Latest / Retail Media Breakfast Club / Amazon Extends Its DSP Into ChatGPT Ads
Transcript
- Amazon extends its DSP into ChatGPT ads (Monday) === [00:00:00] Kiri Masters: Good morning. Happy Monday. Welcome to the Retail Media Breakfast Club. I'm Kiri Masters, and today I have another guest contribution, this time from Jo Lambojiva, who writes the AI for E-commerce and Amazon Sellers [00:00:15] Newsletter, a newsletter read mostly by people running businesses on Amazon as a primary sales channel. [00:00:21] And she published a analysis on the Amazon OpenAI advertising deal that was announced [00:00:30] last week, and I thought it was pretty sharp And rather than writing my own version, I asked whether she would be happy for me to share her POV on this podcast and in the newsletter [00:00:42] Jo's newsletter is definitely one that I [00:00:45] recommend as regular reading. We'll link up to her newsletter in the show notes, and let's jump into her analysis, and I'll follow up with a couple of points of my own at the end. Let's go. [00:00:56] [00:00:57] Kiri Masters: From the 10th of September, brands that [00:01:00] advertise through Amazon Ads can buy ad placements inside the ChatGPT app. [00:01:06] Now, that sounds like that sentence should be the headline, but it's actually the least interesting part of the story. The mechanism is [00:01:15] Amazon's demand side platform. It's DSP, the same console that advertisers already use to buy inventory on Amazon's own website and app, of course, but also inventory elsewhere offsite. [00:01:28] And the most recent [00:01:30] addition to those offsite destinations is these positions inside of ChatGPT. In practice, an advertiser can plan and buy ChatGPT placements using Amazon's shopping and browsing data to decide [00:01:45] who sees the ads. Now, it is a pilot. It's not a general launch. It's limited to a select group of US advertisers, with Delta Vacations among the first to take part. [00:01:58] They were named in the press release. [00:02:00] And this is a supply arrangement, one where OpenAI provides the ad space or inventory, and Amazon provides the advertisers and the buying pipe [00:02:09] But the structural signal underneath, this is the part that is worth your attention. [00:02:15] Because Amazon isn't trying to build a chatbot, it isn't competing to own the conversational interface where most consumers ask what to buy. It's doing something far less glamorous and more consequential, positioning itself [00:02:30] as the pipe that every advertising budget flows through, regardless of who owns the screen at the other end. [00:02:38] And ChatGPT is simply the newest screen on the list. To understand why this matters, we [00:02:45] need to zoom out from the ChatGPT announcement and look at the pattern. Over roughly the past 18 months, Amazon has moved its advertising business into streaming services, premium audio, sports broadcasts, and now [00:03:00] a leading AI assistant. [00:03:02] Different surfaces, different audiences, different content, same role for Amazon every single time. ~Amazon's DSP now stretches~ [00:03:10] ~Hm. Delete that. The- ~Amazon doesn't need to own any of these environments to profit from them. It just needs to own [00:03:15] the buying pipe, and the more surfaces that get routed through that pipe, the harder it becomes for advertisers to justify using anyone else's. So this isn't an AI strategy, it's an advertising aggregation [00:03:30] strategy, and the AI surface just happens to be the latest thing fed through it. [00:03:35] The two-part sales pitch. Part one, price. By several accounts, Amazon's DSP undercuts competitors on fees, sometimes [00:03:45] dramatically. On certain reserve deals, Amazon has reportedly charged close to zero against the 15 to 20% that independent platforms and Google have historically taken. Those figures come from a limited set of [00:04:00] reports and should be read as directional rather than gospel, but the direction is the point. [00:04:04] If two systems can buy the same inventory and one charges materially less, budget follows gravity. Part two, data. Because [00:04:15] Amazon is a marketplace as well as an ad business, it holds first-party shopping and browsing data that no pure ad platform can match. Amazon knows what people search for, what they compare, what they put in their basket, and then agonize [00:04:30] over for three days before either buying or abandoning in a fit of financial responsibility [00:04:36] Paired with cheap access to premium inventory, the pitch writes itself. Buying through Amazon costs less and targets better than buying anywhere [00:04:45] else [00:04:45] So far, that argument has largely held. Amazon's ad revenue has continued to grow, and some of that growth appears to have come directly at the expense of independent buying platforms [00:04:56] Now here's what this actually means for [00:05:00] sellers. Number one, concentration is deepening. The party increasingly positioned to route ad budgets into the newer surfaces is the same party that owns the marketplace many sellers depend on. It runs the retail media those [00:05:15] sellers already pay for, and it competes with some of them directly through its own product lines. [00:05:19] None of that is new, but each additional surface routed through Amazon's pipe makes the dependency a little deeper and a little harder to unwind. That's not a reason for [00:05:30] alarm, but it's a reason to price the concentration into longer-term planning rather than treating each channel as independent [00:05:37] Number two, measurement will decide the market. Whoever standardizes how performance is counted on AI [00:05:45] surfaces will