What if artificial intelligence could revolutionize the e-commerce industry? Join us in our captivating conversation with Ralph Mayr, director of Product Management at Smarter Ecommerce, as we delve into the intriguing world of AI and its impact on ecommerce. From high-level concepts like philosophical zombies to current applications in performance marketing, Ralph shares his expert insight from both a consumer and product manager's perspective.
We tackle the moral implications of AI, including whether it should announce itself if it could be mistaken for a human being. Our discussion explores the state of Google search and its results, as well as the ads served alongside them. We examine the growing pains of new technologies like Google's Shopping Graph and conversational AI, and how these developments may improve search experiences for both consumers and advertisers.
Finally, we dive into the latest updates from Google Marketing Live, touching on conversational campaign creation, Product Studio, and more. Learn how chat-based assistance could revolutionize campaign creation and the implications of AI-assisted tools like Product Studio for product images. Don't miss our analysis of the potential long-term effects of these innovations and the trade-offs between time savings and performance. Tune in and elevate your understanding of the ever-evolving landscape of AI and ecommerce!
What if artificial intelligence could revolutionize the e-commerce industry? Join us in our captivating conversation with Ralph Mayr, director of Product Management at Smarter Ecommerce, as we delve into the intriguing world of AI and its impact on ecommerce. From high-level concepts like philosophical zombies to current applications in performance marketing, Ralph shares his expert insight from both a consumer and product manager's perspective.
We tackle the moral implications of AI, including whether it should announce itself if it could be mistaken for a human being. Our discussion explores the state of Google search and its results, as well as the ads served alongside them. We examine the growing pains of new technologies like Google's Shopping Graph and conversational AI, and how these developments may improve search experiences for both consumers and advertisers.
Finally, we dive into the latest updates from Google Marketing Live, touching on conversational campaign creation, Product Studio, and more. Learn how chat-based assistance could revolutionize campaign creation and the implications of AI-assisted tools like Product Studio for product images. Don't miss our analysis of the potential long-term effects of these innovations and the trade-offs between time savings and performance. Tune in and elevate your understanding of the ever-evolving landscape of AI and ecommerce!
Welcome to Growing Ecommerce. I'm your host, mike Ryan of Smarter Ecommerce, also known as Mech. Today I'm joined by Ralph Meyer, director of Product Management here at Mech, making his third appearance on the show. This time we talk about artificial intelligence coming from high concepts like philosophical zombies and the difference between appearing knowledgeable and being knowledgeable. Then we trickle on down to some of Ralph's own e-commerce interactions with classic search and conversational AI as a consumer himself, and lastly, we get into the intersection of AI and performance marketing. It's a great chat. I hope you'll enjoy it.
Speaker 1:And just before we start, i want to take a moment to teaser two upcoming episodes. First, i'll be speaking ahead of performance marketing in the lighting space. They've got brands ranging from commodity to premium, and we'll discuss how the post COVID boom bust effect led the business to need more insights and better efficiency. This was a trigger for great projects like marketing, mixed modeling and building a pipeline for profit data. And then I'm talking with the person responsible for leading the global ad tech and paid traffic efforts at one of the most iconic brands in the world. We'll discuss large scale product feeds, the way marketing becomes a CTO topic more than a CMO topic, and how the cookie list. Future is an impetus for horizontal initiatives That gets us into what are some of the ways that we need to change how we think and work post cookie. So I'm teesering this to you for a reason I'd love for you to subscribe to this podcast if you haven't, give a rating or review if you haven't, and, the best thing you can do, recommended to a friend or coworker if you haven't yet. Those actions support this podcast And they mean very much to me and the team.
Speaker 1:All right, let's get into it, okay. So, ralph, thanks for joining us back on the show. I think last time we had you, we were talking about like pricing and some macroeconomic factors that were going on. We discussed, like the war in Ukraine and some some really weighty topics like that, and you're one of my favorite lunch buddies. I always enjoy picking your brain on weighty topics. So thanks for coming back on. For people who haven't met you before, why don't you get us started with a quick introduction? What are your skills? What themes interest you? What are you up to?
Speaker 2:Yeah, thanks again for having me. Always a pleasure to have, on the one side, a lunch chat with you, but also to have something that's being actually recorded and broadcast to people. So let's see where this will take us today. Well, as you said, last time I was on, we spoke about macroeconomic trends and how we think those are going to affect consumer behavior, purchase of behavior, and how that might in turn, affect advertisers, and that is also one of the big themes that's interesting me at the moment.
