Subscribe
Share
Share
Embed
10 minutes of expert insights every weekday. Your morning ritual for staying ahead in retail media.
Ep 4 - Are AI Ads the next big ad surface? How retail media can thrive
===
[00:00:00] Kiri: When Open AI first launched its ad experiments, there was a little bit of skepticism around the business model, but none of these platforms can run forever without monetization ads. Coming [00:00:15] and the real questions are where they appear, whether users still trust the answers.
[00:00:19]
[00:00:21] Kiri: This Expert series is sponsored by Miracle, and again, I'm joined by Amelia VanCamp, head of a agentic Commerce at Miracle, to [00:00:30] unpack what a agentic advertising really means and how retailers and brands can prepare. Without guessing formats. So, , Amelia, you say on a, on our prep call, you said the other [00:00:45] LLMs, like Claude mocking open AI for their ad product made you giggle. Why was that?
[00:00:50] Amelia Van Camp: Yeah. , I think what open AI has done and is doing in their tests is very intelligent and also not particularly surprising. At the [00:01:00] end of the day, these are corporations and they have a monetization strategy to any of the avenues that they pursue. Advertising is a very valuable monetization strategy.
[00:01:13] It shouldn't be [00:01:15] surprising given how historic shopping has worked both in store online, that we're now starting to see ads pop up inside of an LLM experience. I think many users. [00:01:30] We're paranoid and still are paranoid that it's going to dilute the experience. But where I think OpenAI has done an exceptional job at setting the precedent early on is they're not letting those ads impact the result and the answer [00:01:45] that's populated inside of the conversation, they're placing it as a sponsored placement at the bottom.
[00:01:51] That means it's not diluting the trust that the user has of the LLM. Thinking that the answers that it's getting that we are getting [00:02:00] are funded by ads, but instead seeing that the answers are still true, but there's also a sponsor placement at the bottom. I think the key is what is the ad that's at the bottom and how relevant is that to you as a [00:02:15] user, and even more so, how relevant is that in the conversation that you're having with the LLM?
[00:02:21] Kiri: Right. And, and so do you see that as a bit of a risk with what we're seeing right now is that the ads won't be super [00:02:30] relevant to the conversation that's, that the user's having.
[00:02:33] Amelia Van Camp: I think it's a possibility. Do I think that LLMs would, would ignore that possibility? No, I don't. I think the way that they approach. The [00:02:45] world with their technology is relevancy on steroids in any avenue that they're pursuing, right? It's relevancy to you as a user. It's relevancy based on the conversation that you're having, the context that you're providing.[00:03:00]
[00:03:00] I would expect the same experience inside of advertising Where I see the, the pressure or the onus of responsibility coming into play is not just with the LLM setting the system up to be able to [00:03:15] hopefully offer a very similar experience to what they've already done inside of, you know, chat, GBT, perplexity, whatever it may be.
[00:03:22] Uh, but now it goes back to the advertiser. Who is placing that bid for that sponsored placement? Now [00:03:30] it's a question in my mind of if relevancy is the topic here and it's that relevancy on steroids, can you meet that need? And if so, how did you meet it and. Thinking back to [00:03:45] some of the earlier episodes where we've discussed just core concepts that we believe at Miracle are critical to participate in ag, agenta commerce, this notion of G-E-O-A-E-O optimization comes back [00:04:00] up.
[00:04:00] Where you're adding in these intent-based attributes that provide a ton of context for a product that are situational to how a user could be searching for a product [00:04:15] or set of products similar, if you can bring that into the advertising space. My hope is that the LLMs will accept that because it's going to create a better experience for advertisers and for users on their [00:04:30] site, and also for advertisers.
[00:04:32] I think it creates a much better experience to really start to connect the dots on some of these core concepts and frankly, technologies and strategies that you're investing in to not have it be. [00:04:45] Piecemeal. This is one thing that we do for ads. This is one thing that we do for our conversational commerce agent.
