#148 Ensuring Authenticity: How Amazon Fights Fake Reviews with Rebecca Mond from Amazon

#148 Ensuring Authenticity: How Amazon Fights Fake Reviews with Rebecca Mond from Amazon Unknown Speaker: I need you. Speaker 1: Welcome to the Fearless Sellers, The Women of Amazon podcast. I'm Joie Roberts. Speaker 2: So we investigate all reports of abuse. Um, so anytime anybody hits report and review, um, you have that option. So the report, uh, report abuse button, it's for anybody who doesn't know it's right next to the review. Um, and you have the option to kind of drop down and say, what kind of abuse do I think that this is? And so every time that that button is clicked, our investigators are looking at those reports to see, all right, what's going on here? And we're taking that information and again, feeding it back into our tools, auditing our detections. Are we making the right decisions? Speaker 1: Welcome to the Fearless Sellers, The Women of Amazon podcast. Rebecca is the head of external relations for customer trust and abuse prevention at Amazon. Playing a key role in ensuring that Amazon remains a safe and trustworthy marketplace by tackling big issues for sellers like fake reviews. I am looking forward to discussing how Amazon is working to maintain the integrity of customer reviews and ensuring that sellers and consumers can confidently rely on product feedback. Rebecca, welcome. Speaker 2: Well, thank you so much for having me. Speaker 1: All right, let's clarify. What is a fake review? Speaker 2: So this is a weird philosophical question that I probably spent way too much time thinking about. Let me start with what is an authentic review and really like what's the heart of Amazon reviews and why we spend so much to protect the integrity of our reviews. So little history lesson, back in 1994, Amazon pioneered customer reviews and a lot of people just did not understand Why would we put positive and negative feedback right next to something that we were trying to sell? There were media headlines that said Amazon has lost their marbles. We got letters saying, no, this is the wrong thing to do. But for us, we saw the value, the long-term value, the benefit for our customers in that transparency, providing them authentic feedback in terms of what it was that they're about to purchase. And so they can have that holistic picture and really get to know, is this the right thing that they want to purchase? And are they going to love what it is that they're going to buy? And so that is really at the heart of what it is that we're trying to provide in our customer review experience is give customers that feedback when they are making a purchase decision, when they're buying something. And so we are looking for reviews that reflect that type of information. And so if you look conversely at a fake review, a fake review is something that would mislead a customer, that would give them information that doesn't necessarily help them when they're purchasing the product or that would give them false information. So another example is if it's biased information. So my mom, if I'm a seller, And if my mom writes a review, we would say that that would be a biased review. My mom loves me very much. She would write me just glowing reviews, just no matter what it is that I'm selling. But then if it's something like if I am purchasing reviews, that would be another instance of fake reviews. So we have a really a list of different types of reviews that we see are fake reviews and that are banned from our stores. And that we do not allow our sellers or our customers or sellers to solicit these types of reviews or our customers to write those types of reviews. Speaker 1: All right, thanks for breaking it down. And you just brought up, you know, your mom would write you a biased review and that's a really good point. I'm a seller. I've been selling a long time and I'm also a seller's coach. Like I coach a lot of people and all sellers, I know all of you listening have thought this too, like, why can't I just ask my cousin or my cousin's boyfriend to write me a review? And you kind of just cleared it up like it's a bias, but can you touch on it a little bit further? Because when it gets a little farther and farther away from you. Speaker 2: Absolutely. When does it actually become a fake review? When is it no longer a bias? And that's something that we have proactive detections that are in place and machine learning models that are consistently making these types of decisions at scale. And it's certainly something, you know, it's a challenge to set up that policy and say, here is the line. When we say this is, you know, this is a authentic review, we can put this on our store and this is in the fake category. But we have, our tools are consistently making these determinations every day, every second to make sure that our customers have the best buying experience. So back to your sellers who are coming to you and saying, oh, why can't we just have our cousin write a review? And it's because we really want to make sure that our customers have that type of information that is going to help them when they're purchasing a product and not just glowing reviews. We want the authentic reviews. We want the real stuff. Speaker 1: Yes, that makes sense. And as you're talking, it reminds me as a seller of Amazon's, one of the values, like customer obsessed, right? That's a big value of Amazon. Speaker 2: Yes, absolutely. We start with the customer and work backwards from there. That is one of our leadership principles. Speaker 1: Yes, and we're all blessed to be selling on the Amazon