#443 - Unlocking Amazon’s Algorithm: A9 Secrets, SEO Hacks, and the AI That’s Changing Everything

#443 - Unlocking Amazon’s Algorithm: A9 Secrets, SEO Hacks, and the AI That’s Changing Everything Kevin King: Yo, yo, welcome to episode 443 of the AM-PM Podcast. This week my guest is Oana, that's right, coming to you from Italy by way of Romania. She loves to geek out on all the Amazon science papers and we're gonna be talking about a little bit of PPC and diving deep into some of the new AI stuff that's going on. What's the difference between all these different, there's like seven or eight different programs that Amazon's got going when it comes to AI. We're going to be talking about all those and how you can optimize and some of the things you can do for that. In the meantime, enjoy this episode with Oana. Unknown Speaker: Welcome to the AM-PM Podcast. Welcome to the AM-PM Podcast. Where we explore opportunities in e-commerce. We dream big and we discover what's working right now. Plus, this is the podcast where money never sleeps. Working around the clock in the AM and the PM. Are you ready for today's episode? I said, are you ready? Speaker 2: Let's do this. Unknown Speaker: Here's your host, Kevin King. Kevin King: Another week, another awesome AM PM podcast and today's guest is really cool. I think you guys are going to and gals are going to really like her. Her name is Wanda. Wanda, how are you doing? Speaker 2: Hey, Kevin, super excited to be here. Thank you so much. Kevin King: Yeah, you're from Romania, but you're in Italy right now, right? Speaker 2: I'm actually from Transylvania. Kevin King: Oh, Transylvania, okay. Speaker 2: Yeah, it's the area of Romania that's very famous about Dracula. Kevin King: It is, it is. Speaker 2: There are no vampires over there, just so everybody's wondering. But yeah, I'm from Transylvania, and I have been living in Italy. I'm in Italy at the moment. Kevin King: There is actually a castle, though, in Transylvania, right? Because I know that, I've been to Romania a couple times. I spoke to a friend over there in Bucharest. Speaker 2: Bucharest, yeah, that's the capital. Kevin King: Yeah, in Bucharest. Speaker 2: It's not in Transylvania, though. Kevin King: No, I know. Yeah, no, I know. It's like three hours away, but they do tours there to some sort of Dracula's castle, so that's just like a little Disneyland kind of thing. It's not the real thing. Romania is cool. There's a lot of people. In the Amazon industry in Romania, so there's you know, I try and Turku is actually over there and Alina. Speaker 2: Alina, yes, my friend. Kevin King: Quite a few people. There used to be like Amazon, what's it, Hooty? Speaker 2: Hooty? Kevin King: Hooty? Speaker 2: Yeah. Kevin King: It was on course. It was kind of like an amazing selling machine type of thing that was teaching all the Romanians. There's a big, big subset of Amazon sellers in Romania, even though Amazon doesn't really They don't ship to, I mean, they'll ship to Romania from other marketplaces, I guess, but there's not like an official Amazon in Romania. Speaker 2: No, there's no like Amazon.ro. There's no Amazon Romania, but you can buy it from Germany, Italy, France, whatever, and they would ship it over there. But it's true, the community that we have there is quite impressive. I know a lot of people that, especially in the PPC world, for some reason, when you look for PPC specialists, a lot, they are Romanian. That's how I actually started. You know, my background comes in advertising. My passion came from advertising. I learned it from Brian Johnson. I actually had access to his 2016 Splinter Product Academy. And that's what got me here today, basically. And I know a lot of people who did the same thing and learned from different online courses, Brian or whatever. And yeah, there's a very big community in Romania. Kevin King: How did you end up in Italy then? Are you traveling or you met somebody or how did you end up in Italy? Speaker 2: My husband. Kevin King: He's Italian? Speaker 2: Mm-hmm, yes. In fact, I'm also Italian now. I have a double citizenship. So it's very easy to travel to the US because I don't need a visa. Kevin King: Awesome, awesome. That makes it easy. You can come over to the show. That's awesome. Were you in advertising before this whole e-commerce game when you got out of school? You said your passion is advertising, so were you doing something different before e-commerce? Speaker 2: I will be honest with you. I tested AdWords for Google at the time. I kind of hated it. It was like, I thank God I got into Amazon because that was kind of like You know, an instant flame over there and I liked it. But I did not like the social media part of it. I was not into social media at all. For some reason there was something about Amazon and how it worked, the advertising on Amazon. Sponsored products and then everything that is involved to what we have today. Kevin King: Were you a seller first and then you really liked the PPC side or did you go straight into the PPC stuff? Speaker 2: Straight into the PPC side of things. PPC and SEO because for me from the beginning was I really wanted to understand how indexation works. I realized how SEO was a big, big component of PPC right from the beginning and kind of like I tried to balance the two for as long as I could up to the point where I kind of leveraged that into becoming a strategist. You know what I mean? Like taking that to the next level when learning and evolving about how Amazon works. But mostly it was advertising and SEO and then diving into psychology for how shoppers buy, how they behave on the search, how you can increase the sales, how you can fix specific products and then kind of like moving into the visual side of things. And yeah, and then getting into science papers, kind of like that was my... Kevin King: So did you know somebody that was selling and they said, hey, I don't understand this advertising. You're like, oh, let me look into it. Or did you just hear something online about people are selling on Amazon? You're like, nah, I don't want to sell. I want to do the math part. I want to do all the PPC stuff. Unknown Speaker: How did that happen? Kevin King: I'm trying to understand how that actually happened. Because usually people start selling and either they don't have enough money or they're not successful or they're successful and they run out of money or they just like, I'd rather I deal with people on PPC that have to deal with inventory and floating all this money. So how did that evolve, actually, when you went straight to the PPC? Because that's unusual. Unless you got hired as a job to work for an agency or something. Speaker 2: I had a friend who I need to thank for. She was already in advertising. She was doing Amazon PPC and she was like, I think you would like this. Can you check it out? You know, because she knew me. That instantly resonated with me. It was mostly about the numbers, following the numbers, trying to understand, create. It was mostly about the creation side of things when you create campaigns and then you monitor and working with the numbers. It's like really crunching them and trying to understand what impact it has. That was the overall power that hit me. For me, Brian Johnson was an inspiration. For so long, I followed only him in order to understand how he did things, how he explained it. Because if you think about it, I can call him crazy stuff. If you think and look back, even in the early days, he came with a lot of strategies to test that were tested and how you can push everything to the limit and try to maximize every strategy, every bit, every keyword, every placement. It was insane. Even when you did not have all the knowledge yet. I would say the level of how sophisticated advertising is at this point, right? Because if you think about it, Kevin, anyone can Increase or decrease bids or create a campaign and just let it be, right? But over time, that kind of like you really had to understand what's working, what's not. Dial into the audiences, how you group them. Why is one campaign or one specific ad group working better than the other one? And kind of like trying to understand the reasons behind it and investigate In order to get to the technical side of things that most likely, you know, people that just scratch the surface would not dive into. Kevin King: But PPC now is so much more complicated than when you first started. When you first started in 2016, it was pretty basic. I mean, extremely. People could pretty easily do it themselves. Now, you need a degree in rocket ship science or something to actually do some of what it can do. It's still behind Google. Google is still much more advanced than even Amazon is. Learning that from Brian, did you learn from anybody else or was it mostly trial and error or working for clients and kind of figuring things out as you went to complement what you were learning from Brian? Speaker 2: So, throughout the years, I will be honest with you, is following Brian Johnson than kind of like trying to pick other ideas that I would test. It was always about testing. It's not because one shoe does not feed them all. And luckily, I easily got into working with different agencies and handle like Hundreds and hundreds of clients of different sizes and that's where kind of like the expertise comes in when you test the strategy and you kind of like understand on groups of clients on what it can work and what budget you need sometimes for a specific strategy or I don't know how you are able to structure, for example, How do you plan a campaign structure? How do you plan that for clients? And yeah, just working with all different sizes. I was lucky to work with brands that most likely majority of you have maybe all of them in your house. Basically, and yeah, it was just a lot of testing and proving what works and what doesn't and working with a large amount of data. And you mentioned in the beginning it was easy, right? And it was easy and when you test and work on something that's easy and then it kind of like evolves over time, you are already in the game and you are able to kind of like test and try to understand what works and what doesn't. So when you become sophisticated, not that I don't encourage people who want to start learning PPC now, I'm not saying don't do that, it's just be prepared to kind of like I understand in a harder way what it can be advertising today. You know what I mean? Because if you start from something simple, you can build it up. We moved from Seller Central to the Advertising Assault. Yay! We have new reports. Yay! We have sponsored brands. Look, you can add a video to it. And when you experience that in a step-by-step, I think it's easier Rather than all together. So that was definitely a plus that I'm thankful for. That I was able to kind of like digest and learn and test at every step. Kevin King: Yeah, you're building on top of your previous knowledge rather than trying to gain all the knowledge of everything at once. It's kind of like drinking from a fountain that's just got a little bit of water coming out versus drinking from one that's really splashing, it's coming out full force. Yeah. So I know if I'm doing my PPC for my accounts, I'm in there. I feel the pain. It's my money. If I mess up on a bid or I mess up on a campaign, it's money directly out of my personal pocket. I feel that pain. I don't want to make that mistake again. When you're doing it for other people, how do you get into that fiduciary duty mindset to where you have to feel the pain? A lot of these PPC agencies, if they mess up, yeah, they might lose the client. I guess they technically feel the pain there, but a lot of times it's the workers. They don't feel the pain. Unknown Speaker: They're like, oops, I screwed up. Kevin King: So sorry about that. I just learned something and I won't do that again. But how do you get into the mindset to where you have this massive responsibility, especially when you're at big scale and you're dealing with a lot of money that How do you handle that? Speaker 2: I will be honest with you. Testing when you have a huge budget is actually easier than when you have a more struggling client maybe that has a limited budget. You know, the balance needs to come in. I will be honest with you. It really comes down to treating everything as it's always your own. Accountability is a very, very big part in Building long-term relationships and that's what gives you kind of like credit and success and I really do encourage everybody to, you know, when you work and do just maybe perform a specific task, think twice and act as in every moment as that's your own brand and that's your own product and that's the money that you're investing because otherwise Unfortunately, you don't grow. You don't grow from a personality standpoint like in the report that you have with the client and you don't grow professionally because it can be harder. From my point of view, accountability is a very, very important thing. When managing big teams, that was something that I was always looking for in the people that I was managing and the people that I work with even today on the accounts that I manage on my account. Accountability is very important. Of course, there will be, you know, things that go south and there are, you don't always have success, that's absolutely normal. But when it's planned and when it's, you know, you have somebody on guard and you know you can rely on them, that really makes a difference. Because if I cannot watch the accounts today, I know that, you know, the person who I put in charge will do that. And it really minimizes, I would say negative effects on what can happen. Plus, creating a structure. So if you think about it, when can things go wrong? When you don't monitor them or when you don't plan them accordingly, right? So if you plan and then make sure that you have a process in place at no matter what level to make sure that there's nothing falling in between the cracks, that's when you have full control and kind of like minimize. Of course, if something happens with Amazon or whatever, I would say planning and monitoring will definitely make sure that you don't make errors like that. Kevin King: So when you said you were managing some teams of PPC people at some points throughout, so when you hire someone for that team, is it better that they have PPC experience or is it better that they don't and you teach them your way and your processes versus them coming in, maybe they learned in a different agency or they learned on themselves and maybe it's not the best way? Which is better, to have someone with experience or to have someone that you can teach that just has the capability, the mindset of numbers and stuff? Speaker 2: It depends at what point you need that person and how fast you need them to come in and act because if I would need to train somebody from scratch, of course, I would expect to take longer than having somebody who already knows the basis and maybe I just get them to learn my way, as you call it, right, which is faster. So it really depends. I don't really have a preferred way. If you ask me like right now, I would say if I need to add somebody to my team, most likely I want them to know Amazon, to have some sort of experience in Amazon and then we can kind of like, you know, refine everything that they know and build it up. But starting from scratch, from zero, it might require, you know, a higher effort. And yeah, it always comes down to what you would expect from the specific role and when do you want that person to actually start working. Because if I hire them with no experience, they will not be able to touch the accounts. They will not touch anything. They will just be learning and shadowing. So, yeah. Kevin King: Forgetting about AI stuff, just the normal stuff, what's got you most excited right now and what Amazon's doing on the PPC side? Is it the video content? Is it the customer journey stuff? What's got you the most excited that you feel like, oh wow, now we can actually really do some cool stuff? Speaker 2: I think it's the focus towards audiences that I see Amazon shifting and I noticed that in the past couple of years. If you think about it, even when they started introducing the audiences in Sponsored Display, which I will be honest, they don't really work that much. I really hope that at one point we will be able to leverage them more. But the focus that it goes towards a profile specifically of the shopper and try to personalize that as much as possible rather than just kind of like throwing the seeds out there and you know waiting for shoppers to go in and search for it. I remember I was at Accelerate last year and I was talking to a search expert. One of the things that I was mentioning was the demographic side of things. You have demographics, but you think about shared accounts, for example. How much can I rely on Amazon information When it's not filtered and the good news was that they were actually working on that in terms of refining and trying to like reducing the noise and trying to understand if, for example, if I'm purchasing from my account for somebody else, if that purchase would not fit within my profile, they would be able to analyze and kind of like exclude that from the demographics. Which I thought it was amazing from my point of view. Kevin King: I think maybe one of the reasons you're excited there is because the audiences kind of plays into the AI side of things too, where it's now going from more keyword-based stuff to more intent-based stuff. And so audiences is a better way to actually target for intent-based versus keyword versus more just throwing darts. Speaker 2: Yes, that plus you can think about like this. If you really get more insights on your customers and you really create that Customer persona and you have all the insights. How can you maximize and improve your conversion rate by making sure that all the messaging that you have and even the, I would say, you know, the imagery side of things, not only the text side, how does it engage with my shoppers? So that plus the AI plus Cosmo, trying to understand more the customers. Now we have Rufus. I think it all kind of like blends in together. Kevin King: Yeah, that was interesting last month at the AI workshop that we did where Chris was showing how to actually create audiences with AI. He used a thousand AI avatars that interacted with each other, legitimately interact with each other on a Reddit channel. And then real people started coming in. And as the real people started coming in, he phases the avatars out that are actually interacting with each other. And within a month or two, you have 30,000 real people in there that are all around an audience. It's a really clever way of using AI to build audiences that you can then market to and pick their brain and sell products to and launch to. It's really cool. In the meantime, if you missed the Billion Dollar Seller Summit in Iceland, we just wrapped that up last week. There are replays available for that. So if you want to catch the replays of the Billion Dollar Seller Summit as well as Elevate 360, amazing, amazing stuff. Elevate 360 right after BDSS. Both of those are available at BillionDollarSellerSummit.com. It's BillionDollarSellerSummit.com. So go check those out and grab those replays and make sure you stay cutting edge and top notch on your selling with what's going on right now with all the things that are changing. We're kind of behind the scenes for a long time. I mean you weren't like well-known out there. I mean people in the circles kind of knew you but you just like this little secret that you're working for some different agencies and you were kind of doing the stuff for them and then all of a sudden You your name Danny McMillan comes to you and and says hey You're smart. I'm teaming up with you me you is Andrew right Andrew Andrew. Oh, yeah And let's let's dive deep. Let's geek out on the Amazon scientific papers and y'all put out this It's the second time that Danny's done it. He did another one a couple years ago on the AI algorithm. And then he did another one where you guys took a look at the scientific papers when it comes to AI-based stuff on Amazon. And then now all of a sudden, now you're out there. So what changed? What made you go from behind the scenes, puppet master, to now being in front of the camera and actually starting to speak in advance and coming on podcasts like this? Speaker 2: That's still something that I want to be honest with you. If you think about it, last year, nobody knew me. I went to this event in Prague. I started meeting people, but nobody knew my name. I was, at the time, managing 70 people in an agency. I think what happened was meeting Danny and just being able to geek out on the A9 stuff. We put white paper together. I think it was on Cosmo and that kind of like changed everything because at that point I started being on podcasts and I realized how much I loved just sharing what I know with, you know, everybody else outside of my team, for example. So that came on. Then it was the word that took a bit of my soul away, the honeymoon period when we worked off the two patents. So that was at, I think it was October last year, but we worked on it like I think it was three months of hard, intense work because I'm gonna be honest with you, Kevin, I don't have, I'm not a developer. I don't have a technical background, right? So as much as I really like to understand, it's all just curiosity and trying to understand how things really work. So when I was given the two patents and I started looking at them, I was like, that's gonna be intense because looking at a patent and trying to understand how the algorithm works, never seen an algorithm in your life is like, Okay, that's gonna change. So that's how kind of like things just happened. I will be honest with you. I'm super happy that I'm able to share this at scale and I have people also reaching out with questions sometimes, right? It's like, oh, I read that in your article. What do you think about this, for example? And that's something I truly, truly enjoy because at the end of the day, it's, you know, It's reading a science paper and if that can have an impact on somebody's account based on what I found, That just made my day instantly. Kevin King: There's a lot of people out there that are speaking. There's people, I call them circuit speakers. There's a lot of people in the Amazon space and you see the same people at every event or at every online summit or whatever. It's a lot of the same people. Some of those people are still selling and a lot of those people are really good people, but some of them aren't selling and they're not doing this. They're talking about things and teaching things that they're actually not doing. They might have heard it from someone or got it from someone and put some stuff together and they're good at collating it into a presentation. Maybe they're good public speakers because they have a lot of experience, but they're actually not doing it, but you're actually doing it and actually like diving deep. You say it the British way, patent. We say it in the US, patent. You can tell the Danny influence there. Speaker 2: Absolutely. Kevin King: They're not doing it, but you're actually in the weeds. Like you just said, you're not whipping up a presentation or something for a podcast or something in a couple days or the night before you got to present it. You're spending three months reading and trying to understand what's going on. I'm assuming probably doing a little bit of testing and stuff like that. There's no pay for that, really. I mean, you're just doing it out of passion and out of just curiosity. It's not like someone, Danny didn't come and say, hey, I'm going to pay you $20,000. Go spend three months researching this. But you get to use it and then for your clients, what you know, and then now you're sharing it out there with other people, which you found out you're putting in layman's terms rather than there's all this Like you said, you're not a technical, you're not a programmer, but you're figuring out what does this say and conveying it in layman's terms to a lot of people and that's really cool. Speaker 2: Yeah, yeah. And again, it's not about the pay. It actually came from a passion. So when I discovered Amazon.Science, at the beginning of 2024, last year, I was like, you know, when you take a kid to a toy store, that's how I felt, literally, because for me, it was always trying to understand more. So you do the test things, right? And you have like your test campaigns and you start understanding and pulling data. I'm very data-oriented at all times, like in anything with A-B testing and everything. And, you know, all of a sudden you have access to why sometimes things might work and why sometimes things might fail, right? And that was really, really important for me to understand the system behind the other side of the coin. Because if you think about it like this, you have Amazon, right? But when you think about selling on Amazon, there's the algorithm side of things, right? And then is the shopper side of things. So when you take care of this side, and you really make sure that, you know, you have You have a good product because it all starts there. You have a good product. Your pricing is aligned with your target audience. You have your messaging in place, imagery that is perfect, aligned with your strategy. On the other side of the coin, you will need to understand indexation, for example. How do I make sure that I'm indexed? Then how do I leverage that? What do I need to do? What keywords do I need to go after? What do I need to start tracking and making sure that I rank on? You know, all that technical side of things, which is how the system works versus how shoppers react to what they see. It's always kind of like a balance between. When I realized that, for me, it was amazing because I really started digging deep into understanding, for example, how Amazon understands the images with the AVEN, for example. Then it came Cosmo. With Cosmo, it was quite popular because everybody was talking about it. When the Cosmo paper came out, it was quite easy to even share in the community. So it all comes down to that. With the patent, it was more difficult on the honeymoon period side of things. But then when it came to Rufus at the beginning of this year, it was definitely easier. And I will be honest with you. So the first time that I went through the patent and I realized how Rufus worked, what came to my mind was Okay, I need to start optimizing and testing what it's in here, right? So before even writing the paper itself, I tried to understand it and then I ran tests on how to optimize for Rufus on our own accounts, trying to see how it interacts, how it pulls information. And even when you invited me on the webinar in Feb, that we had the live webinar, I was so happy to be able to show you like how I was pulling information from Rufus because all of that was done on our accounts. It was just to see how Rufus reacts every time and just making sure that you provide accurate information because at the end of the day, if somebody comes in and finds, for example, they take my piece of advice and maybe improve it, I would be happy. It's like, okay, take this to the next level and just share amongst us because at the end of the day, it's just a win-win situation for all of us in the community. Kevin King: What about the people that think you're wrong? When you guys put out the honeymoon paper, There's basically no honeymoon period. Speaker 2: There's a cold start. Kevin King: You explained it based on all the patents and all the data and the testing and everything, but there's still a lot of people and some well-known people that are respected that say that's bullshit. There is a honeymoon period. I have proof. I know it is. It's this, it's this. I don't buy what Dan and Oana put out. It doesn't work that way. What would you say to those people because they're out there? What would you say to them? Speaker 2: I don't mind being wrong. At least I tried. If I'm wrong, that would just open the conversation to understanding why. Of course, before you put out a paper like that, you really want to make sure that you're not wrong. Working three months on it, you kind of make sure that you don't do that. Between me and you, that was for me the first big thing that he was coming out with. For that, it wasn't. He wrote many brilliant things before that. But for me, I was kind of like at the beginning of this This is my path in my career. So for me, it was really important that that was not wrong. So I'm not a perfectionist. I just really like to make things really, really, really, really well. But if I were to be wrong, and that's what like from the start was, if you guys have any type of feedback, and I remember I hopped on calls with a lot of people from the industry discussing it. What's your opinion? At what point? Based on what? You know what I mean? I don't believe it because I don't want to believe it. I think that's maybe immature to act like that. So I am ready to be proven wrong as long as we learn. We get something positive out of it. Kevin King: I think some of the people that say that are people that it could be other factors that are playing in. They're just attributing it to what they call a honeymoon. Anthony Lee is the person who coined that term. It's not an Amazon term. It's not. He coined that when he was at Zonguru or Zonblast, I mean. Zonblast, like in 2015. It just kind of stuck and that's what everybody has called it. You guys kind of disproved it and it has always been this thing where you know the Chinese sellers used to say no it's two weeks or it's seven days or someone else say no it's 30 days or other person say no it's actually 45 because I have the evidence here and there's all this guessing and then someone says something and everybody else just runs with it. They don't validate it. They don't check it. They don't read the papers about what Amazon shares at Amazon science and of how their thought process and how this stuff works. And you guys dove into it and you're like, no, this is the way it works. And I believe your side versus the other side. So that's why it ran in the newsletter and everything else. And it caused a little controversy out there. So the same thing's happening kind of now with, I see it a lot with sellers with AI. And some of these software companies, Helium 10, Jungle Scout, I know the owners of all of them I said, hey, what are you guys doing about AI? And they're like, well, what's there to do? That's in simple terms. They don't say it quite like that. They're paying attention. Of course. I'm like, well, you're going to be out of business unless you pivot or add some additional modules to your stuff because your keyword stuff, which is a core of your business, you've got other things, accounting software and shipping software and other things. It's a core of your business and keywords are not going to matter nearly as much anymore in the way people are reverse engineering. Where's the opportunity based on keywords and where's the opportunity of my competitors don't have these keywords and I can put them in? That still matters right now, but it's on its way out where that matters. A lot of them don't believe that. They're like, no, Amazon's not going to change. That's not going to change. And so the same argument right now is with the people that Different opinions on honeymoon period, there's different opinions on how AI is going to affect search and I think AI is going to My opinion on it is you have people doing five million dollars a year on Amazon right now that are not paying attention or don't believe or not optimizing properly that are going to go to $500,000 in a couple years. And they're going to be like, what happened? And it's because they're not paying attention. Right now, in my opinion, you've got to ride the line. You've got to do the old way and you also need to prepare and start doing the new way. And the new way is going to start taking over. And just like you said earlier, you love the SEO stuff. And SEO right now, Google's a little freaked out because of, and a lot of the people in SEO world, which is a big business, are a little freaked out because they're losing some of their control of, you know, Backlinks still matter right now and there's still stuff that matters, but as more and more AI comes in and more and more people start using Perplexity and OpenAI to do their searches for products even, you know, it's only like point, I just saw a stat on it, it's like point, it's less than 1% right now of people that are doing actually searches in GPT or Chat, you know, Perplexity or OpenAI or Claude. To search for products, but they expect that to go to 14% I think by the end of this year and continue to grow even more over the next couple years. So who knows how fast this is going to go? It may be slow. It may happen this year. I don't know. It's moving pretty quickly, but at some point in the next six months to four years, it's going to be completely turned upside down. And a lot of people aren't preparing for that. So can you explain from the Amazon point of view, what's the difference between like Amazon Comprehend, like looking at images? I'm sorry, recognition, looking at images, to Amazon Comprehend, to Cosmo, to Rufus, to now there's another one. There's all these terminologies and people are like, well, what's what? What do I need to pay attention to? What do I need to do? Can you first just start off with just a basic primer on what's the differences between Comprehend, Recognition, Cosmo and Rufus? And I think what's the other one? Olympus. Can you just basically just explain those five? Speaker 2: Don't go into detail. Kevin King: Just explain what are each of those five and how are they different? Speaker 2: Absolutely. So for those who think that Amazon is not going to change, The bad news is that Amazon already did. And if you look from how it evolved, like with adding Cosmo, right, in 2024, we knew that that was about to change in terms of how Amazon sees products. But that was kind of like the tip of the iceberg because So, we've been trained to look at SEO. You mentioned SEO, Kevin. From a lexical standpoint, what does that mean? Word-to-word matching. We wanted to make sure that we have repetition, we have all those keywords in. The more we had, the more we indexed, the more we ranked. That changed drastically in the last couple of years because Amazon actually enforced the policies To avoid that, to avoid keyword stuffing and repetition because they leverage AI already and they've been doing that for some years. It's just that it kind of like fell off the radar for everybody is now it's like it's written on the wall and we have all this proof, especially with the update that they did on the algorithm with the last patent, the one from 2023, November, where it clearly shows how Let's talk about how rapid the process of data is and how easy it is for Amazon to recognize your product, what it is, understand it, and then do an estimate on where it should be on the search. With this ad, if you think about Cosmo, Cosmo is just the last layer that they added Fascinating thing, if I can call it like that. For me, it's fascinating how Amazon evolved. If you think about it, from the start, they always said that they have the customer at the center of everything that they do. And we knew that, right? It's just that it took them some time to kind of like focus on preferences and try to estimate what shoppers want. There are a lot of science papers out there that show how Amazon leverages AI for OCRs, for recognizing images, other than what you just mentioned on the AWS. When you look at recognition, that's a tool that can help you test your own images in order to understand how Amazon Possibly can see them, if that makes sense. So if I have an image of a mug, right, I want to make sure that I run that through recognition, which is just a tool, a free tool. Everybody can have access to AWS and recognition. In order to see and make sure that Amazon's tools recognize my item as a mug, right, or as a product or whatever it is, you want to make sure that it's labeled correctly. Why is that important? Because when you upload it on your site, you would You see only how shoppers interact, but you don't see the side of the coin that I mentioned earlier, which is the system and how it recognizes. Because Amazon utilizes different recognition systems. So for example, they have AVEN, which is the last one that they added last year, in which it kind of like tries to pull attributes from an image and understand the product. So if until now I used to sell, I would say, a V-neck t-shirt, and I would call it round just to get indexed also on round, Now comes Amazon in and says, no, this image is definitely not around. It's a V-neck, right? So they use Avon. That's one of the tools, for example. Then they want to make sure that, and this is fascinating for me, they also have a sort of like a safety net because AI hallucinates. So they created systems that help, for example, on Avon's side, make sure that it doesn't hallucinate. That's called something like Throne or something like that. So the system behind it, once you get into the science papers, is quite complex. But it shows you how much Amazon evolved and it evolved mostly in the last couple of years with the leverage of AI. So those who don't believe that Amazon is changing, it already did and it will continue to change. We do expect to kind of like evolve from here. Comprehend, on the other side, it's the text side of things. You want to understand the sentiment analysis. I did this and I tested. This is the interesting part. When I discovered Comprehend in the beginning, I was really focused mostly on the titles. I really wanted to understand how Amazon understands the titles in terms of the positivity score. And what I found over and over again, every time that I came out with a title that had a higher positivity score, my conversion rate in the A-B testing on Amazon would always improve. Because for some reason, it was more natural language, it resonated better with the shoppers, and therefore you would get better conversions. So for me, that was like, okay, now in my processes, whenever we would need to look at title optimization, making sure, you know, the basics are still there are important and following the guidelines, but then utilizing the AWS Comprehend as a tool in order to kind of like test that positivity score and then go and test the new, the new title in the AAB testing, but it never failed. Like 100%, it was always better conversion on the, On the version with the higher positivity score in the sentiment analysis from Comprehend. Kevin King: So how can someone actually use Comprehend? What's the process there? Can you explain that real quick? Because it's free, but you said something about scores. Can you just explain what that means and how someone listening, like, this sounds cool, Oana, but how do I do that? Speaker 2: Absolutely. The first thing that you would need to do is make sure that you register and get an account on AWS. And once you are in the console, you just need to type in comprehend. And then at that point, you get into the comprehend section and there's one box where you put the text in it, right? Like you copy paste the title, for example. And once you hit analyze, you can move down onto the page and you have the sentiment analysis. It's quite easy and straightforward and it will give you like if it's neutral, positive or negative and it gives you the scores on the text itself. And you kind of need to go back and forth between different versions of the title. So it's not like, you know, don't be concerned if in the beginning you don't get it right. It took me a while until I got my first title with the keywords that I wanted, the format that I wanted and a higher positivity score. And then I built a bot around it to help me with that. But yeah, it takes a bit of back and forth, but it's definitely worth it. Kevin King: You can do it with your bullet points and with your description and all that stuff too, not just your title. And by making those changes and getting a higher score on all those and then updating your listing, a lot of times you can pretty quickly see a lift in conversion rates and in exposure on Amazon and start showing up more places and oftentimes a little bit higher. Speaker 2: The traction is definitely better. And again, I'm always testing everything that I do. No change goes without testing because I really love data. And it never fails. It absolutely never fails. For some reason, the positivity score and the sentiment analysis, even a small improvement is always beneficial on your title. Kevin King: What is Cosmo and Rufus? What's the difference in those? Speaker 2: If you think about Rufus, Rufus is just your personalized shopping assistant. That's what they built it for. That's what it's intended for. Think about that person in the corner when you walk into a shop ready to help you out. That's Rufus. You want to engage with it. It's your choice. You prefer typing in the traditional way, let's call it, the searches and you type in keywords. You can do that. Or you can, you know, ask Rufus for help through questions and it will respond. First of all, it responds to your questions and then it recommends products. That's Rufus. With Cosmo, it's different. Cosmo is already applied on the system itself. It's integrated with the algorithm, and it just helps fill in the semantic gaps. What does that mean? It helps understand and recommend better products, connecting them to shoppers on every search. And here we are talking about traditional searches. What was fascinating was the ability to kind of like look beyond the traditional word-to-word matching. And that's what they wanted to understand. They wanted to go further in terms of what's the intent behind a specific search. The same principle was applied to Rufus and trained on Rufus. The only difference is that with Rufus, it's your choice if you want to interact with them versus Cosmo. Cosmo is there and we know from the science papers that it has been integrated from 2024 and when the paper came out, it was clearly explaining how it makes the connection between specific products and the I would say the meaning behind the specific search and at the time there was that famous example that everybody used with non-slippery shoes for pregnant women, right? So if somebody would search for shoes for pregnant women, the Cosmo framework would help combine that knowledge together and say, okay, it needs to be non-slippery. Not necessarily that we look for pregnant woman's shoes per se, but the benefit that lies underneath it. The same with coats. For example, you look for like a warm winter coat. It can come, for example, with a specific, I don't know, material that can help with that because Cosmo is able to connect the features to that intent when it comes to shopping and searching on Amazon. And the same thing happens on Rufus because even in the patent, it says that all it does, it connects The main purpose would be to connect features to benefits and to understand them clearly so it can then rank not only the responses that it gets but rank also the products afterwards. Kevin King: An example of that might be where it knows if I have a listing and a title, I don't know, it says wool, W-O-O-L, wool coat. Speaker 2: Wool coat, yeah. Kevin King: If someone goes into Rufus and types in, I need a winter coat for I live in Canada and I need a winter coat. Rufus will go and it will know, even though it doesn't say winter coat for Canada, which is the keyword matching way, it knows that wool coats is appropriate for Sub-zero temperatures in Canada. And so it will show that as a result. So that's kind of just a quick down and dirty example of what you're talking about, right? Speaker 2: Absolutely, yes. And Rufus would come in and would also explain that. It would be like based on, I don't know, shopper's reviews, this coat would be what you are looking for because blah, blah, blah. And he would give you like the reason behind that based on the information that he can pull. With Cosmo, if you go into the traditional search and the search bar and you just type in, the same product would show up. It's just that you don't have a response. So they're very similar in terms of... Kevin King: You don't have to understand our rationale. You just have just a list of stuff. Speaker 2: Mm-hmm. I just wanted to mention that clearly explaining your product Is more important than ever at this point on Amazon. Kevin King: So how important would it be in my pictures in this case this example of the wool coat to actually make sure I had one of my infographics showing that this is I don't know some special sheep wool from the high mountains of the Andes or something or whatever. How important would that be in the factoring of Cosmo and Rufus and the AI? Speaker 2: It definitely makes a big difference and if you think about it, it makes more of a difference from the shopper's perspective. So if you do have that information on one of your images, Rufus will be able to not only respond, you know, with acknowledging that that's the product that you might want to buy, but also pulls the image itself. If it cannot find that information or connect it to the image, it will only respond back as a text. And we know that images engage better and you tend to have a better conversion rate when you utilize visuals. My two cents would be make sure that whenever you have a differentiation point or I would say one of your strengths, make sure that you show them on images. So when Rufus will reply back, It will pull the images and therefore your chances of conversions actually skyrocket in that moment. Kevin King: What's Olympus? That one a lot of people haven't heard as much about. What's that one? Speaker 2: I haven't found the paper on it yet. I will be honest with you. But it comes down to product recommendations. From what I researched, they mentioned it just comes down to the relationship between Let's talk about products and how they get recommended and the interaction between different types of products when you buy them. And that was something that also was mentioned with Merlin. I'm an AI that helps, it's called Merlin, that kind of like understands and analyzes the behavior when you buy different products and making sure that when they need to recommend a product would be suitable. I will make an example just to help out. So if you buy a phone today and a phone case, right? Two months from now, Amazon will want to recommend another phone case for you to buy, but not the phone again, because it understands you will not purchase another phone. You don't need it. You need a case to protect the phone that you already purchased. Kevin King: Because it figures in two months you've worn it out or damaged it or something. Speaker 2: Yeah. Kevin King: Okay, so when you are going in and looking at a client's listing to optimize it, and I've heard stories of some people where you've gone in for sellers and took a look at their listing, like, oh, you need to add this and add this and do this and do this, you're missing this. And then a lot of them are blown away, like, holy cow, I had no idea I needed to do all that stuff. What are some of those things that you're going in and doing? What's some low-lying fruit that people that are listing, that they should go check right now on their listing and make sure it's there or modify? Speaker 2: The number one thing I would say, go and check your back end because that was kind of like, you know, in a house, whenever you have things that you don't really want around within the house, you kind of like put them in the basement. That's kind of how we treated our backends for so long and all of a sudden product attributes and the backend of our listings are way more important than we think because all of the systems that we mentioned, Cosmo, Rufus, Avon, Merlin, they all pull information overall on the listing and a big part of it would be the backend. So making sure that that's optimized That's the place that I would start. Kevin King: Let me go and download the category listing report and looking at every single field and making sure there's something in every single field that you How do you know what to put in some of them? Because some of them, Amazon has, you cross-reference it to the flat file report. You've got the little tabs. It'll show you, these are the values that should be in there. We're expecting one of these five values. And sometimes there's no values. Sometimes it's like adult kids and women and you're like, no, this is not the value I should put in here. So how do you know which ones to do and what you should put into them? Speaker 2: It will depend from product to product, but the ones that are mandatory, those should not be left out. Those that are optional, of course, it will come down to you understanding if it's relevant to adding the information or not. But let me tell you this, if a field is, it is optional, but it can be, I would say helpful from an optimization standpoint. Please don't avoid it. Literally, just make sure that you put in the information. I don't know. No easy example comes to my mind, but maximizing all that, it really makes a difference. Another thing would be people think that you have your A-plus or your premium A-plus, and I really encourage you to have a premium A-plus, not a normal A-plus. You would neglect the The description in the backend. From testings that I did, I realized that Rufus is able to pull the exact information, kind of like how I phrase it, from my description that's in the backend. And