The Retailer That’s Obsessed With AI

0:00I always think spring is a perfect excuse to dress a little softer and a little bit more romantically. And apparently Macy's fashion experts are saying that pastel dream is the trend to know right now. What I love about it, it's not really going like full head-to-toe pastel. It's more about those softer details that make an outfit feel fresh and a bit elevated. Think delicate florals, lace, ruffles and sheer layers alongside flowy dresses or tailored pieces in lighter tones. It feels polished, pretty and very spring, but still easy to make your own. And you'll find it all at Macy's.

0:46Cathaleen Chen: Hello and welcome to The Debrief from the Business of Fashion, where each week we delve into our most popular BOF professional stories with the correspondents who created them. I'm Executive Editor Brian Baskin. And I'm Senior Correspondent Sheena Butler-Young. We talk a lot on the debrief about AI, but for much of fashion, artificial intelligence has been approached with cautious exploration, excitement, but with a healthy dose of skepticism. Then there's Revolve. Yeah, I used to think we talked a lot about AI, and then I listened to Revolve's last earnings call where AI was mentioned 29 times.

1:20Cathaleen Chen: And just to put that in perspective on Apple's last call, and just to remind you, Apple is a tech company. Revolve is a fashion retailer, for those who don't know. Apple cited AI, I think, 14 times. So Revolve was double Apple on this. They are obsessed with AI. And that's not new news for Revolve. Just months after ChatGPT launched in November 2022, Revolve released what it said was the first AI-generated billboard at Coachella. It has since gone on to build its own in-house search tools that are actually moving the needle on conversion. But the real test is coming now. Revolve has started to open its first physical stores, and it's a real question whether they can take this data-first, AI-centric approach and inject that into the world of brick-and-mortar retail.

2:07But can an algorithm actually tell you where to move a mannequin to sell more clothes? Can it turn a physical boutique into a living, breathing A-B test? Today, we're joined by retail editor Kat Chen to help us understand why Revolve can't stop talking about AI and whether the tech can truly power its next era of growth. Kat, welcome to The Debrief. Thanks, guys. Happy to be back.

2:29Cathaleen Chen: Yeah, welcome back. So Kat, before we dive into the AI of it all, I think you have to answer a question that is probably on the minds of a few of our readers, maybe the over 40s who haven't been to Coachella in a while, which is, what is Revolve and why are we spending an entire episode on them? Well, you know, when I think of Revolve in relation to technology, I always think of the chicken or the egg analogy. Because while Revolve looks like a shop bop or a net-a-porter, but for younger Gen Z and millennial customers, Revolve is actually built like a data science company. Its co-founders, Michael and Mike, were co-workers at a software company.

3:07And they didn't set out to build a fashion company. They were interested in e-commerce at a time when e-commerce was just in its nascency. And at that time, there was very little penetration of fashion online. And so fashion just sort of fell into their laps as a white space. And that's how they tell the story. And so is Revolve a tech company? Is it a fashion company? It's definitely both. But technology and data science in particular is its foundation.

3:33Cathaleen Chen: Yeah, it's funny to read early profiles of them because they're talking about this idea, Like what if every click on a retailer's website was a data point? And back then in 2003, that was the hot new idea. And today everyone's doing that. So they were really early to that game. Totally. And I think for a while as a reporter, it became really tedious to cover because then in the 2010s, every direct-to-consumer company was billing themselves as exactly how Revolve built themselves, right? Like every purchase, every click is a data point. And, you know, it's funny with the Allbirds news today, but a lot of that didn't pan out.

4:11Except Revolve is still very much a living, breathing data science company. And AI has been just a really big unlock for them.

4:22Cathaleen Chen: Sure. I mean,$1.2 billion in sales last year definitely a success. And I think the other side we do need to talk about here is what they represent culturally and fashion. I mean, they were really known for that festival wear Coachella thing. But it seems like they've managed to outgrow that, which is a good thing because I think that aesthetic has kind of faded a bit, right? For sure. And I think they've also pulled back from their investments in Coachella marketing in the past couple of years. It's definitely skewed toward a very young consumer. They are known for wedding guest dresses. And you couple that with amazing customer service.

