Growing Ecommerce – The Retail Growth Podcast
Google Marketing Live: Ads in an AI-First Search Experience
Join Mike Ryan for a discussion of his highlights from GML 2024. Was the event overshadowed by Google I/O? Do advertisers need to worry about the biggest shake-up ever to Google's SERP? He answers these questions and unpacks the news around core ecommerce feature announcements like profit bidding, interactive Shopping ads, and improvements to asset and video reporting in Performance Max campaigns. All in all it was a good GML for paid teams, but a worrying one for organic teams and publishers. Mike explains it all in this episode.
Join Mike Ryan for a discussion of his highlights from GML 2024. Was the event overshadowed by Google I/O? Do advertisers need to worry about the biggest shake-up ever to Google's SERP? He answers these questions and unpacks the news around core ecommerce feature announcements like profit bidding, interactive Shopping ads, and improvements to asset and video reporting in Performance Max campaigns. All in all it was a good GML for paid teams, but a worrying one for organic teams and publishers. Mike explains it all in this episode.
Welcome to Growing Ecommerce. I'm your host, mike Ryan of Smarter Ecommerce, also known as MEC, and today we're going to talk about Google Marketing Live, also known as GML. It was a very exciting GML this year for e-commerce. E-commerce received a lot of new features from Google, or has new features on the radar, I should say and it was also a very good GML for AI, as you can imagine. So we're going to dig into that. We'll talk about features that are planned, like profit bidding, also shopping and search ads, coming to AI overviews and Google Lens and some other interesting campaign details that were discussed at Google Marketing Live.
Speaker 1:I joined from the gate, catching a flight to Birmingham, so it was a really fun setting to watch EML and I hope you enjoy this discussion and remember if you enjoy this podcast, please share it with a friend, give us a review on Apple Podcasts or Spotify, wherever you listen, we really appreciate it. All right, let's get into it. So, as I mentioned at the top, it was a really good GML for artificial intelligence. You know, I could kind of save my thesis statement or my summary for the end. Maybe that would be a more dramatic effect, but I'm terrible at that, so I'm just going to get out in front here with it.
Speaker 1:I had the feeling that Google Marketing Live this year. I felt that the thunder was stolen a bit by Google IO the week before. I mean, of course, that's their big developer conference. Gml is their big marketing conference. You can see this one of a couple of ways. I guess you could see, maybe, how well their roadmaps are aligned together, like how much synergy there is or whatever buzzword you want to use, or just that you could see how single-mindedly Google is pursuing artificial intelligence at the moment. There's been a lot of criticisms about this strategy so far. I mean, one of the big announcements from IO, which by default becomes a big announcement for their marketing conference as well, is the fact that this kind of project or experiment, formerly known as SGE, is now rolling out to the broad public in a feature called AI overviews and basically there's going to be AI generated content on the main Google search results for all users worldwide, with time starting in the US at first, and people have been criticizing the inaccurate results and potentially dangerous results and is this actually what searchers want? And so on and so forth. There's plenty of discussion about that.
Speaker 1:So far, I think my impression is that Google kind of blanked here. I mean, this is their main revenue product. This is the core of what they do is search, and I do feel that they've adopted this a bit hastily and even if it has taken a bit of time, because the funny thing too is that a lot of things at GML were announced last year. At GML, I think they did a good job last year, kind of staking out the vision. They introduced new things like product studio that integrates or kind of combines generative AI with your feed and your images so that you can generate better product images. And it became clear soon after Google Marketing Live last year that AI would be coming to the search results and that there would be ads served in those placements, and so what happened this year was largely it's partly a confirmation of that. It's to some extent a fulfillment of that. It's moving from a more theoretical standpoint toward the actual roadmap.
Speaker 1:But yeah, this was all prompted by basically, openai and Microsoft, their partnership in particular, was really threatening. Satya kind of said that, more or less said he would make Google dance, and it's true because Microsoft Search didn't win any market share over the past year. Microsoft Search was super fast at integrating generative AI and it wasn't a needle mover for them. And yet Google seems to act on the premise as if it will be a needle mover for them now. It's an interesting assumption.