effectively set the terms for anything built on top of them. The theoretical appeal of conversational advertising rests on the richness of intent people reveal when they talk through a decision in detail. [00:05:58] That appeal only converts into [00:06:00] real money if the intent signal survives the shift from organic conversation to commercially influenced, and if it can be measured credibly [00:06:09] Both of those remain open questions. And number three, watch the access threshold, not the pilot. [00:06:15] The most practical thing for operators is to track whether AI surface inventory eventually becomes available at lower minimums or at lower prices through the same console that sellers already use for existing campaigns. [00:06:29] If it [00:06:30] does, the question stops being structural and becomes operational [00:06:33] Bottom line, in the meantime, the smartest move is the boring one. Understand the structure, watch the access thresholds [00:06:42] And resist the urge to treat every new deal as [00:06:45] either a crisis or an opportunity. Sometimes it's just a pipe getting longer. Are manual processes slowing down your RMN [00:07:00] momentum? With GrowthLoop's composable commerce media solution, retail media and ad teams can build and launch precise audiences in minutes, activate across DSPs and ad platforms, [00:07:15] and deliver transparent ROI reporting that drives advertiser revenue. [00:07:21] learn how retail media leaders at Costco, Fanatics, and goPuff use GrowthLoop to deliver stronger [00:07:30] results for their brand partners fast. Visit go.growthloop.com/breakfast. That's [00:07:45] go.growthloop.com/breakfast And just wrapping up here, this is Kiri again with some postscript comments. Joe's takeaways here are pitched at small and medium businesses, typically with a heavy Amazon presence, and here's [00:08:00] how I'd translate them for enterprise brands and retailers operating a retail media network. [00:08:05] Number one, could other retailers offer something similar? Absolutely, and nothing is stopping them, and that door has seemingly been open [00:08:15] since May. OpenAI listed its buying routes publicly. There are four holdco agencies, plus Adobe, Criteo, Cargo, Pacvue, StackAdapt, plus a self-serve ads [00:08:30] manager in beta for advertisers of any size. [00:08:34] None of these partners have exclusivity And Criteo alone had over 2,000 brands running ChatGPT ads by August of this year. The big question [00:08:45] is what kind of participant any future retailers who want to jump on board with something similar want to be? [00:08:52] ~The reality is no other retailer has its own DSP~ [00:08:52] ~The reality is Walmart Connect is really the only other retailer with its own DSP, which was itself built in partnership with The Trade Desk, and that partnership, uh, officially~ [00:08:52] ~Reality is the only other retailer that operates its own DSP~ [00:08:52] The reality is Amazon is the only retail media network with its own DSP. There are a couple more that [00:09:00] are custom co-branded DSPs in partnership with The Trade Desk, like Kroger Precision Marketing, Target Roundel, Instacart, Walmart Connect and a few more [00:09:11] So any other retailers that want to go down this route are really [00:09:15] reliant on The Trade Desk [00:09:17] to actually build that integration [00:09:19] Now there are other routes here. Target, Albertsons, and Williams-Sonoma all have been in ChatGPT as advertising partners since [00:09:30] February, buying ads for their own products. Amazon is doing something else. They're bringing other brands' budgets into the ecosystem, and that requires owning your own DSP. [00:09:42] Number two, offsite has a new [00:09:45] meaning. Offsite is a major growth story for retail media [00:09:49] this is the place where a lot of projections about retail media growth are reliant on the on-site sponsored product ad ecosystem is... There's still some [00:10:00] opportunity there, but we're getting close to capacity, or at least close to capacity of what retailers are willing to put on-site. [00:10:08] And so off-site is the big growth opportunity, and now Amazon has added this new and compelling [00:10:15] surface bought through a console that most brand and agency media buying teams are already logged into. And so when a CPG brand is deciding where some of their incremental off-site ad dollars go, the [00:10:30] idea of just putting those ads inside ChatGPT is certainly an easy button [00:10:36] And in this case, the early bird gets the worm. If that is the first [00:10:41] And simplest access point for most brands to start [00:10:45] spending on ChatGPT ads, some of that budget is going to just stay there due to inertia later on, even if there is in the future a cheaper, faster, better way of buying that ad inventory directly. [00:11:00] And finally, Amazon strengthens its USP as a DSP. This was a point that Jo made in her piece. [00:11:07] Many retail media networks offer retailers specific buying interfaces or managed services as their only [00:11:15] buying access points, and the answer from most brand teams has been that they'd rather work in one place. And so every surface that Amazon adds to its DSP strengthens that preference. And [00:11:30] finally, this news is worth keeping in proportion. [00:11:32] This is a pilot with a handful of US advertisers. OpenAI's ad stack is still in its infancy, and pilots die all the time. So this is not a permanent [00:11:45] fixture in the landscape necessarily, but consider it another road sign on the path of Amazon and OpenAI's ad ambitions. That's it for today. Thanks for listening, and I'll catch you tomorrow [00:11:57]