4:59So they have two-day delivery. They have free returns on anything. And so they also have a very high returns rate. I think over half of all of their purchases are returned. But that's built into their business model. And so even if you're not into that Revolve sort of millennial goes to brunch and Coachella aesthetic, the customer service is undeniable and it's irresistible, in my opinion. Can we go back a little bit to the foundation of this? Because Brian's right. There were a lot of companies. I think you said the same thing in the 2010s. there were going to be tech companies that do fashion, by the way.

5:36That applies to Nike and other non-DTC brands as well. Revolve now is calling itself having a culture of AI today. What does that evolution look like? What does that look like in real time to have a culture of AI as a fashion tech hybrid company? So the way that Michael and Mike described it to me is this both a top-down approach where there is a lot of encouragement from the two of them. You know, Mike is actually an engineer. And in my last interview with them, his partner, Michael, was like, if Mike wasn't here, he'd be starting his own AI company. And so there's that directive from the top, but also from the bottom up where individual employees are encouraged to play around with AI, test it out, and then present it to the entire company when it's effective.

6:28So a lot of these applications of AI, and there must be dozens of applications from internal reporting of research and company data to using AI to turn some images into videos. There's got to be dozens, if not hundreds of uses that Revolve is deploying. And one really interesting anecdote that they shared with me was that even if something improves efficiency or output by just 1%, that's considered a success.

7:06Cathaleen Chen: Let's talk about some of those uses. What have been some of the most successful ways that they've rolled out AI? I think the most remarkable example is how they've built their own program for search results on the website. And so it used to be that every product has tags, descriptor tags. So color, texture, silhouette, however many tags that can fit into the description of a product. But AI allows retailers to sort of automate that tagging process by analyzing an image of the product. Revolve used to use a third-party tool that does this for them. Except they've now built their internal tool that is so effective that it's effectively outperforming the industry leader, this third-party software that they were using.

8:05And I can't remember what the statistics was, but double digit increases in both engagement and in conversions through this search tool that they built alone.

8:17Cathaleen Chen: Can you give an example to illustrate that? I mean, what's a search that this kind of tool can process that like a standard, you know, keyword to meet tag search can't? Absolutely. So the example that Mike and Michael gave me was like, if you search party dress, so there's no color. There's no shape. There's no sort of concrete descriptor. And what AI, their AI tool is able to do is pull up anything that is sequined, for example, or textured or has a mini length. Like party dress might not exist as like a meta tag for any of these items, but the AI is able to pick up on these attributes as related to something fun for a night out, a cocktail dress, et cetera.

9:05et cetera. And so obviously, you know, this is a matter of trying to give the customer exactly what they want. And that's the end goal for any fashion retailer, for any brand, right? And so I think the best uses of AI that Revolve is exemplifying is in this sort of search, recommendation, personalization. It's not necessarily fully personalized, but it is anticipating the desires of the consumer. And that also shows up in what the homepage looks like. If I'm on Revolve often, my homepage will look different than yours, right, Kat? I think one of your sources in the story talked about personalization and curation being a big opportunity here for AI.

9:49I think the example it said was like, otherwise it would look like Amazon from the early 2000s, basically, if they didn't do this. I will say the personalized homepage functionality is pretty common in retail. Essence actually does it really, really well. But it's a very great function for these websites that have tens and thousands of SKUs. Revolve, I think, has over 100 ,000 products on its website. So parsing through them is not only difficult, but very much not something that you want to do. You know, especially let's say I'm shopping for a dress to wear to my friend's wedding in the summer.

10:33They have hundreds of thousands of dresses on the site. I'm open to styles, but I know what I like and I know what I don't like. And articulating that is very difficult for me. It's very difficult for the average consumer. And so being able to click on a dress and then see recommended dresses that are sort of related in aesthetic to that dress, how close those recommended dresses are to the dress that I've landed on is critical. That is what they're trying to achieve. And they're using AI to sort of match my implied desires with the other products that they're serving up. And this comes up in many ways in terms of how Revolve is using the AI recommendation tool.