that's really powerful if you think about how you would want Rufus to respond maybe to specific questions. When it comes to details to the product. So make sure that you utilize that section. Again, that's something that it's not necessarily, you know, you can put a word in it and that's it because you think about it, you have your premium A plus and that's what matters on the front end. But the reality, the backend description helps and really makes a difference, especially with Rufus. So I would really make sure that that's maximized in terms of the characters that you can add. Kevin King: Do you think Rufus, it's in testing, it's kind of like in beta now and there's some people that really love it and there's some people like, this is giving me garbage and it's making some really bad recommendations. So they're fine-tuning it and they're getting it. It'll get better and better with time, especially as more and more people use it and learns more and more. But do you think it's going to get Moved into that search bar toward the search bar right now where you can go still type keywords and stuff and find results Do you think that's that search bar? It's gonna be more like a it's gonna get replaced with a Rufus type of thing Do you think they'll always be separate where people can search by keywords and they can search with a Rufus interactive type of? Speaker 2: That's interesting. I think it will kind of become a hybrid in between the two. One place but for like you either choose to ask a question or you choose to type in a keyword, most likely you will start getting the same type of responses back at one point. I think it's actually growing and you are right, especially in the beginning. I remember when they released Rufus, everybody was like, oh, this is just horrible. It hallucinates. It would pull information that would not be there. But then we learned why it did that. I haven't seen that in a while. There are no more posts complaining about Rufus. It doesn't show that it hallucinates that much anymore. In fact, from all the interaction that I'm doing, I'm not seeing hallucinations anymore. That shows and proves how fast it learned. Because if you think about it, it will soon be a year since they released it to everybody in the US. And in the beginning, it was garbage. But then it really started getting better. So from my point of view, it's like, you know, it's still a learning curve for us how to interact and how to use it. But ultimately, they will blend it in. Kevin King: Yeah. At the end of February, Amazon released the new Alexis LLM and and you know at the time of this recording we haven't had time to play with that yet to know but what what are your thoughts on voice AI and voice AI search and everything now unfortunately when we're recording this we haven't had time to play with that but but that's out there now and how do you think that's gonna affect because a lot of people are now are just talking to their phone you know Siri is now powered by a new system it Siri couldn't do that can make it can really interact with you just like you know open AI came out where you can call an 800 number call 1-800 ChatGPT or whatever it was and you can actually talk to it and now that Amazon's doing that and then Apple's doing it with Siri, where do you think this is going to go when it comes to voice search? Speaker 2: I'm a type person. I always like to interact with my phone through text for some reason. It's just a personal preference, but I cannot not notice and admit the fact that people like interacting with their phones through voice. We know how easy it was to release Alexa and then with Siri and how people interact with that. Now it's just taking the game at the next step. And again, I think it will just continue to grow from here because if you think about it, it's easier to talk rather than to type because we are so used to multitasking so many times, right? So you can take notes or ask it to help you with a specific search. Ultimately, I also see us doing that when we are looking for a product on Amazon, like, hey, help me, you know, buy this product. I want it like this. And then maybe it comes back with some specifications in terms of this is what I found. I think this is what you would like. Because it, again, comes down to a personalized experience. Kevin King: Well, I'm looking here at the time. We've already spoke for an hour, so they're going to get mad at me over here at AM, PM Podcast. The editor's like, we're going to have to finish, continue this on another one because we could still geek out on a whole bunch more cool stuff, I think. Speaker 2: I would love to. Kevin King: If people want to reach out to you, want to find out more about you or follow you or learn about, I know you have an agency and stuff. What's the best way for them to get a hold of you? Speaker 2: LinkedIn would be my place to go. That's where I kind of like share everything with the community and My go-to place at any time. I'm not a very social like Facebook, Instagram person, so LinkedIn is definitely the place for geeks to be. If you have any questions or anything, just reach out on LinkedIn. I will get back to you as soon as possible. Kevin King: For someone listening, how do they spell your name and find you on LinkedIn? Speaker 2: It's Juana. And then once you start typing Juana and you go with O-A-N-A. Kevin King: A lot of people don't know how to spell that. Speaker 2: Yeah, O-A-N-A. So Juana, it's definitely. And I assume that if you're in the Amazon business, The algorithm will help you once you start typing P-A-D-U-R. The good part is, Kevin, that I'm the only Oana in the industry, right? So whenever people ask me, what's your name? I'm Oana. Don't try to pronounce my family name. It's almost difficult for Romanians as well. I'm exaggerating, but it's not an easy name. I don't want to put a burden on anybody. Kevin King: How do you pronounce it? Speaker 2: Normally, it's pronounced Oana Padurariu. Kevin King: Okay, we'll just stick with Oana. Yay! Speaker 2: I told you. Kevin King: Well, I appreciate you coming on and sharing today. This has been fun. Speaker 2: Thank you so much. It's been incredible and I look forward to the next one. Kevin King: Gonna have to do another podcast with Oana just to talk more about what's happening in AI, especially as everything's changing. Want to get her back on maybe later this year. To actually talk about the latest stuff that's going to be coming out when it comes to Amazon and AI. If you want to keep up with the latest that's going on in that, in the meantime, you can always subscribe to the Billion Dollar Sellers newsletter. It's BillionDollarSellers.com, BillionDollarSellers.com. Totally free every Monday and Thursday. I put out a brand new edition. So if you're not a subscriber, check that out at BillionDollarSellers.com. We'll be back again next week with another awesome episode of the AM-PM Podcast. In the meantime, remember that whatever you pay attention to expands. Where your attention goes, energy flows and your frequency is what you frequently see. Speaker 2: Take care. We'll see you again next week.