[00:04:51] This is one thing that we do to help with search and discovery. It really starts to bring, you know, a key theme back to all of those different [00:05:00] arenas. So my hope is that, uh, this whole notion of relevancy on steroids is something that we're really adopting and seeing, uh, as a, a leading factor inside of advertising on LLMs.
[00:05:13] Kiri: Yeah, so on, on that [00:05:15] point, the, the relevancy on, on steroids. Love that art. Trademark. Um, and we, we talked a little bit about these intent-based attributes and how important they are back in episode one of this series. What are some things, some practical [00:05:30] things that you're advising retailers on to get ready for this brave new world?
[00:05:37] Amelia Van Camp: So first and foremost, the world that you're already playing in today, data rich wins the race. It always has. It's [00:05:45] not a new concept. That concept still exists again, and it's all inside of that. Trust concept. How can you make sure that LLMs can trust you so that they can populate you back into the conversation?
[00:05:59] And how do you ensure [00:06:00] that your customer, who hopefully will end up converting and purchasing your products, can create that flywheel of trust as well to to come back time and time again, whether it's on your site or the LLM. So data [00:06:15] becomes your biggest friend inside of this race. How you think about data.
[00:06:21] Again, cut it into core data sets that have already been a standard requirement to play inside of e-commerce for years now. [00:06:30] And then add on top of that this metadata layer that includes intent-based attributes to be able to play inside of Ag agentic. I think those two concepts, one is likely something you're already [00:06:45] doing.
[00:06:45] I think assessing it. How robust is your data today, your core set of data? What systems are you using? Where do you feel there are gaps? Do you frequently have issues with pricing anomalies? Do you frequently have issues with [00:07:00] core data sets missing in a catalog? Start to really look at that and assess how strong you feel you are in that core set of data.
[00:07:09] And then from there. Start to look at what are these [00:07:15] organizations who offer solutions to solve for intent-based attributes? What concretely is under the hood? How do they do it? Do they do it for multiple LLMs? Do they do it for multiple models? Is it asynchronous? Really start to get into that, [00:07:30] that mindset of, Exploration. I think everyone is very exploratory right now and, and wanting to experiment. But make sure that A plus B equals C. With any of the vendors that you're looking at, make sure that what [00:07:45] they're showcasing to you is real and that they have facts to back up the claims that they're making so that you don't end up in a clunky system where one side of your data is really strong and the other is a bit [00:08:00] weak.
[00:08:00] Kiri: Would it be, would it be fair to describe what you are currently doing at Miracle as a orchestration layer across those different components across the product catalog ads? How does, does. At work, how would you describe it?
[00:08:13] Amelia Van Camp: Yeah, we do actually [00:08:15] describe it as an orchestration layer. Um, again, our history is in multi merchant. Order orchestration. And so we've had this backbone in how we've built our products and how we've built our expertise of thinking through [00:08:30] not just one single set of information and how you handle that, but multiple sets of information that are handled by different parties.
[00:08:38] How do you manage that? As an orchestration layer, and so inside of the world of AgTech, the [00:08:45] beautiful thing is we have many of our capabilities that already hit to these points of, you know, core e-commerce practice that are required. You're able to do that in an environment that is. [00:09:00] Open-ended inside of ag, so for multiple LLMs, for multiple models, which is new for some retailers in terms of how they play in the landscape, and you're able to plug in the ENT components that are necessary for [00:09:15] the here and now on this topic of discoverability and ranking ag channel connection.
[00:09:21] Really doing that inside of an orchestration layer. Where you can see the full flow, transact the whole way through [00:09:30] and ensure that from soup to nuts, you're well covered, not just on data, but you're well covered on a process and a system that's going to enable you to meet that customer's needs, whether it's [00:09:45] on an LLM or perhaps on your marketplace.
[00:09:48] Kiri: So one area that. brands that sell to retailers struggle with is how can they participate here? And especially in the CPG category, a lot of these brands [00:10:00] 99.9% of their sales go through retailers, they do very little themselves, and they're wondering, what can we be doing to help push our retailer partners along or give them better content?
[00:10:12] Powerless in this equation? [00:10:15] What should they be thinking about?
[00:10:16] Amelia Van Camp: No, they're certainly not powerless. I actually think they've been given an amplification, uh, a bit of a loudspeaker inside of ag agenta commerce with LLMs. And the reason I say that is if you think about a [00:10:30] brand, a brand has a reputation that it's standalone. To the retailer that, or the set of retailers that it may distribute its products or sell its products on.
[00:10:39] And that brand is something that is considered by the agent [00:10:45] that is determining what products, what brand to surface back to the user in that conversation. And so it creates an onus for brands to think the same way that retailers are having to think today [00:11:00] about how do you present yourself. To frankly, the digital world.
[00:11:04] Full stop. What is, you know, if you have a subreddit thread, what is that conversation around your brand? How can you contribute? How can you, [00:11:15] you know, in some cases, influence that What assets do you have on YouTube? Uh, even Wikipedia, which every time I say that I feel a bit like I'm aging myself talking about Wikipedia, but it is a data source that these agents scrape from.
[00:11:28] What does your Wikipedia [00:11:30] page say about you? These are independent.
[00:11:32] Kiri: pr,
[00:11:33] Amelia Van Camp: Exactly, exactly.
[00:11:35] Kiri: information?
[00:11:36] Amelia Van Camp: These are things that are. Outside of the control of the retailer that you're selling your products through, but they [00:11:45] are inside of your hemisphere of control. And so starting to think about again, specifically for agentic, how do you influence those?
[00:11:53] What is your plan to influence? How do you want to populate inside of an LLM? [00:12:00] If you are a brand that maybe doesn't have a strong. Direct to consumer site where you sell all of your products. Maybe it's not, you know, it's 5% of where your sales come from. Start to think about how you [00:12:15] can really feed that information both through the retailer.
[00:12:20] But also on your own site to optimize for your products showing up in the first place. So these themes of consistency, these themes of like [00:12:30] data rich, data integrity, are your responsibility as a brand as. It's also questions you can be asking the retailers that you work with. How are you handling intent-based attributes?
[00:12:42] Have you done a front end health assessment to make [00:12:45] sure that your front end is readable to agents so that I feel a bit better about knowing that my products are surfaceable. In your environment, what is your agentic strategy? Have you launched a shopping agent? If you have, how is [00:13:00] your shopping agent determining what products.
[00:13:03] To list back to the user. So I think the, the name of the game is think still very much about your brand strategy and certainly work with the [00:13:15] organizations that you're distributing your products through or selling your products through to really understand what their agent to commerce strategy is and how you can work to support that to ensure that you are also forward thinking.
[00:13:27] So intent based attributes, I'll say it [00:13:30] until the end of time, until it's not relevant anymore. Uh, but this is a key concept, not just for retailers, but also for brands. It's very much thinking about how your products can be surfaced inside of a conversation. When you [00:13:45] have those intent based attributes that are built in, that optimizes for your product, surfacing inside of that conversation.
[00:13:52]
[00:13:53] Kiri: Intent based attributes, relevancy on steroids. Those are the, the, the two things to take [00:14:00] away from this episode. It's about intent ready, product truth, so that targeting is. Relevant, and accessible. Thank you, Amelia, for joining me for this four part series.
[00:14:13] Amelia Van Camp: You are so welcome. This was [00:14:15] fun.
[00:14:15] Kiri: And if you're just joining us on, on this episode, go back and listen to the full series. We spent three episodes talking about discovery moving upstream. PDP is a homepage and all the shrinking [00:14:30] signals that are coming to the fore with AI enabled commerce. So thank you to Miracle for sponsoring this expert series, and I'll catch you tomorrow.
[00:14:39]