platform. And good reminder, we got to play by the rules and not push on the bias. So I always say to anyone who asks me as an experienced seller, I'm like, if you're asking me this question, like if it's okay to ask for this review, then you probably have that gut feeling that it's not okay. Speaker 2: Well, it's really important for sellers to familiarize themselves with our policies because there's definitely instances where it's like, okay, like, oh, I'm not sure if this is okay. So, for example, a seller can absolutely ask for a review. Please leave a review, but they can't condition it on, did you have a good experience? Then leave a review. If you had a bad experience, can you email us? That's then bringing in some of that bias again that we're not necessarily, we're not looking for on our stores. So if a seller were to ask their customers for reviews, they need to leave, ask in a neutral manner. Doesn't matter if you've had a good experience, a bad experience, anything in between, we want to get your reviews. Speaker 1: Yeah, and you can do that with a product insert, right? Speaker 2: Absolutely. Absolutely. But again, just have to make sure that you stay by that neutral tone. And then there's also an option within Seller Central that sellers can click on a button and that will email the customer and ask for a review in a templatized format. Speaker 1: And what's with the five-day policy? Like, you have to be five days from when they received it to click that button. I always get that, too soon, can't send it. Speaker 2: I think that we want to make sure that customers have an opportunity to use the product, familiarize themselves with the product, and not just, you know, get the product, look at it, and say, oh, this seems legit, and then write a review. So, we want to make sure that there's that experience there, too. Speaker 1: And so how are you detecting? I know you mentioned like, there's like an algorithm to it. But like, I imagine that there is a lot that goes into detecting and also preventing fake reviews. Speaker 2: Absolutely. There are thousands of data points that go into these determinations and these detections. And again, these are just happening simultaneously. They're happening at scale. Our machine learning is constantly, these models are looking at new reviews before they're posted on our stores. We are going through behaviors, we're going through relationships, we're looking at past patterns of behaviors with the seller, with the customers, and all of this is happening before the review is posted. And even after the review is posted, we're looking at new information, new types of abuse, and then we're assessing, okay, wait a second, did we make the right decision? And letting that review up there, do we make that right decision and making sure that the, or in taking a review down? So we're constantly assessing and auditing these decisions. Speaker 1: You just mentioned the word relationship and that was like a key thing to me that there is part of the algorithm, the machine that's tracking, looking at relationships. So it might not be just a relative, but like, what does that mean? Like somebody you've sent a gift card to in the past? Speaker 2: What things like, so it could be relationships between competitors. We're looking for that type of abuse. Like if all of a sudden you're seeing a whole bunch of negative reviews that are showing up on a product that's, and there's some suspicious activity associated with that. But then most importantly, so one of the main things that we have seen and the biggest trends that we have seen has been this rise of what we have called a fake review broker industry. Years ago, a lot of the abuse happens directly on our stores and is orchestrated on our stores. But our tools and detections got really good at stopping that type of abuse. And so the bad actors had to start orchestrating the abuse off of our store. I work on social media and on encrypted chat services and creating groups where they, the fake review groups. And so they would tell sellers that this is, you need to be using our service in order to be successful on Amazon and then work with like a grassroots group of reviewers that would be willing to write fake reviews for an incentive or for some sort of a discount. And so what we've done in the past few years has really targeted this fake review broker industry and gone after the brokers and shut down these groups, shut down the social media service, the chats. And so that way we're stopping the fake review abuse from the source. Last year, or sorry, in 2023, we had, we had taken legal action against over 140 bad actors that had attempted to solicit fake reviews. On our store, and these legal actions are working to really reduce the number of fake reviews, not just on Amazon, but elsewhere, because if this broker is trying to target Amazon, they're also trying to target other retail stores, other types of services that use fake reviews. And so, you know, it's really important for us to be working together in order to stop this broker industry. Speaker 1: And is there, what are the consequences for them, the people who have those like grassroots rings, and then also consequences if, for sellers, like if we do it too much? Speaker 2: Yes. So we've been taking legal actions against these fake review brokers. We've been taking a very aggressive stance against this with zero tolerance for fake reviews. We want to make sure that we are encouraging an environment where everybody is able to play by the rules in ensuring a trustworthy review experience. In terms of sellers, we take a variety of enforcement actions, and these can include warnings, they can include suspension, and of course blocking I'm the seller from the store. This is something that we want to make sure, again, it's so important for us that we're maintaining the integrity of these reviews. And so we take strong enforcement actions to stop the abuse on our store. Speaker 1: Yeah, competitor abuse, you brought that up and I see it Well, and I've experienced it myself actually. I mean, I don't have a total proof, but sellers listening can kind of like relate, but you see these one star reviews come in like three, four at a time in like a span of two, three weeks. And like, it's either a competitor or somebody who's just, you know, attacking me right now. But I sell borderline, it's a natural cream, but it's borderline supplements. And I feel like that category can be like ruthless. And, and as a seller, it, it really, I can handle, like, I'm tough. I can handle a one-star review, but when I see like three, four of these come in and my star is dropping to like 4.2, it's, it's really, I know I click that report button, but I don't feel like it goes anywhere. Speaker 2: So we investigate all reports of abuse. So anytime anybody hits report and review, you have that option. So the report abuse button, it's for anybody who doesn't know, it's right next to the review. And you have the option to kind of drop down and say what kind of abuse do I think that this is? And so every time that that button is clicked, our investigators are looking at those reports to see, all right, what's going on here? And we're taking that information and again, feeding it back into our tools, auditing our detections. Are we making the right decisions? And, you know, there's definitely instances where I've heard from sellers and I would go, that can't be, why am I getting these, you know, these one-star reviews? Unfortunately, you know, it's something where if we disagree with what they're saying, that's not something that we can take action on. But, you know, we understand that there's just sometimes you get that spidey sense and that's why we're always looking at those reports of abuse and why we invite our sellers. We invite customers to please, if you do suspect that there is abuse, To report that because we do take that information in and we are looking at that as part of one of the many signals that we're looking at. Speaker 1: And then what'll happen is I have like a really cool marketing team that does design for my products with me and they will get heated over the one-star review. And then there's like four of us, five of us like hitting the report, report, like let's tackle this. Does that help at all if more people report it? Speaker 2: No, especially if we can tell you guys are all right next to each other hitting report, report. So no, that doesn't necessarily change it. But again, it's something where we're looking at these signals and we are constantly ensuring that we are updating, taking in this new information and using it to improve our detections and make sure that we're providing customers as well as our selling partners. We know how important reviews are for our selling partners. And we want to make sure that we're providing that great review experience for them as well. Speaker 1: Yes, I, you know, I started selling, I think it's been like five years and I was on this big race to get reviews. Like, okay, I gotta get reviews, tons of reviews. And maybe this will make all the listeners feel better. I launched my hero product, like my product that does the best in sales for me right now, over a year ago. And I am, I'm in like, I'm at like 93 reviews. Like I haven't even passed a hundred reviews yet. And I keep telling myself like, it's okay because they're authentic and they're growing. And I am shy to hit that review button. I don't know why, but I'm like shy to automate it or use it. So I use that gift that Amazon gives us to be able to send those, the emails out. And I've still gotten, you know, close to over a hundred. So patience is part of this new age of selling. Speaker 2: Yeah, patience and we understand how hard it is for sellers that there's almost like this chicken or egg situation where how do you get a review without a sale, but how do you get a sale without a review? And so that's one of the tools that Amazon had created was Amazon Vine program. And it was intended just for this to support this situation and to help sellers get that first review or to help slow moving products and to get those reviews and to get that to get customers visibility into those products. So, absolutely, you know, hit that review button and, you know, encourage your customers, encourage everybody to write the reviews. I think it's so important because it is quantity, but it's also the quality. And you may have seen that we have the new feature on our reviews where the review summaries are at the top. And so those summaries are something that I personally, as an Amazon shopper, rely very heavily on those. And I look at those. I sometimes also look at the reviews. I'm a working mom, don't always have the time to scroll through all those reviews. And so it's great just to have that summary up there. And so, you know, as you mentioned, that quality, it's, you know, those quality, authentic reviews that helps with that summary, making sure that you have a rich summary up top. Speaker 1: Yes. And thank you for bringing up the Vine program. I've used it. How do you select your Vine people? Speaker 2: Yeah, so we have a robust community of Vine reviewers that are selected based off of