11:24It's been very effective for changing landing pages. So if I'm looking up address on Google and I go to a Revolve link, instead of it just being the product that I found and saw and chose, the landing page now could feature recommended items. And so obviously, if those recommended items are on point, then I as a customer is going to spend more time on Revolve browsing and increasing the chances of conversion as well. I'm glad you mentioned that maybe that's not as differentiated because I think when we were talking earlier about even the search function, to play devil's advocate a little bit here, I remember being at Shop Talk in 2023 and meeting a bunch of AI upstarts and they were talking about this opportunity to change search.

12:13So people would search for or brands would build on the back end that if they had a blue dress, they would name it like Cerulean meets something chic. And no one would know to search for that. So even that's not particularly innovative if you if you pull it back a little bit. I say that to ask, if you're a company that's not Revolve and you're considering whether to build versus buy AI in this hyper AI world, how would you describe what the difference is with Revolve's decision versus doing the opposite, which is just to buy something? I think Revolve has the capabilities to build. I do not think every retailer has the capability to build.

12:52Well, that's one part for sure.

12:54Cathaleen Chen: And what's the advantage of that? I think is just better integration, better control. And in the case of Revolve's search tool, it's just way more effective. But I want to make one point that I've realized doing reporting for this story, which I typically don't cover AI. This was a crash course in AI for me. But I think when you hear certain functionalities of AI being described to you, it all sort of sounds the same. So like what Revolve is doing this tool that they built with search, it's not a novel tool. Like you said, Sheena, you've heard about this at Shop Talk years ago. I think the real difference, like you and I, like the three of us, we would have difficulty picking up on.

13:45Like the real difference in how these systems work are probably just super wonky. And at this cursory level of us just sort of describing what these functions do, there's not much different, like, you know, what, I don't know, Nordstrom is doing with search. I'll bet if they were to tell me about it, it would sound to me quite similar to what Revolve is doing. Right. But I think the difference is in how they're able to test and iterate. And that's something that only comes with when you're building these things internally, with your own team of engineers. And also how your organization is not just sort of focusing on these separate functions as silos, but together.

14:31And I think that is the secret sauce of Revolve.

14:35Cathaleen Chen: That's a really great point. I wanted to switch gears and talk about an aspect of Revolve that we haven't really covered, which is their private labels. They seem to launch one of these every other day, and a lot of them are quite successful. And I'm sure AI must be playing a role in them deciding what to launch and how to market these. Absolutely. They're pretty adamant that they're not using AI to design, which I think is, you know, it makes sense that they would communicate this. AI is very much embedded in the designing process from generating very fast sketches to informing the design team, the merchandising team with company data.

15:24They recently launched this tool where it's kind of like asking Siri or asking ChatGPT, like, oh, what can you tell me about last week's sales on this collection or that collection? Anyways, private label is a huge aspect of Revolve's strategy. And the idea is that they're a data company. They know what people are searching for. They know what there is scarcity on. And so they've been able to leverage that knowledge into more than two dozen private labels across Revolve and Forward, their sister site. And it's a huge opportunity for them because one, higher margins with sort of this environment that we're in right now where there's just so many uncertainties.

16:15There's so many possible different supply chain disruptions, you know, from every angle. I think private label reduces some of the variables for retailers like Revolve. But they recently launched sort of their marquee private label. It's their namesake private label, Revolve Los Angeles. They say that it's doing quite well for them. It's definitely something unique. It's something that's totally thought out with its own design team. It's occasion wear. It's very, very dramatic. And it's expensive. It's way more expensive than most things on Revolve's website. This is sort of very much the opposite of internally how they're leveraging AI.

16:59Externally, they're doubling down on brand, which is a really interesting play. So Kat, so private label makes sense on its face how you would use AI to identify the gaps. You launch your Revolve namesake private label brand based on that. You have marketing. The creative element is obviously very complicated ethically and otherwise. Then you have physical stores. To me, that feels like a tougher test. Can you tell us how Revolve is using AI when it comes to this physical store rollout and where they are at with stores right now? So they're not. And this is a problem that they – this was, you know, my first draft.