Speaker 1:I don't want to be too critical of Google here, but I feel like there was an episode where I spoke with De Reuter from Bullcom and you know he works at a. Bullcom is a big company and it's small compared to Google, but he talked about the way that they make decisions. It's like steering a freighter and you know everything is by degrees, everything is gradual, and this is kind of the risk of a classic innovator's dilemma for a company the size of Google that they do become risk averse, that they do become fossilized or rigid. Don't want the opposite either, because an organization can become sort of brittle, in a way, and not quite as spry, flexible, nimble as they used to be. So you need to find that's the dilemma. You need to find the middle ground, and I feel like the pace and extent at which they're adopting AI now is potentially going to be challenging for them.
Speaker 1:But let's get into how this shakes out for marketers in a bit more detail. There's a whole big question marking over organic search. Maybe I can bring someone on who knows organic search better than I do. That wasn't really the topic of Google Marketing Live. It's all about paid media, and so they show, speaking of the fulfillment of past visions, show these kind of Google Glass-like experiences. Ironically, I think it was running on Meadows Ray-Bans, which is super funny, but not on their own hardware. But yeah, you know where. Basically, you know you can do visual search with your phone or with glasses, whatever the case might be, and just have shopping integrated into that experience, and I think that's the fulfillment of a longstanding wish for Google. So that's something where I feel like, okay, they've had this vision in place for a lot longer than Gen AI was around or whatever is going on with Microsoft or OpenAI, and that feels more true to the spirit of what they want to achieve.
Speaker 1:And they talk about features like circle to search, which is pretty cool. You just have an image and you, on your smartphone, just draw around it with your finger, whatever part of the image you want. Let's imagine it's an outfit. This is, of course, their fantasy that you take a photo of someone's outfit and you circle the shoes or circle the skirt or the handbag or whatever, and then they can serve you listings of similar products to that, and that feels to me more connected to users and somehow plausible. They claim that the Circle behavior they got that inspiration from users, so I think that's something that's really cool. When it comes to the AI overviews, yeah, we'll have to see.
Speaker 1:I think it's a really interesting time because the way Google search marketing paid search marketing, of course originated is like keyword targeting and later on, with shopping, you're not targeting keywords but Google is handling this matching between queries and products, so it's still very query driven. We're moving to a point here if people really adopt the technology in the way that Google imagines, we're moving to a point where people will have these very lengthy prompts and they're having a conversation with the AI and you can't distill it to keyword targeting and it comes down to, potentially, to things that could never be targeted by keywords. Let's see like Google will handle the matching there. They'll handle the serving. This will be served through normal e-commerce campaigns like shopping performance max. I don't know. I still feel that ultimately, despite all of the noise, let's imagine there's some lengthy sentence or experience, that conversation that they're having with a search engine or with a chatbot, whatever the case might be where they're doing this kind of extended shopping. In the end, it does come down to specific keywords, or you can distill something out of that, I think, to match products that Google will need to do that.
Speaker 1:The question is, yeah, is that something that's exclusively the domain of AI and the platform? Could humans do this? Would it be worthwhile for humans to do this? These are the questions, but I think what will be really interesting is is the volume going to be as significant as imagined? How will the performance be on this type of user and this type of traffic? Is it going to be more intentful? Because the user is doing a lot of qualifying and they're going to be really ready to convert it, so you'll get this really nice high converting traffic, whereas maybe otherwise, if they just click through a standard search result page and a standard shopping ad, maybe they'd be a little further up in a browsing phase and your website would have to do more heavy lifting. These are very interesting things to imagine about how this will play out, but it could lead to higher click-through rate. It could lead to higher conversion rate and we'll see if this is priced any differently than normal inventory. And then the big question is will we know any of this? Is there going to be segmented reporting available?
