The definition of agentic commerce just changes in that these chatbots just become the surface area with which people make purchasing decisions and they partner with broader e-commerce platforms like an Amazon to execute those decisions. People are still figuring it out, it’s a new category, so I wouldn’t say it’s dead per se, but the evolution of this will absolutely have to be partnering if OpenAI and ChatGPT have these grand ambitions of hitting 100 billion dollars in advertising revenue. The two categories that you need to have to hit a 100 billion dollar advertising revenue is CPG and e-commerce. If you’re unable to hit those two categories in digital advertising, then it’s going to be very challenging to build a meaningful scaled advertising business. The quickest path towards advertising budgets on those two categories is to partner with Amazon.

Not just in gaining advertising revenue wallet share, but it’s also gaining customer mindshare of building habits to start your purchasing journey on OpenAI and ChatGPT. Because if those customer journeys and those shopping habits continue to start and exist primarily in Amazon itself enabled by Rufus, then OpenAI and ChatGPT will just continue to be a niche behavior for e-commerce. A partnership definitely is something that we’d like to see, but TBD.

ES: I read an interesting piece recently where the author makes the case that big technological disruptions have historically pushed advertisers to agencies. Is that what you’re seeing and why agencies and brands hire GG? Because look, this is too disruptive, it’s too frantic and dynamic. We’ll let you be the experts on managing AI and all the applications to advertising. Is that a reasoning that you come across?

AE: I think that’s a fair thing to say. Many people say agencies are like cockroaches; everyone’s been decrying the end of agencies for over a decade. I’ve been building ad tech for almost a decade and everyone has always said cool you’re building software products so this is going to kill the agency, and no that has never been the case. Agencies have strengthened their position within the categories that we’ve operated in over the past decade. Despite us building really great software products, invariably the customers of those software products have always tended to be agencies.

Now, I think there are two competing factors here. One, brands in-housing is very much a trend. If you look at total US advertising revenue, brands in-housing their media is at about a third of total media spend and it’s growing in low single digit percentages and has been for the past few years to the detriment of holdcos. Holdcos have been losing their market share, independent agencies have been growing as well as brands in-housing has been growing. Now, you could make an argument to say that AI will be a catalyst to increase the trend of brands beginning to in-house. But even still, what does that increase look like? If it’s low single digit percentages does that get up to mid single digit percentages and end up being at 40 percent of US market share by 2028? That’s still not the majority of advertising revenue.

One of the things that I believe to be true is whenever there is high degree of complexity and high degree of new and change, brands hire experts to guide them on that journey of technological change and understand all of the variables at play. Whether that be programmatic media buying, retail media, marketplace commerce, agentic AI, brands will effectively hire an agency to perform the service of executing that media channel to gain optimal results and provide that elevated level of service for them.

If you think about the broader trend within agentic AI right now, many VCs are promoting the notion of AI-enabled services. Service businesses that have historically had to operate on services margins in which they need to linearly scale head count and team members as they scale revenue and customers could now theoretically exponentially increase customer and revenue while maintaining the same team, thereby operating more like a software company. That might be another tailwind that leads to agencies continuing to proliferate is that agencies are able to operate on a significantly higher operating margin and will be valued at a significantly higher multiple because of that margin.

Our perspective here is that we build great software products to enable our customers to achieve their goals, whether that be agencies or brands. It just so happens that we’ve started building on the Amazon DSP, which is largely controlled by agencies, and agencies have put up their hand to hire GG because they realize that GG is simply able to dramatically reduce the operational cost of managing this media channel and immediately improve advertising performance. I’m not here to say our AI is better than this person’s AI for the advertising performance piece, but rather it’s actually quite simple. Heads of advertising at many of the agencies that we work with have effectively written playbooks to buy media and report on media, and these playbooks are often constrained with human hours in mind. The human hours that actually get allocated to this are probably a fraction of someone’s job.

So what we say to our customers is hey let’s rewrite these playbooks unconstrained by human labor. Rather than thinking that a human can only allocate two hours of their day to this, let’s assign this work to an AI agent unconstrained by human labor in which AI is always on, always executing, and rather than humans arbitrarily deciding numbers for bids and budgets and bid adjustments, let’s allow an AI agent to use a proprietary ML model to decide those numbers with a higher degree of precision and accuracy than a human ever could in a manner that’s always on. Generally speaking across our customer base we just see immediate advertising performance improvements when they begin to adopt GG because GG’s able to operate at a level of scale and precision that a human ever could. That’s why agencies hire GG and we’re really excited with where we’re at and to continue to grow.

ES: What’s next frontier? You started with Amazon, where do you go after that?

AE: If you think about who our customers are generally speaking, programmatic media buying teams at large agencies. It’s in our best interest to ensure that we are touching as many surface areas that that team operates in. Now that we’ve been in market for over a year, a head of media at a large agency will now go to us and say okay I get it, we’re on board, GG has transformed the way that we work. But GG has unfortunately only transformed the way that we work on this one media channel and our teams operate on others.