the, you know, they're given high quality feedback, they're based off of their trustworthiness, and this community, you know, we hold them to a high standard. We want to make sure that they are providing our sellers and our customers with a great review experience. And so we, yeah, so again, it's based off of the quality of feedback that they give as well as their trustworthiness. And we protect that community. We make sure that the seller and the Vine reviewers do not necessarily have, that they're not able to contact each other. We are collecting the product. The Vine reviewer then can select which products to write reviews on. We think that creating that barrier is really important so that way we can still maintain that integrity of the Vine reviewer community. Speaker 1: What happens when, and I'm sure you see this a lot, and I see this a lot from sellers, the Vine reviewer gives them one, two, three stars. The seller is reporting a Vine reviewer. Speaker 2: So again, a fake review is not necessarily a review that the seller disagrees with. And the Vine reviewer, we are looking for not just the five-star reviews. We want the authentic feedback. It holds the same principles as our entire customer review experience. We want to make sure that the Vine reviews are just as authentic as any other review, if not more so than a customer would be reading. Speaker 1: And I hear a lot of people say the Vine reviewers are required to give stars across the board. It can't just be all fives. Speaker 2: I'm not sure if that's a requirement, so I would have to double check that. Speaker 1: Awesome. All right, that might be a myth that we're busting here. I'm still going to tell people that to make them feel better, though. Speaker 2: Okay, there you go. Unknown Speaker: I love it. Speaker 1: Yeah, I laugh, just side note, I laugh when I get a three-star review. I'm like, who took the time to leave me a three-star review? Like either give me a one or a two or like a five, but a three, you're like, oh, the product's okay. Like, come on. Speaker 2: Yeah, again, I've spent a weird amount of time thinking about reviews and that's something anytime, like I'm in an Uber or reviewing a restaurant, it's something I think about, like, when's the last time I've left a three-star review and isn't that like, you know, what's... What does that mean? So, I love the fact that you do see those three-star reviews. I feel like it's, you know, somebody's putting some thought into it. And, you know, when it comes to our reviews, we want the good, we want the bad, we want the in-between. And I feel like that three-star hits that in-between. Speaker 1: It does and now with the sentiment that it has at the top and you can see the positives and then there's always like a little bit room for improvement. So as we're all product developers when we're selling on Amazon, so it's an opportunity to improve your product. Speaker 2: Yeah, it's great feedback for sellers to take in and to make improvements on their products and or to make adjustments and you know, just to improve their sales and improve the customer experience. Speaker 1: So if a seller is really, really frustrated with reviews and the report button didn't do anything, like, like what options do sellers have with Amazon to raise attention to them feeling like they're getting a lot of negative reviews that are And I'm here to talk to you about how we can make sure that our customers are not, you know, I don't know, for lack of a better word, malicious or meant to hurt their business. Speaker 2: Yeah, it's continuing to hit report abuse on new reviews. And because, again, we're looking at all of those reports of abuse and just continue to work through that system and making sure that that information is getting to our teams. There are thousands of dedicated professionals that are looking at this. These are investigators. These are data scientists. These are people whose eyes light up when I talk about LLMs and what's the potential for, you know, the next generation of AI and detection for fake reviews. And so these teams take this very seriously. This is our livelihood. And it's these decisions, they need to be made at scale. They need to be right. They don't take lightly enforcement action against a seller. You know that these are their livelihoods and so they're very careful and they want to make sure that they are making the right decisions. And so it might not be something that the seller always agrees with, but know that there is just so much thought that goes into that. And process that goes into it. And it's always iterating. We're always here. It's always day one at Amazon. So we're always making sure that it is something that we're improving to ensure the customer and the seller experience is just fantastic. Speaker 1: Yeah, I think that's fair and I can, you know, know from this conversation that there's a lot of resources going into it and it's like a very serious matter and it's evolved so much over the years that I've been selling, which, you know, Amazon's been around a lot longer. Let's talk about that, like looking ahead, the future of reviews. Consumer behavior and AI technology. There's so much advancements at Amazon. So what else, what can we look forward to? Speaker 2: Well, before I go into this, I have to caveat with the fact that I majored in government and so I don't speak tech very well, but I'm going to do my best. But if I say something wrong, just please don't come after me. No negative reviews, please. But I would say that when I talk to our data and our tech teams, one of the things that just makes their eyes light up and get really excited about is when we do talk about Gen AI and the potential for that to detect abuse at scale. So what we generally do with our AI tools is we're looking at behaviors and patterns and seeing like, all right, is this a normal behavior or is there an anomaly? And what's happening here that would signal an abuse. What LLMs are giving us the ability to do is to also look at anomalies in language. And to make some of these determinations at scale that right now we are using, or maybe not right now, but in the past we've used more our investigators and people have had to make some of these decisions. So let me give you an example. Let's say if I get a product and I see an insert in it and I say, oh, this insert, it offered me $25, but I still hated the product. That seems like a pretty legitimate review. It's unbiased. That's my honest opinion. But our tools might look at, say, the flag and say, oh, she got an insert. That's probably a fake review. And it would take somebody to kind of look at that and say, oh, no, no, that's a legitimate review. Let's keep that up there. But what LLMs are now giving some of the capability to do is to look at some of that at scale and to detect those types of anomalies and make those determinations and keep those types of reviews, those authentic reviews up. While taking the abusive reviews down so it's it's massive improvements and some of those those determinations Making them faster and making sure that we have that better experience But then I'd also say just you know, even the summaries those review summaries. I think that that's such a cool customer experience we're making sure that customers just have the right information in front of them and When they are deciding whether or not to purchase the product. So I think that that's just a really cool area. I'm excited to see that evolve as a non-tech person. I think it's great to see it and to learn much more about it. So I think that that seems to be where the future of reviews and authentic reviews and review experience might be going. Speaker 1: Yeah, that's, that, that does make sense. And when you say LLM is that's the language model or what is that? Okay. Just making sure I understood. Speaker 2: Awesome. Speaker 1: Yeah, it is. It is also, um, the AI is cool, but it's the human element of having to adjust as a seller. To slowing down and being more aware that the technology is coming and it may flag things. So that brings me to my next question. If the technology is flagging something, is there like a bucket and then somebody on your team has to go through and review it all? Speaker 2: Hi. So our technology is making these determinations at scale. And then we have a team of auditors that are consistently inspecting those decisions and they're consistently inspecting the models. And there are some times they're They're looking at, all right, is this model making the right decision? Is there changes in what's happening with this model that we need to look at? Like, why are there more or less of a certain decision being made? And so we always have these investigators, these auditors, as well as just the data scientists themselves, the ones who are creating these models and updating them that consistently have their fingers in the pot and making sure that these are the right decisions. So it's really this interesting intersection between both with the human investigators as well as the machine learning models that are working together just to consistently improve the system. Speaker 1: All right, now I know when I hit that report button, I'm like, all right. Speaker 2: There's magic that goes on behind the scenes. You have no idea. Speaker 1: I'm like, I'm making all of you work harder today because I'm reporting all of this. Speaker 2: Exactly, you're making us earn our dollar, so do it. Speaker 1: Yes, well, it's job security because there will always be reviews. Awesome. So you've been working at Amazon for quite some time now. Speaker 2: Yes, I guess a little bit over four years. Before that, I was working for a trade association representing the toy industry. So safety, consumer trust is always kind of near and dear to my heart, as well as kind of supporting small businesses and making sure that they have a great experience. It feels like it's kind of like a natural trajectory for me. Speaker 1: Yes. Well, that's awesome. Congratulations on your success. What city are you in? Speaker 2: I am in Arlington, nearby Amazon's HQ2, second headquarters. Speaker 1: Oh, okay. So it's the second biggest in Arlington. Speaker 2: Yes. Yes. How about you? Where are you based out of? Speaker 1: I'm in Austin, Texas, and there's a pretty big Amazon presence here. Speaker 2: Yes, there is. There is. Austin, Texas that I had no idea had the large boating industry, but it flurries up today. Speaker 1: Yeah. Well, you know, in the summer when it's like in the hundreds, we all have to either be on the lake or in a pool. Speaker 2: Absolutely. Well, in the summer over here, it's just 100% humidity, so you might as well be in a lake over here. Speaker 1: Yes. Awesome. Well, Rebecca, it has been an absolute pleasure to have you on the Fearless Sellers, The Women of Amazon podcast. And until next time, stay fearless. If you're already selling on Amazon or you're looking to get started and you want my help, go to amzfearless.com to book a free strategy selling session. We can see if we can help you out. That's amzfearless.com. Talk to you soon. Thank you for listening to the Fearless Sellers, The Women of Amazon podcast. Until next time, stay fearless.