Speaker 1:When we look at the way that Google has approached reporting for Performance Max, for example, they've typically had not a lot of segmentation available. There are scripts that offer workarounds to this, but you can't really see where your ads are serving. When they do break out PMAX placements specific placements they end up just having the impressions and no other performance metrics. Or, similarly, if you look at the search term report to understand what searches triggered ads for Performance Max, you'll see things like the click-through rate and the conversion rate, but you won't know what it costs you. You won't have any cost-related metrics, so you can't know, like, okay, what was actually my return on ad spend for this search term or for this category of traffic, et cetera, or what was the cost per click. So Google is very much offering these kinds of insights on their terms, on their terms, and I think when it comes to, yeah, these AI overviews and these new types of placements, there's going to be a lot of pressure from advertisers. I think there's a lack of trust, like, okay, pmax was building on all kinds of things that people knew about and were comfortable with and existed before, and maybe you could get away with less transparency. But this is an entirely new category of placement, seemingly, and I think advertisers will want to know where their money is going and how it is performing in regards to that. But Google has gotten that message.
Speaker 1:I want to talk about another announcement from GML. This was my personal highlight, I think for a lot of people, this slide came up that said like more reporting and controls for performance specs, something like that, and the crowd burst into spontaneous applause. The presenter seems a little caught off guard. Yeah, but it shows you what the market actually wants. Marketers want to have control and transparency into their ad buying. But we were just talking about how limited some of the performance max reporting is and there's now going to be asset level reporting. So this is assets are like the Lego bricks that go into creatives and the creative is like a little cluster of Lego bricks. So assets we've previously not had reporting at that level of detail before. That's new. It's exciting because it does open the door for some level of creative optimization.
Speaker 1:I'm hearing rumors I don't know if this is confirmed or not but that there's only going to be like conversion rate and conversion value, like revenue, that will be reported here. I don't know how many metrics will be there, wouldn't surprise me. But another thing in this category is that YouTube placements there's going to be reporting for that, as well as support for exclusions. So I think these are really positive things. Google is listening. They're not unmovable on these topics. It's similar. In the past, they offered these limited search term insights for Performance Max. They offered brand controls for Performance Max. So they are listening and they are making compromises. I won't say that they're giving advertisers exactly what they want, but they're meeting in the middle.
Speaker 1:And we just mentioned this topic of brand exclusion. Another very cool announcement from Google a big criticism or sticking point with these creatives and assets, especially like automatically created assets or, for example, assets that you might make using product studio, which we mentioned before. It's basically an AI powered photo editor, which we mentioned before. It's basically an AI powered photo editor. These things you had no brand control over them and so you just have to prompt and in the case of like product studio, and then filter for the ones that seem kind of on brand or not. But now you can supply Google with like fonts, brand colors and some other basics so that it should produce output. That is more on brand for you.
Speaker 1:And again, there's an age-old debate in creative optimization, which is not my specialty, but the idea is basically like yeah, are branded ads actually just for the gratification of the CMO or do they perform better? When and where is the appropriate point to really brand, or how much should you just be focused on good quality creative? You know, famously some folks out there are saying that ugly ads perform the best, not pretty ads. And at a slight digression on that point, I was just reading there's some interesting new research suggesting that even like an experienced creative or like an experienced marketer, their picks on what creative is going to perform and what's not is only like 2% better than random chance. And some would argue on the basis of this research that the most effective way to put your creatives out there is to put a lot of creatives and measure how they're performing. I believe John James, a friend of mine online, was making a statement like that, and that's actually exactly the philosophy of Google here. That, and that's actually exactly the philosophy of Google here.
Speaker 1:They're trying to generate a lot of creative procedurally as well or through AI. They want you to submit a lot of creative so that you have a high ad strength controversial metric that indicates basically how well you're supporting the kind of liquidity of assets that the algorithm wants, and they're going to mix and match and your CMO might not like how the ads look that serve with Performance Max, but in theory they should be the most performant ones. So it's an interesting bit of research that, wherever you fall on this topic. I'm not a creative, so I'm more agnostic to it. I'm sure there are some creatives out there bristling, but it's an interesting view on that. So I feel like we've been moving at a brisk clip here. I don't know how that feels to you as a listener, but just to recap, we talked about how shopping ads will serve in AI overviews, which is known for about a year at this point, but it's actually going to start happening now, as much as we know that there's going to be more detailed reporting and controls for Performance Max, including these asset level reports, youtube placement reports, youtube exclusions, and that you're going to have brand control over AI generated assets as well. So that's a really nice blend of new opportunities and ways to reach and also new controls. So I liked all that so far.
Speaker 1:Say what you will about AI overviews and if it's for the good or not for the consumer experience. I think advertisers are kind of set here. They don't have to worry too much. They'll reach those users and the market will speak for itself. Either people will adopt this technology or not. If they do adopt this technology because it works and they like the results, the ads will be present. If they don't, then your ads will be present in standard search results too, and Google will adjust the mix accordingly, because this is like something like 80% of their revenue and they can't mess with that. So I think, as an advertiser, there's nothing to be overly concerned about.
Speaker 1:I want to talk about at least one more thing while I have you, and that is profit bidding. So, yeah, this is a big one. This is on a lot of people's wishlists for a long time. Just to summarize the way that Google Ads works it optimizes in an e-commerce setting. It will typically optimize toward conversion value not toward conversions, but towards conversion value, and most commonly, that value is revenue, although it can be anything that you want to put in there. So then, if you take a really popular efficiency metric like return on ad spend. That's going to be the ratio of your conversion value, whatever it is usually revenue, the relation of revenue to ad spend, and it's been popular for a few years now for people to, through different solutions, bring in profit into the system. So that way you're bidding with the relation of profit to ad spend instead of revenue to ad spend, of profit to ad spend instead of revenue to ad spend. And the other way of kind of handling that has been a workaround, which is to basically create buckets like what margin class is a given item in is it a high, medium or low margin and supply that via the feed. But of course, actually optimizing toward profit is a much more direct route of doing that. So I'd have to check.
Speaker 1:But I guess probably three years ago anyway, google rolled out an interesting kind of extension to a standard tracking, your conversion tracking, and it's called conversions with cart data or CWCD. You'll see that acronym that's right, another acronym Rolling around in the future. I promise you that. So the premise or the function really not the premise the function of conversions with cart data well, it's about that cart data. It will tell you about your order details or the composition of the order, which items actually got sold. And this is a really interesting data set.
Speaker 1:I've used this in the past for myself because you can look at normally like an item will get clicked let's say it's a $70 shoe and then whatever happens after that, it's all going to be attributed back to that clicked item. So that person might buy the shoe in two sizes because they're not sure about the fit, and it'll double up the order value to 140, of course. Or that person might buy an entirely different brand of shoe or a different category of shoe. They might buy that shoe plus shoelaces and insoles. You never know what's going to happen. But all of that revenue or profit gets attributed back to that product that was clicked. And by implementing you get to understand that. And this is very cool in reporting because you can look at these substitution effects in terms of brand substitution, product type substitution, item ID substitution. You can see it all there and here's where it comes around to what we're talking about.
Speaker 1:Profit bidding. You can also, like that's on the tracking side, is CWCD. You can also in your product feed. There's an attribute called cogs for cost of goods sold, of course, and you can submit the cogs for every item ID and at that point Google is able to not only understand the composition in terms of what's actually being bought here, but also the margin composition of that. That's where it gets pretty interesting. So when you combine those tracking, that tracking implementation and that cost of goods sold attribute, you can enable going forward profit bidding. That's for performance max as well as standard shopping campaigns.
Speaker 1:And you know, implementing profit bidding is very exciting. But also it's a big change in any account, because in the past the algorithm was optimizing for certain order values, and when you're then arbitrarily kind of or from the algorithm's perspective changing those values because your margin will act effectively as a multiplier and you can have products in the same category but with different margins and all kinds of different scenarios can come up. So suddenly, from the algorithm's perspective, some products are being hit with this. You know, a 10% multiplier and others with a different percentage can be confusing for the algorithm to understand what's going on, and so learning periods become a concern. In a topic like this, how is this going to affect my account performance? And now Google is promising that this should be pretty seamless. I think that this is credible for two different reasons. I mean, they say that it shouldn't overly affect learning when you implement it. And also you can then switch back and forth between revenue and profit bidding seamlessly, and I find that even more credible.
Speaker 1:But let's talk about just implementing it in the first place. The thing that happens when you on your own, let's say, use server-side tagging kind of solutions to implement this, or offline conversion, upload different ways of getting this data in there, google doesn't know that you're doing that. It doesn't know what happened to the account and why. It just knows that all of a sudden values changed a lot and by different amounts and so on. So that's quite confusing. But when you're activating a feature, google is aware of this and they can perhaps mitigate that to a certain extent because they understand the change that is occurring. That's significant.
Speaker 1:Then, when it comes to switching back and forth between profits and revenue, I mean I don't really know the use sense it has that data. It doesn't have to guess or model the effect. There's no counterfactual, it's tracking both and then you're switching. This is my understanding of why that's seamless to switch. That's why that makes sense to me From the timing. I think it's good timing on Google's part. I think that's seamless to switch. That's why that makes sense to me From the timing. I think it's good timing on Google's part. I think that's the growth at all costs kind of mindset. Let's say yeah, anywhere from maybe 2019 through 2021, early 2021, that kind of a mindset is going away. People are a lot more concerned about first order profits, reaching profit immediately, not making, not sticking their necks out with these lifetime value bets. The environment is just too challenging right now for most and they need to have that insured profitability. So I think from that perspective it's really attractive.
Speaker 1:We've seen a lot of advertisers setting up the conversion with cart data in the past, although it was a bit kind of in want of a use case, and you could set up some advanced reporting. If you know how I I don't know how many people really took advantage of that configuration, but the big sticking point to me, or the what it all hinges on, is the cost of goods sold, because you have to be willing and able to share that with Google and I think that there probably is still going to be a place for some of these third-party tools like Profit Metrics, just putting a name out there. No affiliation, because it's a server-side setup and you can also do your own server-side setup but you're not sharing some of your most sensitive data with Google. My kind of concern is that Google has a product graph it's called the shopping graph. You know they they look at e-commerce products across a lot of different advertisers and they're they're able to kind of connect the dots, you know, especially with high volume products. Then I can really drill it down to the item level. I'm sure also that you know they make categorical connections and all kinds of connections. But in principle, in theory, there's nothing that would stop Google from making a pretty realistic guess about what your margin would be. Because if they have enough advertisers submitting COGS and they've got that at the item level and they can also look at the brand level and the category level and they can kind of map out, let's say, that interquartile range or the median, they might in principle be able to segment that by vertical and by market and by company size who knows or by budget. They could potentially be able to make a pretty realistic guess anyway about what your margins are.
Speaker 1:I don't know if Google is motivated to do something like this. People get into this tinfoil ad area when it comes to sharing their profit data with Google. I've asked advertisers if they're willing to do this and twice I had people laugh in my face and I wasn't even advocating for it, I was just asking. I wasn't even advocating for it, I was just asking. But there wasn't such a clear use case back then. Now there's a clear use case and we'll have to see how much this does or does not motivate people and what Google can or will do to kind of win trust here as well, if they'll offer any kind of special security assurances or it all remains to be seen. But in the back of my head I've thought this for a longer time actually is that if enough people do this, then it wouldn't necessarily matter anyway. Particularly, you know, let's say that there's a map bound like a minimum advertising price bound product and a given vendor or brand is, you know, they're going to have a typical cost cogs. It maybe varies a bit from advertiser to advertiser, but in the end those businesses will be pretty range bound in terms of how much profit they're generating and I think there's certainly cases where it could be rather easy to surmise or where median profitability would be pretty accurate.
Speaker 1:So last thing that I've got for you today, I want to circle back to the generative AI theme, because certainly Google kept circling back, so I'll take the same liberty. Someone was counting how many times they said it in like the first few minutes. It was a lot. But we talked earlier about some of these AI-supported placements, like Google Lens, in terms of the visual AI that empowers that, or AI overview placements in terms of this chat GPT-like for lack of a better word or this large language model, these automated answers that will be embedded in the search engine result page. But that's not all they had on offer for generative AI and for these kind of supportive features.
Speaker 1:So I want to also highlight a couple of cool ones. Coming to shopping ads, there's going to be a 3D spin that's available, where basically they're going to be able to connect the dots between pictures of, like, different angles of an item and they'll be able to generate what that looks like in 3D and then it's going to be there in the ad itself. So this is a cool move toward more interactive PLAs or product listing ads, where you can actually then give that thing a spin. They're also they've tested virtual try-on in the past, which is in the direction of augmented reality, but you basically, in the first place, it's about fashion and you select, for example, a model with similar size and then you can they're going to use AI to show what those clothes, how that would fit on the model, and so you can get a much more realistic picture there, and I think these are real positives both of these because they just help to qualify users. I talked earlier about how, hopefully, people who are interacting with AI will be more self-qualified or more AI-qualified, and it should drive good conversion rates on your site when they land there. I think that spin and try-on are just more examples of, because there's just this disconnect between this digital or virtual environment where you're shopping and then the fact that you're ordering physical goods, so I think those should be really positive.
Speaker 1:They're also adding video highlights, which is another cool thing in the lines of interactivity in these product listing ads, where you can, yeah, circle back to what we said earlier, this debate between branded assets and ugly assets, so to say. But you can go either route you can have these nice branded video highlights right there appearing in the search engine result page or not, to say that user-generated content is ugly, but or you can go for UGC and if you've got proven winners from either of those categories can go for UGC, and if you've got proven winners from either of those categories, more branded type stuff, more UGC doesn't matter. You can use them both and you can add those so that they'll appear as part of the ad unit, which is a pretty cool innovation when you take those things together. So very supportive of fashion in the first place, but fashion is the largest vertical, so that's fine. But also coming to like beauty soon with virtual try on, which I think is great for skincare, makeup, cosmetics. So there's one more thing which is pretty cool, changing about how these ads can appear.
Speaker 1:Google has been working on new customer detection for a while now, partly with your first party audience lists, which will should be easier than ever to use because they're also bringing their clean room on board. That's going general availability it's called Google Ads Data Manager, but a way that you can then work with that first party data. What everyone's always talking about is distinguish new and returning customers and then you can offer discounts exclusively to new customers. So that should again be something that can help actually drive the incrementality of these channels to give you a leg up compared to competitors and make sure that you're giving targeted discounts instead of just blanket discounts and you're discounting for your returning customers, so these are all really cool innovations. I think that's all stuff that I have to say about Google Marketing Lab. That's really positive.
Speaker 1:Just to revisit my thesis statement from the beginning I don't know how lucidly I stated it back then, so maybe now I can say it more concisely there's an extent to which a lot of this feels like things that they announced last year at Google Marketing Live, that they announced last year at Google Marketing Live. There's an extent to which this feels like Google IO stole their thunder a bit. But the feeling that I have is that it's sorry, but this is the best metaphor I could come up with, so forgive me. It's like an anaconda swallowing a deer or something like that, or a crocodile, whatever eating a really big meal. And that's what happened. Google took a big old bite of the AI pie last year at Google Marketing Live and I feel like they've been spending the year digesting that. Some would say. Maybe they've gotten some indigestion, some acid reflex, acid reflex heartburn, because not everyone's totally satisfied with these AI features. I'm trying to remain neutral on that, but they're still digesting and maybe now we're getting to the point where this stuff is actually going to start hitting the market. That's what they're telling us, so I think there's nothing to do but wait and see.
Speaker 1:I don't think advertisers should be particularly concerned. I think this technology is generally a lot more concerning to publishers and to people who depend on organic traffic. I would be more worried about that. So that's, of course, another part of your business too. I'm sure and yeah, certainly there'll be discussions about. Does this just increase dependency on advertisements? Is it right or wrong that the AI is training on people's data and then auto-generating answers and then the only way they can stand out is with an ad? There's plenty of articles being written about that by publishers because they're upset, so I'll let you read those and draw conclusions there.
Speaker 1:But that was it. That was my GML recap. There's way more so if you didn't catch GML recap from Google's blog, or you didn't watch it or you didn't see the millions of posts on LinkedIn. This is just my two cents. I'm not going to talk about every single thing that they announced, because it's just too much and it's all been talked to by now, but I wanted to show you the things that popped out to me and what I think about them. So thanks for listening. This podcast is produced by Smarter eCommerce. To learn more, go to smarter-ecommercecom. And again, we really appreciate if you support us with a review, with a referral to your friends. This all helps get more people listening to the podcast. Thanks again, see you next time.