If we think about where the teams that use GG right now operate, it’s the other enterprise DSPs that we spoke about. DV360 and Trade Desk are generally speaking the other surface areas that our teams log into and so very naturally we’ll begin to expand to those other enterprise DSPs so that GG becomes this orchestration layer across all three DSPs for any media agency and brand.

ES: Going back to this idea of how this wave of disruption is unmanageable for an in-house marketing team and so it actually pushes them into agencies, is that a short-lived or a persistent effect? You could make both arguments. It’s short-lived because they just need to get comfortable with the application of AI, and once they do and there’s all these SaaS tools they’ll be able to bring that back in-house and there’s less reliance on an agency. Or is it that now once we’ve given this task or this project to the agency, there is almost like a lock-in because they have all the domain expertise and they’ve got our campaigns being managed in this way, so I’ve almost cut myself off from learning it and there is just this permanent dependence. Which is it do you think?

AE: I think it’s a little bit of the former. On the latter piece brands are smart and are able to straddle the lines so that they’re not dependent on their agency on a long-term basis. But getting back to that initial point, complexity breeds an agency strengthening their position with their clients and brands just continue to hire agencies across new degrees of complexity. I believe that AI and agentic AI will continue to be a tailwind for agencies at least for the next two to three years. Three years and beyond, Eric, we might be able to blink and buy a TV ad. There might be Neuralinks in CMOs in which you don’t even need a team in three years from now. I’m not here to opine on the long term, but at least in the near term I think agentic AI and all of this complexity and new technology shift will definitely be a tailwind for agencies.

ES: I saw a video on YouTube of a woman who had built this apparatus to control drones with her brain. It was fascinating. It was almost scary, I don’t want to say horrifying, but it was also very impressive. I’ve got two questions. Talk to me about shoppable TV. What is Amazon doing there and how is it progressing?

AE: Shoppable TV is a niche behavior and a low single digit percentage point of conversion rate. Shoppable TV would be buy a product, add it to cart or scan a QR code. I think the notion that shoppable TV will become this ubiquitous behavior across TV media buying is immediately dispelled when you understand the low degree of market share that it’s gaining even in Amazon’s TV buying apparatus, in which Amazon is by far the best possible channel for shoppable TV to become relatively ubiquitous. It’s very much a niche behavior.

We recommend to our customers that for every TV ad that they begin to buy against, they have a shoppable creative and let GG decide which creative gets shown the most based on conversion and engagement and audience performance. But it’s very much a niche behavior because we don’t engage with our TV in the same manner with which we engage on our phone. A common friend of ours, James Borrow at Universal Ads, said something controversial in saying that TV is actually the second device and the first device is the engaged device that we have in our hand. We don’t want to shop on that second device, but the second device is there, it’s changing purchasing decisions. I just continue to think that shoppable TV is very much a niche behavior. People will make investments in it because why not, but we’re not going to see it become ubiquitous anytime in the near future.

ES: That’s fascinating. I totally agree with all that, especially the idea of the TV being the second device. My sense is you don’t need shoppable TV in order to do credible attribution here. You’ve got timestamps, you’ve got IP address because someone’s at home on the couch, their TV’s connected to WiFi, their phone’s connected to WiFi. You essentially have nearly deterministic attribution anyway and someone’s much more familiar and accustomed to buying on their phone than they are with their remote control.

AE: Exactly.

ES: Let’s go to LiveRamp. Was that a defensive tactic against platforms like Amazon simply expanding their reach to non-platform inventory?

AE: I’m not an expert in Publicis’s strategy and LiveRamp as a product, but everything that I know about holdcos over the past three to four years has been generating profits on non-services revenue lines, particularly around principal-based media buying and proprietary audiences. If we think about Publicis’s identity spine, they have Epsilon, which is their audience graph, and they’re able to mark up CPMs across all of their media buys by enriching the identities of Epsilon and making it ubiquitous across all of their buys.

Now with LiveRamp, they’re likely going to have the ability to collaborate first-party signals across brands’ data, across retailers’ data, across a variety of other multi-tenant first-party data collaborations and mark up the fees that they would charge by creating proprietary audiences against it. Effectively what this does is it strengthens their existing data layer across publishers, across media buyers, across brands to further monetize across principal-based media buying and proprietary audiences and that’s my belief of why the LiveRamp acquisition was strategic for Publicis.

ES: Adam, this was fantastic. You’re like the canonical MDM podcast guest because we just get to go really deep on a topic and everyone gains a lot of fantastic insights. Talk to the audience about GG. How can they interact with GG, how can they hire GG?

AE: If you are an agency or a brand who’s buying Amazon DSP ads, we’d love to talk to you. Go to ggcode.ai and hire GG and you won’t regret it.

ES: Adam, have fun at Cannes.

AE: Thanks, Eric.

Comments: