Everybody, thank you for joining. My name is Nikhil Devnani. I'm Bernstein's U.S. emerging internet analyst. I cover a variety of marketplace businesses, one of them being Wayfair. It's my pleasure today to have with me, Ryan Barney, who's the Head of Investor Relations at Wayfair, and Fiona Tan, the Chief Technology Officer at Wayfair. Thank you both so much for being here.
Thanks for having us.
Thanks for having us.
Fiona, just to give everyone a little bit of background, Fiona is Wayfair's Chief Technology Officer. She leads global innovation for the e-commerce platform. She's been with the company for about six years. Prior to that, she was the head of technology at Walmart, where she also spent a little over six years. She's sort of been on the front lines of two major e-commerce platforms, and in a role that I think is really relevant to the discussion we're gonna have today. We're very happy to have her with us. Wayfair, overall, I would say as a business for as long as I followed them, has prided itself on, you know, being very data-driven, very technology-driven.
There's about 2,500+ employees at the firm, so half the workforce centered around engineers, data scientists, product managers, and so I think a very interesting lens on the broader topic of agentic that we're gonna get into today. Before that, maybe, Fiona, can you just set the stage? How has Wayfair used data and machine learning over the years to build the e-commerce experience-
both for consumers and then also, of course, for sellers as well?
Yeah, for sure. Again, thanks for having us. Yeah, being a digitally native company, you know, I think we have been using data, machine learning, the sort of more traditional predictive ML for quite some time, right? If you think about the category that we're in, very emotive, very style-based, there aren't really any brands, I guess, that you could anchor on. The ability to understand style, represent it, ability to help our customers search, provide recommendations, et cetera. Also I would say the other big area, again, being digitally native, marketing technology, predicting optimization around, like, how we would spend.
Those are all the things I would say drove a lot of the investments in data and predictive ML over the years. Like I would say, over a decade. There was a lot of focus on that 'cause it was really critical to us in terms of how to run the business. In a way, the evolution to now the more so generative AI, right, the capabilities there, it's sort of like we had a good training, I would say, coming into it, and now being able to leverage that and scale that across a much, say, larger footprint across our e-commerce stack. The things that we were doing, we can do better, and there are new things that we can do with the capabilities, right, if you think about it.
That's been that nice kind of evolution for us. It's helped us in terms of accelerating the use of the new capabilities.
One thing I forgot to mention in my preamble, if you would like to submit questions, please do so via the QR code, and I'll weave it into the conversation as well. I have plenty to get us started. You know, one thing that Neeraj has talked about recently has been the tech re-platforming.
Mm-hmm
... that the business has gone under in recent years. What is it that Wayfair is now able to do today that you weren't able to do when you first joined five, six years ago, because of the tech re-platforming initiative?
Yeah. And interestingly, before my experience at Walmart and Wayfair, I, you know, spent the first 20-some years of my career in enterprise platforms. I think that was actually good training for what was to come with re-platforming.
If you think about, again, being a virtually native company 20+ years, you know, the platform that we had was really scaled for entrepreneurship and speed, time to market, right, of features and capabilities. What we did with re-platforming was essentially move it, right, from an on-prem solution to being on the cloud, right? From that to really focusing on sort of modern decoupled platform architectures that would enable us to move, still move quickly, but to be able to scale and perform and be reliable, et cetera, right? It also provides us those building blocks, like reusable building blocks, and it sets us up actually really nicely for you think about the API-driven and now sort of agent-ready APIs.
It is now all possible if you have the stack that we have now, right? If you say five years ago, we were building a lot of bespoke experiences, so everything that we wanted to do, we could do. We'd do it quickly, but it would all be bespoke, and then it gets to a point where it's really hard to maintain. I think the re-platforming and the timing of it was actually very fortuitous because it allows us to accelerate and take advantage of all the capabilities because we now have a good foundation to be able to do that. Testing, iterating, getting capabilities out quicker, those are all things that the re-platform really allows us to do.
Is there a cost benefit of that as well, or is it predominantly the ability to move quicker and build things faster?
I guess you get benefits from that, right? Being able to, you know, test more, bring capabilities out faster to customers, and then also the reusability of the stack. Honestly, I think our developers also, you know, appreciate the ability to leverage different components, and, you know, we're also allowing them to work with cloud-native capabilities, so it's also building up their, you know, again. It's a good learning for them as well.
You've described, I think, three strategic outcomes when it comes to AI, you know, the first being reinventing the customer journey, the second being embedding AI into internal workflows as a company, and the third being making Wayfair AI-ready or a partner in various ways. I'd like to start with sort of the last piece first around the partner and the ecosystem. I guess your primary set of partners today is always your supplier base, right?
Mm-hmm.
How are you using AI to enhance the experience for suppliers, making it easier to bring inventory online? What have been some of the benefits and product rollouts on that front?
Yeah. I think that's been an area, honestly, that again, we've been investing in for a long time. We have, you know, 20,000 suppliers. Again, with our category, the supplier information or product information is much more heterogeneous. Right? If you think about it, you don't have standard products. You know. Again, it's style, dimensions, all of those things are much more important in our category. It's something that we've been investing in for quite some time. The unlock that you get now with the frontier models, right, the how it's trained on the world's information, the multimodal capabilities have really allowed us to focus in on catalog enrichment so that our suppliers can onboard a lot faster.
They, you know, we can help them with the, again, the heterogeneity of the information they give us. We can enrich it, we can augment it, we can correct it, we can ensure that it's accurate. That's really powerful because it allows the suppliers to enable them to kind of get on our platform a lot faster. It allows us to ensure that the product information is as accurate as possible, and we can iterate with them very quickly and then bring that up to our customers, who also see the benefit of it, because I think they're, you know, being able to find the right product that and with all the most accurate information and we can matchmake the customer and the products much better.
I think that's been an area where, again, we've been investing in, but the ability to do that now at scale is, you know, not comparable to what we were doing before. In fact, we just had, you know, we work both with Gemini and OpenAI very closely, but, there was a case study OpenAI just published, I think I want to say like two weeks ago, on exactly this topic about how we've been able to work with them, to really, you know, disrupt how we do catalog. How our suppliers kind of onboard and manage their products on our platform.
Are there benefits, too, with if you're taking over? I don't want to say more control, but you're having a heavier hand in helping suppliers bring inventory online and describe things better?
Yeah.
Is that better for off platform? When I say off platform, meaning is the inventory then showing up on a Gemini, because it's sort of you're solving for what the LLMs want to see when they're scraping for inventory?
Yeah. Absolutely. I would say it in two ways, right? One is that everything that we can do on our site to make it better for our customers, again, from an accuracy, the representation of it, being able to see it in different ways, that's always really important. It helps our customers. It also does help, you know, when the platforms are say scraping our site, right, so they're able to pick up that information. I think the other thing that's been really helpful too with both Google and OpenAI is that we have integrated our feeds directly to them, right? What happens is that they're getting the most up-to-date, most detailed, accurate product information, inventory information, location, et cetera.
That enables them to be even, you know, more able to sort of represent our products when a customer, you know, when if you're asking a question to Gemini or to OpenAI, that is more commerce related or you're interested in a couch, whatever, they have the most relevant information. I would say, you know, making sure that everything that we do to make our product details better is helpful, then directly integrating with the platforms. The other thing I would say, on top of that too, we're working with them on besides the catalog feed that we're providing them, are there also additional attributes that could be particularly interesting, because of the AI native platforms and their needs being a little different than the more traditional sort of Google search kind of surface.
We're also working with them on that. It's like what are these additional attributes that they would find useful.
With respect to how consumers engage with, let's call it AI enriched listings on wayfair.com, have you measured and noticed differences in conversion rates, improvements in conversion rates with this bucket of inventory versus other inventory?
Yeah. For sure, I think there's been some measurable lift, I would say. A lot of it's just around, again, you know, if the information is accurate and our customers are able to get that, it just makes it better. Imagery is another thing that's obviously, we've been able to leverage the multimodal capabilities of the LLMs to be able to do, right? If you think about in the old days, if you know, if the supplier just gave you an image of this couch or chair, whatever.
Small couch.
Yeah, small couch. It would be, you know, you'd see it and you could see it maybe from different angles. Now what we can do is we can take that, and then we have our models that can discern, okay, what style it is and then what other products we have within our catalog that would go well with it. I can show you this couch in a, you know, in a setting, right? Before you could do it, the supplier would then have to essentially have a fairly expensive photo shoot, get all the items set up, take a picture and have that available. Now I can do it at scale.
That's something that, you know, for customers when they see that it's a lot easier for them to imagine that in their space. Then you were at Shoptalk, right? You saw, you know, we're also soon to be launching a product that takes it even one step further, right? Now you can see images of room settings in different styles and deep. The main thing there is that these are pictures of room styles with real products that are buyable, you know, shoppable on Wayfair. Then the next unlock is you can then upload a photo of your space, and then you can see those products in your space, right?
It gets it closer and closer to you being able to really visualize, not just one item in your space, but like, you know, a collection of items in your space. That's something that, you know, again, it's like a kind of at the bleeding edge of what the models can do and what we can do with the models. We're looking forward to be able to launch that soon.
When you're building these on-site experiences and you think about your positioning competitively within the market for furniture online and home goods online, do you think the moat is in the quality of the model you're using, or is it more so in the data you have to create a very niche and specific shopping experience for this category? Like, what do you think the differentiating factor ultimately becomes as you look out?
Yeah. You know, I think, you know, the models are constantly improving, so you definitely wanna leverage it as much as possible, and I think you also wanna design your architecture in a way where you can plug and play the latest and greatest and the best models. Also, sometimes you wanna optimize the performance latency 'cause you wanna make sure to do that. I think the differentiation is really around all the things that we have around it, right? The data that we have about the products and our customers, you know, our understanding, our sort of very deep understanding of style, you know, the things that we do with dimensions and accuracy, and then not to mention, you know, again, we've got 20,000 suppliers.
Some of them have their inventory, you know, housed within or fulfilled within by Wayfair. The rest are, you know, drop-shipped from the suppliers, having all of that information as well. It's the selection, it's the assortment, it's the delivery capabilities that we have. I think when you marry all of that together and you can leverage the multimodal capabilities and the world knowledge that the LLMs have, I think that's the real unlock, right? It's not just the models. The models are now enabling us to do things that we couldn't, you know, do before from a generative and sort of visualization capability standpoint. Underlying principles are still really important, right? Like I said, the data, the selection, the assortment, et cetera, and the delivery fulfillment capabilities.
One of those product rollouts recently has been the Discover tab.
Yeah.
Telling Ryan I'm moving apartments in three days, so I'm in the furniture market again. You know, I've sort of been going through it myself personally. You know, what have you seen in terms of consumer response and engagement around that as a sort of launching pad to conversion rates and overall interaction with customers?
Yeah. I think customers that are engaging with the Discover tab, you know, they are pushing conversion lifts on that. I think a lot of it's around, again, if we can provide an experience that's helpful, right, and that you haven't, you weren't able to do before, right? Hopefully if you've been playing around with the Discover tab, and then soon, hopefully with Stylist, that's the thing it's gonna allow you to do, is that you now are able to visualize. Again, it's a category where it's super hard. Like, if I had this couch, I actually don't know how I would describe the style of this couch or chair. I don't even know if it's a couch or chair.
It's if you can see it or if you can just take a picture of it, right, and our models can understand that, and then it can recommend things to go with it, et cetera, and then it can show it to you. I think just the trust loop around that. Again, it's all based on accurate information. It's based on products you can buy. I think that's pretty powerful. That's where we're leaning and we are seeing that lift in conversion with customers who are engaging in that way.
How do you think about on-site search on Wayfair evolving over the long term?
Yeah.
I think the platform is now able to ingest more conversational.
Sure
... queries. Do you think that becomes the default of how consumers search, and is that a good thing 'cause you get more context? How do you see just search changing and what it means?
Yeah. I mean, if you think about, you know, honestly, e-commerce in the last 20 years, it's been pretty similar, right, in terms of how you find things. It's been a search bar and some filters. I again think, you know, in our category, we've been probably even more disadvantaged by those limitations because it's such a visual category and again hard to describe in, like, succinct keywords, right? I think about it more as not just that you can type in longer queries. It's really a combination of the ability to have a multi-turn conversation, but it's not just typing, right? It's how do you incorporate voice into it? 'Cause sometimes you're trying to describe something, it's a little easier to talk it through rather than actually typing it in.
I think the visual aspects obviously, right, with the generated imagery. Then the context that can follow you as you're shopping, understanding that, hey, you know, you're coming in this time and it's you're remodeling your bathroom. What would be helpful to be able to surface to you across pages? I do feel like it's not just gonna be a longer search bar, and that's kinda why what we're working on is how do I come up with something where it is gonna be more multi-turn. You can do more things with it. It's more iterative, and it can have the context across pages, and it can incorporate sort of the imagery, video, voice very seamlessly.
That's something that actually we're working on, very excited about, 'cause I think now we have the tools to really be able to disrupt the shopping experience, which again, I think particularly for our category, it's gonna be a game changer.
That being the Wayfair Stylist search experience?