17:38As Brian knows, I filed my first draft posing this sort of challenge that Michael and Mike were facing is that they recently ventured into physical retail. They have a very small store in Aspen and then a store in Los Angeles at The Grove. They're looking at potential new stores on the East Coast this year, but very slow rollout. This just really cemented the idea to me that Revolve is a tech company because Michael was describing how frustrating it was that they're not able to track the impact of moving a single rack or moving a single mannequin. you know, a few feet away from its original position.

18:23And I was like, well, that just sounds like merchandising, right? Like a merchant would sort of have the intuition to make those decisions and they are things that are not rooted. These decisions are not rooted in science. They're not rooted in math. And that's the beauty of, you know, visual merchandising and what, you know, people with those backgrounds are able to do in a retailer. I don't know. It sounds like if they could find a way, Kat, when I was reading that quote, it sounds like if he can find a way to do that, he's going to do it. Or both of them. Yeah. I mean, they're working on it.

18:56They're working on how they can sort of quantify a lot of these retail decisions. And it'll be really interesting to see how that plays out. I'm very excited about their retail expansion. I think how they will do it will be different from how other retailers approach it.

19:14Cathaleen Chen: Any theories? Should we speculate a bit? I mean, to get back to that question Sheena posed at the start, I mean, can an algorithm actually tell you where to move a mannequin and to sell more clothes? That doesn't feel far-fetched to me. I don't know. That doesn't feel far-fetched. And maybe even a robot can move the mannequin in the next couple of months. Like, we're not far from any of this. Yeah, it's hard to really visualize because if that data was being used to run these stores, how the stores look to us won't actually change all that much, right? It would just be every sort of minute decision behind it would be more informed.

19:51And I wonder if that's how it's going to ultimately manifest is that Revolve stores won't look totally different from retail competitors, but that they'll just be more effective.

20:06Cathaleen Chen: I think the solution here is clearly robot mannequins. Mannequins that move. But in all seriousness, I actually, I'm curious to see how they approach it because this is very similar to the pitch that Amazon made when they were opening grocery stores and convenience stores. And what they did was they loaded up these stores with cameras and sensors and all this automated stuff that customers generally hated. And it was a huge flop. But I suspect Revolve will be smarter about how they approach this and more like what you're saying. Yeah, I think so. They're taking it really slow. And there hasn't been like a huge PR push behind promoting their stores.

20:49Obviously, they're talking to me, so there's a little bit of it. But I think they're very much taking a measured approach. And I think if a new store, if a new Revolve store opens in New York this year, that'll be, we'll definitely have to go and do some reporting.

21:07Cathaleen Chen: Yeah. It's worth mentioning that we talk about them like they're a tech startup, but they've been a public company for years. They're really more in line with another online retailer or conventional retailer at this point. They're not trying to grow their sales 10X or 100X in two years or anything like that. I'm sure they'd love it if they did. I mean, they have been very vocal about the struggles of multi-brand retail in other parts of this landscape, Saks, Essence. Like, they see this as an opportunity. But one, you know, one really interesting observation just from talking to people in this space, not related to Revolve technology, but related to how they build their relationship with brands.

21:54I have heard that Revolve and Forward are amazing partners. I mean, the bar's kind of low, right? You pay on time and you're a great partner. But I've really heard across the industry great things about how they operate in terms of their third-party brands. And I think that, again, is something that we can't see as consumers, but I do think it makes a difference. Well, I think that's a nice place to end it. I love that you said that Revolve is taking it slow and steady to the tune of 29 mentions of AI on their earnings call. Really slow and steady. Not on AI, just on the other stuff. Kat, thank you so much for joining us.

22:32Thank you for having me. This was so fun. Please be sure to check out Kat's article, Why Revolve Can't Stop Talking About AI, at businessoffashion.com. This and other stories are available to BOF professional subscribers only, and you can find the links in the episode notes. You've been listening to The Debrief, produced and edited by Olivia Davies, Angel Nemov, and Eric Bria. I'm Sheena Butler-Young.

22:53Cathaleen Chen: And I'm Brian Baskin. We'll be back next week with a new episode. Thanks so much for joining us and be sure to follow us wherever you get your podcasts.