Speaker 2:So in my current role, I'm the director of product management here at SMEK And that means I'm largely responsible for our product strategy figuring out, you know, what products people want to see in the future and how we're going to build those. And I know you might always refer to product management as sort of a mixture between building the right product and building the product right, and I do see it very much the same way. So my if I could describe it really briefly there's like three topics that I'm concerned with. One is desirability, so what type of product do people actually need? Second one is feasibility, in terms of can we actually build this, you know, given the resources that we have and the detectives out there. And third one is viability, which is can we actually turn this into something that's commercially viable as a product? And those things map back very nicely to the discussion we had last time about the macroeconomic environment, because that obviously influences what types of products and solutions advertisers actually need and what we SMEK should therefore provide.
Speaker 2:And I'm a personal background. I have actually a background in software engineering, so I come from a computer science degree, a bachelor's and master's And coincidentally, one large aspect of my master's degree was artificial intelligence and machine learning, and I think that's going to take us into another really interesting trend that has been affecting advertisers those last couple of months.
Speaker 1:Yeah, thanks for that introduction, And one other thing I'll mention about Ralph is that he's he's written a book as well, a work of fiction and no affiliate link on that. I just want to shout out what an amazingly versatile mind and hard worker Ralph is. But you also, besides all that you you have a really great blog, which is can be found at Ralph Meyercom, and you wrote an article lately about generative AI. So tell us a bit about what you're, what you're reading and thinking lately. Let's start at this kind of higher philosophical level before getting more down into the e-commerce needs.
Speaker 2:Sure happy to. As you said, one of my, my hobby horses actually is blogging a little bit on the side And I'd like to look at things from from a variety of different angles and primarily from the philosophical angle, that many of those new developments And it's always interesting to look at, for example, the very recent development we've seen with chat GPT, where on the surface level it looks like you know, computers suddenly have passed what's called the Turing test in computer science. Some context on that maybe. the Turing test is a very simple idea of how do we test if a computer is sort of intelligent on the human level. Well, the simplest idea you could have to test that is just have a human being, have a conversation with that computer And if the conversation goes on for really, really a long time and the human being can't distinguish if on the other side there's a computer or there's another human being, then the computer has sort of passed that, the Turing test.
Speaker 2:Now, if you and I know you had the experience and many of our listeners probably as well if you're going into a long conversation with the chat GPT for example, you might actually have the experience right. You might feel like there is a omniscient human being on the other side who can answer all of your questions and then help you with everything that might come to your mind. And that's a very interesting philosophical point there, right Like when do we distinguish between something being artificial intelligence or just a computer on the other side, something that begins to feel very human?
Speaker 1:Yeah, definitely, i'm reminded to have. I mean, this comes from the realm of science fiction, but this, like empathy test that exists in Blade Runner, and I think that there's a stunning degree to which this technology can also pass, not just on intelligence, but in appearing to express emotions and stuff like that, which is totally fascinating, although newer instances like Bard, for example, my understanding is that Google is putting a lot of safeguards into place to purposely constrain it so that it shouldn't even really use first person, It should not emote. They really actually want to constrain that and shut that down, at least within certain contexts.
Speaker 2:That's true, and we've seen that happening actually and we've seen that unfold actually in real time with Microsoft's new Bing, where the initial versions that they put out there for beta testing had no such safeguards in place And there were kind of immediate attempts to, you know, make it talk nonsense and make it look like a deranged human being. It was quite easy to prompt the chatbots to give you very weird answers, like to make it seem like it's a human on the other side that has some mental problem or something, or have it appear to be sentient when it actually is not, and Microsoft has put some safeguards on that. So they published on how they restricted certain terms in the output, how they are instructing the AI not to respond to certain queries and how they are just limiting the level of emotional expression. And in the latest version, you can also, as a user, configure in the beginning sort of how emotional or how serious you want your conversational partner to be. But that to me and, as you said, google is definitely trying to do the same thing with Bard, because the last thing that they want to do is have a similar PR disaster as we've seen with the new Bing, after this first couple of very weird utterances by the AI have been publicized. But what I want to say on that is one thing that I wrote about very recently is to sort of an unassuming consumer on the other side. You're very quickly tricked into believing that you are having a conversation with a real human being. So it looks like this thing actually wields the pool of language very, very well and it seems to be just someone you can trust on the other side.
Speaker 2:But if you look at the tech that's underneath all of that, the tech is mostly based on statistical interference and on predicting. Based on what the user said and based on initial constraints, and stuff like that is trying to predict what the next reasonable word in a certain conversation is likely to be. But that has very little to do with, for example, what is factually correct or what is helpful or what is useful. And to give you one more example on that, that's a very early chat GPT example that has been floated in the media quite prominently was an author who basically asked chat GPT a very simple question. He asked what's the latest book by this and that author, by himself, and the answer that chat GPT gave with full confidence was the last book that this author published was this and that title and it was published in this and that year.
Speaker 2:And if you read that statement you would say, yes, that sounds correct. The guy wrote this book in that year. But actually it was completely wrong, like it was the wrong title, it was the wrong publishing year and it was not the latest book but a book that you wrote like five years ago or something. But it was presented to the person at the other side of the prompt with very high confidence And if you're not fact checking that, you would immediately be tricked into believing that that is true. And that to me, when I put on my skeptical hat about all the technology that I think can send us into some very, very dark places, like if we can't tell whether or not we can trust the output of these things and these things seem to be very, very confident in what they're presenting to us we might run into Soviet problems.
Speaker 1:Yeah, we already live in this world of like truthiness and post fact environment and all those kind of phrases that come up And it does speak in this confident matter of fact way. Maybe you could just unpack before we move down into the sort of the e-commerce level. I'm packed one or two concepts, like I'm interested in this idea of philosophical zombies that you wrote about, like it's certainly a cool phrase. Maybe you could explain that idea to us. And around the steam of appearing intelligent versus actually being intelligent, because that's sort of yeah, we said that chat GPT, it's just trying to predict the next word in any given sequence And this has turned into sort of magical unlock for pretty cool capabilities. But why don't you?
Speaker 2:unpack this for us Definitely So? I mean, the term of the philosophical zombie was actually crafted, i think, back in the 70s or 80s, when philosophers really engaged with the question of consciousness, never trying to figure out what it is that makes us human beings conscious and how to distinguish that from other things. Like, the panda molding in my hand is definitely not conscious. My pet dog might be conscious to some degree, and you and the other side have a very high assumption that you are conscious, right, but how do we make these distinctions And how do we actually decide whether or not something is conscious? And obviously that has a lot of implications, particularly on moral philosophy. Right, i can, for example, do what I want with my pen because we decided it's not conscious, but there are rules and regulations about how we can treat other conscious beings. So it's important to make the distinction and to find out whether or not something possesses some level of consciousness. And the philosophical zombie is this idea of could there be? is it conceivable that there are entities that look and feel and behave and speak like human beings but actually are not conscious? So there would be no inner experience, right, you would have somebody whom you can have a prolonged conversation with, but there's no internal representation of the world that's going on on their end And philosophers have been arguing about that. So there's this conceivability argument to say well, is it even possible to appear to be human without having consciousness, or is consciousness just a necessary precondition to have an intelligent conversation? I think it's consciousness of the stepping stone that you need to have in order to be as intelligent as a human being. That's an interesting question. And then the second question is if it's conceivable, then what would such a being need to look like? For example, if we would replicate all the billions of neurons in the human brain, if we just replicate that in a computer, would that be a philosophical zombie? Would there be sort of an inner state of of, of in a stream of experience? would there be consciousness? would there be sentience, or would this just be? would it be morally okay, for example, is to turn off the computer? Or in that situation, would it not be morally okay to turn off the computer because you would actually kill something that's conscious?
Speaker 2:And to bring this back to chat GPT, what do many people argue is that, with sort of passing, what we could call the Turing test the chat GPT has achieved something like being a philosophical zombie. You could argue that this appearing to be as intelligent as a human being, but not having any conscious experience on the inside that that actually makes chat GPT this philosophical zombie. So if you would think of it in that terms, then this first question, like, is it conceivable that there is something like a philosophical zombie? that box could actually be checked, because, yes, we have chat GPT now and that is the philosophical zombie. And the question, the larger question than is what does this mean in terms of how we interact with it, what does it mean in terms of morality And what does it mean in terms of, yeah, how do we train these things to interact with us as humans?
Speaker 2:If you remember, a couple of years ago, google actually unveiled their conversational chatbots that would, for example, make a head resting appointment for you or that would reserve a table for you at a restaurant, and the way they would do it is that thing would actually call the restaurant or would call your head rest and would say hey, i'm Ralph Meijer's personal assistant and I would like to book a table for two people at 8pm tonight. And then there was this huge discussion whether this thing would have to announce to the other person that it is a robot, because voice synthesis was so good at that point already that if you listen to that, you couldn't distinguish whether or not that's a person or a robot, and that's something that the people are thinking about a lot these days. From the regulatory standpoint, does an AI have to announce itself that it is an AI if it could otherwise be confused with a human being? But you know, as I'm rambling a little bit here, but you asked initially what's on my mind these days, and that's a lot of what's going on.
Speaker 1:Ask a question, get an answer. No, that's why I brought to you Anuril. You always have a lot of information entry command and a lot of deep thoughts around that, which, by the way, is another topic you mentioned. I'll refer people to that article. Go check it out, because you mentioned this kind of threat to deep thinking and to attention that this kind of technology could pose, which I don't disagree with. I think there's already a lot of threats to deep thinking out there And, um, i certainly I feel like I face more challenges with that than I did as a younger man, and some of it's maybe me, but I like to think some of it is environmental, so I can be absolved of any guilt there.
Speaker 1:But let's bring this round, then, to toward e-commerce, toward digital advertising. Let's just apply some gravity here and make this you know, bring this down onto the philosophical clouds and into the dirt and grime of everyone doing business out here. So you recently broke a bike chain or no, i don't know. You have to tell me what this bike part is, because I don't even know how to say it. It looks like a French word And you went to Google looking for answers. Yeah, and you interacted with two different kinds of interacted with Google sort of index of the world. It's it's classic, classical search engine, and you interacted with some other tools as well, so can you tell us about that?
Speaker 2:Yeah, definitely. So. The thing is, i very recently picked up cycling again and I got a nice kind of wind. That road bike that I mean, she is a beauty, but she also has her her own personality and apparently, you know, from time to time things break and things need to be fixed.
Speaker 2:Because it's, yeah, not a very new model, and me being an obvious in all of that, i recently turned to Google to just figure out, you know, what's what's broken today and what needs to be repaired. And how do I do that? And actually that I used that last instance of such a problem to compare a couple different search engines and try to get a realistic view of where we actually are when it comes to building intelligent machines. I don't know, there's a big term, but ultimately that's the goal of many of these things, right? And so what I tried to do I was googling for sort of the making, the manufacturer of the bike and the term change your radar, because that's apparently the thing that I thought I had broken. And if you take a very sober look at the results that Google would spit out for that and list them on my blog, obviously you see that very few of those web pages are listed on the first result page are actually relevant to me as the consumer in that situation. But it's factual information that might, to a large degree, be accurate, but it's not helpful. It's just yeah, it is nothing that I can actually work with. It doesn't teach me about what a change your radar is. It doesn't tell me how I figure out how to repair it. It doesn't tell me where I can purchase a new one or which one I need for this particular bike. It's just a very generic list of results. And that points actually to an underlying limitation that we still have with how we're representing information and making it accessible by a search, which is this 1980s or, at the latest, 1990s concept of we're just indexing everything. So we're parsing all the text that's out there on all the web pages And in recent times we also started parsing all the videos, parsing all the transcripts, trying to understand all the images that are out there, and then we're building a huge database where we have all those keywords in there And when a user searches for something like a change your radar, then we just bring up exactly those pages that mention the term change your radar.
Speaker 2:Most often, that's pretty much like Google works and pretty much like all the other search engines work today, and that, obviously, is not necessarily that helpful. What would be a lot more helpful, in my view, would be a more semantic understanding of the world, and what I mean by that and I know Google has been working in that direction with their knowledge graph and with their shopping graph on the other side is trying to build a semantic model that says OK, we know that, bicycle, for example, there are different categories of bicycle, there are road bikes and there are mountain bikes and there are e-bikes and there's bicycles consist of parts, and one of those parts is a change your radar, the other one is the chain itself and the other one is a titer. And then if a person searches for something like a track change your radar. You could say well, i know track is a manufacturer of bikes and track makes road bikes and this particular road bike that he mentioned has this particular change your radar. And then you could make the semantic connections and could point to much more helpful search results.
Speaker 2:To some degree, google is doing that. So, for example, if you're looking for the name of a famous person on Google, it will point you, it will bring up those info panels on the right side We see the person's name and age and everything That is structured semantic information. It's parsing that also from the index that it's building, but it has some logical understanding about the thing that's going on there. It knows that's a person. It knows a person, it has an age. It knows a person, it has a website. It knows a person has a biography and stuff like that. And that, to me, is one of the very much overlooked concepts in knowledge management, knowledge retrieval today that we would need to put a lot more emphasis on if we want to make things like chat, gpt and all those chat-based search models a lot better.
Speaker 2:Because ultimately, what happened with my change your radar dilemma was I tried to do the same thing on ucom, which is one of those new chat-based search engines, very similar to the new Bing, which would have worked just as well, but just wanted to give you a chance.
Speaker 2:And if you go to them and do the same thing again so you search for what change your radar do I need for this and that bike, you also get results that look reasonably accurate And you can ask it where can I buy one of those? And it will point you to a couple of websites that would sell that stuff. So that's good, and that's a bit more helpful actually than the static search results from Google that we've seen, however, and that's where sort of the problems begin. The start, that's where the thing starts falling apart, actually, is when you make a very tiny change to your initial prompt to the machine, like, instead of saying, what change your radar do I need for a track alpha bike, when instead you add in one little typo in there, for example, or you just use a slightly different term because, like me, you have no idea how to spell the Railer right Sounds a little bit French even.
Speaker 1:The way it's written, at least is if it's pronounced D-Railer, i've got no problem, but the way it's spelled is quite different.
Speaker 2:Probably both of us are butchering the terms. If there's any bike enthusiasts out there, please correct us. Yeah, or any French speakers? Any French speakers? thanks, definitely So.
Speaker 2:The problem is, if I make a tiny mistake in the input to mix metaphors here a little bit then immediately the wheels start coming off, the thing starts hallucinating, it starts pointing to inaccurate results, it starts even producing results that are not helpful and that are even wrong, and it's extremely hard for the user to say beforehand whether I made a mistake. That would lead the machine to doing something weird. I can look at two results, two prompts and two results, and I can't distinguish whether one of those is prone to be factually incorrect, for example. There's no way for me to tell that other than going to another search engine, another tool, and trying to cross check and cross validate these results. And that's a big problem, actually. And that brings us back to the semantic modeling that I mentioned before, because those new search engines they're based on deep learning and large language models. They have very little modeling, they don't have what Google possesses with the Knowledge Graph, they don't have that underneath And therefore they can make even less semantic connections between the things you're looking for and how they relate to the stuff that's out there on the web, and that, to me, is a very big problem, because we are putting a lot of trust into these machines.
Speaker 2:Like I said beforehand, they appear to be human, like if I'm chatting with that thing, i might get the impression that I'm actually talking to a bike specialist there, but the fact that I made a small typo and my bike specialist, on the other hand, suddenly recommends something totally different and points to nonsensical results, that actually is a problem. That's because I can't distinguish that and because I would put a lot of trust into the information that's presented to me. And to close this off to the close of the more pessimistic take on this, i think we need to be very careful about how much trust we put into these systems, particularly when we use them to make more consequential decisions, like how to repair our vintage bikes.
Speaker 1:Yeah, exactly, i mean to offer some counter perspective. I think that these are growing pains and we have to see how that's going to evolve, but I don't know to your point about the way Google search is functioning and behaving here. Yeah, they've definitely been working on semantic knowledge And I think there are updates around Google, bert and all kinds of things here. It's definitely more advanced than just keyword density or something of the past. But I mean, fact is I see I'm looking at a screenshot You would search 2010 track 2.3 alpha front derailer, and I don't know if they were missing another kind of intent where, like, that query is somehow quite specific, but also it doesn't necessarily say what you want to know or do with that. But it's just pretty garbage. What they brought to you Like there's some page from 2017 about a stolen bike is the second result, and it's kind of inexplicable the way that this SERP looks. So I am wondering about that And as a user, i feel like the search engine result page has somehow I don't know gotten a little bit worse or less usable over time. I'm less satisfied with Google as a product than I used to be, so I'm sort of glad that there are some new technology coming in to start things up, and I hope that it will be a net positive.
Speaker 1:But another thing here is like about the ads that you were served, because I feel like a tricky thing. Google has these knowledge graphs, as you've mentioned, and they even have a specialized knowledge graph, a product graph called the shopping graph, and this is primarily used to power and serve ads. As far as I know, i'm not sure to what extent they use it beyond that, but you did have a very specific query and the ads that you were served though it's very hard to know which of those ads is correct can actually deserves your click, if it's really compatible or if they just bid higher. It's super opaque And I think there's a problem there where the organic results were totally weird. The ads were almost more relevant, perhaps because of the shopping graph or the product graph, but it's still a challenging thing to manage as an end consumer.
Speaker 1:And then to your point with the conversational AI, that this isn't quite there either. It's just a kind of an awkward phase right now, i feel like, and or maybe our expectations are just constantly raising and raising and raising. But I am optimistic that there will be really good shopping experiences and browsing and consumption experiences possible in with conversational chat And I don't know. Some people would say that that's you know, i think it's. Some people would say that's a very capitalistic use case or something like that. I don't care, this is an e-commerce podcast. It's a use case that I'm excited about that. I'm optimistic about where I think that Google as it is right now isn't really fulfilling that experience. If you go to Amazon, it's not gonna be much better. Amazon's experience has degraded a lot lately too, like something's gotta give here, and I'm optimistic that conversational AI could be that something to help kind of improve, help manage the dilemma of choice or the paradox of choice and manage some of these things that are going on.
Speaker 2:Wow. so that's a lot to unpack there. Let me just add a few remarks On your point about sounding capitalistic. when we're talking about online advertising, there are some people out there who think that this whole model of online advertising is just broken. You know, particularly in the media world, people say, well, you know the fact that everything is sponsored by, by ads sort of it killed journalism and stuff like that. And the problem is, i don't on a very high level I would somehow agree with it. But on the other side, this is the world we're living in, right, and we as the consumers, we're getting an endless amount of free stuff out of that because we're living in this world that's fueled by advertising.
Speaker 2:The concern that I'm having is, if you're giving me that free stuff in exchange for presenting me ads, then I would appreciate if those ads were highly relevant to me, like in that conversational exchange that I had with youcom about my change e-railer. I explicitly asked where can I buy one of those? And if I'm doing that as the consumer, i have no problem at all with a sponsored result being thrown in there, right? If the chatbot then says look, you can buy one of those at these three bike shops. That's the one I recommend, you know, and that's the one that's a sponsored result and stuff like that. If the thing that it's presenting to me is highly relevant, i'll give it's the right product. if it's the exactly e-railer that will fit onto that exact bike, which, you know, making that mapping can be quite confusing for an amateur like me, because there's a thousand different variants of these things and all the manufacturing by one brand and it's really hard to figure out which one you need. If I get a recommendation for the right one, i'm definitely willing to click on an ad And if you look at the shopping ads that Google brought up for that particular query we mentioned earlier there, you see the big problem.
Speaker 2:It's like I'm getting 10 to 15 different ads. The SERP looks like a, you know, like a classified page from a newspaper. There's 10 different ads. None of them are relevant, or some look on the surface level to be relevant, because I'm getting 10 different change e-rails but none of those e-rails actually would fit onto the bike that I mentioned in the query.
Speaker 2:So again, as a user I feel like bombarded with irrelevant advertising because I'm not gonna buy any of those unless I'm pretty sure that you know that's the e-railer that will actually solve my problem. And I think if we're talking about advertising, you know, fueling a lot of what's going on in the world I think one concern that's often overlooked is how do we make those ads more relevant to people. When we're talking a lot about increasing efficiency, like you know what happens, how do we bring the right ads to the right people and stuff like that. But there's also a question of effectiveness, like if you can be at the right point in the customer journey and if you're in the right context and stuff like that, your ad is just gonna be a lot more relevant and that's gonna ultimately boost your click-through rate And that's gonna boost it a lot more than just throwing another 10-year relevant ads at me at some other point in my customer journey.
Speaker 1:Yeah, the sort of points is toward the kind of worms here. I totally agree that relevance is key here. And then the others this whole question about how regulators feel, which it can be quite different ultimately, because I guess we're probably headed into an area of more contextual advertising And I don't know, maybe the ask is too big. Like are we talking about an edge case here with this derailer, like you know, in terms of the compatibility, that's probably listed, you know, in a manual somewhere compatible with the following models or something like that. And I don't think we can necessarily expect the Google's product graph to have that level of information, particularly on an older model and everything like that. But AI there is kind of the prospect that AI, if it's read every little piece of thing out there on the internet, that it could have this kind of awareness. But yeah, let's see where that's gonna head. I mean M's, m's, m's, M's, m's, m's. To the point about ads in these experiences, i agree I'd like to see good, relevant ads in there. I think there's a great opportunity for Google to add a buy box or do different things here. Search and shopping campaigns are for sure going to be present in these kind of experiences and maybe new native formats as well. To me, a buy box seems like an obvious win for Google or something they could try, but there's many factors at play here. Another thing before I let you go, ralph, i just want to quickly pick your brain about.
Speaker 1:On the last episode I talked about updates from Google Marketing Live, like conversational campaign creation, which is basically sort of like using Microsoft Copilot or Google Duets. It's one of these. You know this idea of using generative ads, you know, to create a new platform. You can use generative AI integrated into a productivity suite in this case integrated into ads to help you create your campaign. And then Product Studio, which is sort of like you could almost think of it as like an AI assisted Photoshop, again in the UI which is going to help you enhance your product images. You can use like generative fill on the backgrounds, or the AI can automatically fix resolution, upscale resolution or stuff like that. But what do you think about these kind of technologies?
Speaker 2:Yeah, we've seen it at GML 2023. We've seen a couple of very interesting developments there that are ranging from, in my humble opinion, from, you know, being toys for people that are sort of dicing on the cake, to some things that will bring us closer to fundamental changes in how the whole Google Ads universe is going to operate. To start with the ones you mentioned before, like these chat-based assistance when it comes to creating campaigns, I feel like that's more on the bells and whistles part than actually, you know, something that's going to change fundamentally how the world works. I don't doubt that that's going to be helpful for some people. It's going to probably make people's lives easier. It's going to speed up the workflow from zero to having your first campaign running.
Speaker 2:But on that topic, I'm a bit skeptical how big the impact of that is ultimately going to be. The way I'm thinking about it is consider the use cases you have on your iPhone with Siri, for example And yes, I can ask Siri to put in an appointment for tomorrow at 3 PM, and it's going to do that right, And I can do it in a conversational way. But ultimately, as a user who has done this once or twice, I'm just faster in picking up my phone, clicking new appointment, clicking Save and that's that. So I think there's some use cases where people are going to make use of those chat-based interaction models, But there are others where people are just going to be faster in the existing Google Ads UI or in the Google Ads Editor or in stuff like that. Then, when it comes to doing these things at scale for catalogs of tens of thousands, hundreds of thousands, millions of products, I'm still a bit skeptical whether that's going to help people who are managing those large types of accounts.
Speaker 2:The second point you mentioned about product studio and having AI help you ultimately improve the quality of your ads. I guess that this Photoshop-like tools like get rid of the background and stuff like that. That's definitely interesting, because I think it's biting into the market that some external companies external to Google Head in the past that have tried to provide these things on top. There were things that integrate into your workflow and would make sure that all your product images have, for example, the same white background, And having these things being provided by the platform itself that's certainly relevant and helpful. I think it's going to put some pressure on the market of tools that are helping you manage your Google Ads account a bit better.
Speaker 1:Yeah, i agree. I mean, none of these capabilities are necessarily new per se or unique per se, but having them situated right there in these interfaces is a massive home field advantage for Google, which will definitely put pressure on these third parties. Yeah, about the conversational campaign creation, i don't know. I talked about this on the last episode, but I think the time savings are quite clear. But in the end it just remains to be seen Is using this kind of an approach actually does it perform well, better than a human approach? Where's the trade off there? Is it the difference between?
Speaker 1:for some businesses it's like I couldn't have done this at all without this tool, obviously massive for them And then there's other larger companies where they have to do a more complicated calculus on that of what is really the increment of this. What are the trade offs? How does it affect my performance? Because if it's saving time but hurting performance, obviously this could be a bad outcome. So this kind of stuff needs to be great. I can go faster toward destruction, awesome. And I'm not saying that's going to be the case, i'm just saying that this is totally unproven and untested so far. So I think that's our big job ahead of us is to test.
Speaker 2:Yeah, totally agree with that, And I think maybe I'm over simplifying here, but I do see this as a long tail problem. One thing that Google definitely wants to tackle with this is how do we make the onboarding experience easier to get all of those merchants who are not Google Ads if there's still some of them out there How do we get those folks onto Google Ads really quickly And how do we make sure that they're using it at scale and spending as much as possible? That's the long tail of the market, and then there's the short head of the market of really big spenders, really big advertisers And, as you said, it remains to be seen how much value they can grow out of these tools. There's definitely something in there, but there's also a role around all these discussions about what's the PPC manager's role going to be like in one year, five years, 10 years time, what's the role of the agency going to be like, and stuff like that. There's a lot going on in the ecosystem, and what Google unveiled at GML this year definitely adds to that.
Speaker 2:There's one final topic I want to touch on while being at this topic of GML. I'm not sure how big of a bust it actually made in the news, compared to the generative AI experiences and stuff like that, which is Merchant Center Next, where I think this points a little bit more into a strategically more relevant direction. There's a few of the things that there and we're trying to read the tea leaves here, obviously but a few of the things that they're trying to do with this is reducing the emphasis of structured product feeds in favor of Google themselves parsing your web shop, parsing your website, grabbing the data that they think is relevant and then using generative AI to just create all the different ad formats out of whatever they define on your web page. Right, and that's going to put us into a more interesting. It's going to put an interesting pressure on the ecosystem and raise an ecosystem of companies that help you with feed management, feed optimization essentially getting data from your core business systems into such a shape and form that the different ad platforms can work with and transforming it and doing all these steps that are needed to create feed for Google ads and for Facebook and for Amazon and for all these things. It's going to be interesting how that segment of the market is going to be affected by what Google is trying to do with Merchants in the next, and also it remains to be seen how much of those auto-generated assets that we also see with performance max more and more, how that plays into the picture And, as you rightly alluded to before, how that impacts performance at the end of the day. So we have seen I mean, we've been monitoring that market very closely and we've seen initially some very disappointing results, particularly on auto-generated video ads. But this can and this is something where I'm really a tech optimist I think that this can improve really quickly and we're going to see a lot more auto-generated assets that are of high quality and of high relevance and can be produced at a scale that is hard to replicate if you're trying to do that manually or via an agency And I think Merchants in the next is going to play a bigger role there than we assume at the moment And this question of trying to eliminate the feed or bypass the structure product feed, that's going to be interesting there as well, and it sort of brings us back to the initial question or the initial discussion that we had about how much structured information or how much semantic modeling does a company like Google actually do and how much does it understand how these things connect with each other.
Speaker 2:It's going to be very interesting because on the one side as with my bike troubles we've seen that Google has, despite all these claims about the knowledge graph and the shopping graph and stuff like that, it's still relatively dumb when it comes to figuring out that a track 2.3 alpha is a road bike right.
Speaker 2:It doesn't know that, despite all the efforts that have been going into the knowledge graph and to bypass that, obviously it introduced things like the shopping graph and introduced things like the shopping feed, where the burden is on you as the retailer to provide the structured information right. Provide Google with the system manufacturer. This is the brand, this is the price, this is the retail price, this is the shipping costs and all these things. You have to provide that in a structured way, because Google just does not have any way to come up with that information if you don't provide it that way. Now, with putting less emphasis on the feed in the future and Google trying to retrieve that information on their own, basically grabbing it from your website, it's going to be very interesting to see how accurate it is and then how relevant the resulting ads are going to be. So that's my more strategic take on what we've seen at the GML.
Speaker 1:Yeah, good points you raised. I think with Merchant Center Next and the possibility to automatically populate this data It does usher in. It brings us closer to an age of, like total M2N automation where your feed you don't need to supply feed, you basically just need a URL and you could have your feed generated for you. You could have all of your campaign entities generated for you, your assets by and large generated for you or enhanced for you, and so on. And I think that this is to your point about late adopters versus, like these high-end enterprise businesses. This definitely feels mid-market.
Speaker 1:It remains to be seen if the people who haven't adopted yet are going to magically adopt now.
Speaker 1:I'm not sure about that, but generally speaking, i do think it speaks toward these under resourced small and medium businesses. I think an irony of performance max is that it felt like it should be best for SMBs but it kind of wasn't, and Google has struggled to find a way to really like they, simplify their product more and more, seemingly in the hopes of bringing on these smaller businesses and these late adopters. And yet, at the end of the day, it increasingly becomes a budget game because it just ushers in more and more of this age of average, where it's harder to differentiate, where you end up having a very average campaign average feed, i would imagine. If everything is automatically generated, and then what is there to do more at that point but outspend in order to kind of call attention to that? But I don't know, we're getting into speculative territory here and also I think we're at the end of our time box. So, ralph, i just wanna thank you again for joining us on the show Always great to have you back.
Speaker 2:Thanks for having me.
Speaker 1:Thanks for listening to Growing Ecommerce, and if you enjoyed this podcast, please consider sharing it with coworkers, friends or within your professional network. We really appreciate it. This podcast is produced by Smarter Ecommerce, also known as Mech. To learn more, visit Smarter-Ecommercecom ногapscom.