Wayfair Stylist is just one part of it. I would say that, you know, the ambition is to infuse it across the commerce journey, right? Not just where, you know, where you're looking at, I wanna reimagine my room, but, you know, sometimes it's just you wanna just replace one piece of it, right? Or you wanna do something that's more like a remodel. We wanna be able to help you through all of that.
When it comes to what you've seen and learned from the initial integrations with the larger generalized sort of LLMs out there, what kind of behavior are you seeing from users as they come through from a Gemini or a ChatGPT? You know, what's been the experience overall around how their behavior might be the same or different to what you might otherwise get in terms of direct traffic to Wayfair?
Yeah. You know, I think it's still early days, you know. I would say it's a channel, right, it's still nascent, but it's growing quickly, right, I would say. It converts well because again, you know, somebody on that platform has done a lot of research. They're using the platform to research and been recommended a product from the platform that is on Wayfair. When you click through to Wayfair, you know, it's fairly high intent, right? It converts well. Now it's like how do we work with those platforms and we were sort of founding partner with Google on their UCP, which is a very broad-based protocol for commerce.
Then also with OpenAI on what they're trying to do, and you all probably know that, you know, they pivot a little bit away from sort of the instant checkout experience, but they're really focusing on discovery. Again, working with them to make sure that, you know, our products are discovered well and represented well. In OpenAI's case is how do you build out the ChatGPT app experience so that you can also help drive some of that.
An audience question here, whether you could quantify the percentage of traffic, to the extent you're willing to, throw that in there. Then if I could add onto that maybe qualitatively, is it incremental traffic to Wayfair?
Well, I appreciate the attempt, but we're not gonna quantify it. It's still very small. To Fiona's point.
Yeah
It's very nascent today. It's growing quickly, but, you know, that's the small number growing to a slightly larger small number.
Okay.
Yeah. I think we'll learn over time, you know. The main thing is we wanna be where our customers are at, right? We wanna make sure that if for customers that do wanna start on a native AI platform and they wanna do their research there, we want to show up well there. We wanna be able to help them find us there.
With the integration with Google around UCP, sort of what was the uplift to be part of that rollout? What went into building that? To the extent you've sort of looked at ACP from OpenAI, like how do they differ?
Mm-hmm.
What goes into all of that?
Yeah. We were, like I said, founding partner with Google on UCP. This is again where the replatforming really helps, right? 'Cause we now have a more modular stack, and we can actually work with them on sort of like what are the APIs and what are protocols. The initial focus was around you know, obviously integrating the feed, having them be able to discover us, and then being able to check out, so for a customer to be able to check out. The protocol is a lot broader, so it allows us now, you know, recently they, the release was around like how can you then do more on the pre-purchase side?
How do you also integrate in things like, you know, loyalty or promotions, et cetera, right? We're working with them, and enabling that. You know, I think again, it's very broad-based. You'll see over time they'll continue to evolve and more and more capabilities will be added to that protocol. It was important for us to be part of that initial group. You know, again, we've had a very long and productive relationship with Google. I mean, you know, when I say we moved from on-prem to cloud, you know, we're on Google Cloud. That's been good to be able to do that. Then on the OpenAI side, similar, you know.
The protocols are different but also pretty similar. As I was saying, they're really focused on the discovery part now. Initially, you know, they were also very focused on the instant checkout. They've now moved to discovery, so we're working with them again to just make sure that our products are discoverable. To the original questions around replatforming, it's really set us up nicely to be able to do that.
You've touched on it a little bit, but if you could elaborate on it a bit more, you know, what are these LLMs looking for, in terms of, you know, when you think about like GEO being the new SEO, right? So what attributes do you think matter the most such that the Wayfair product shows up better than the Amazon product or the Walmart product or any other competitor you might have?
Yeah. A lot of it is very similar to what was before, right? Whatever is available that is that can be accurate, that is very complete, that will help our customers find our product and get the information that they want, that's most accurate. A lot of the hardcore work that we've done around the catalog has been, you know, which we have been doing for a long time, is continues to be important, so that first-party data. Then for us, the things that are maybe more nuanced and more specific to our category is like our understanding of style, right? Is a big one just because it is style and trends, those are all very important in our category.
The accuracy around dimensions and you know all of that is also. Just specific nuanced knowledge about the category is important. That's all showing up. As I mentioned, I think you know working with OpenAI and Google, I mean, they're also looking at like what are the additional things that they might find to be helpful in a conversation. They're also learning 'cause a lot of it is driven by what customers are asking and how they're interacting with them. As they are understanding that, you know, we're also, they're also letting us know you know what could be additional attributes that we should pass in, right? Versus just them being able to gain it from our site.
A lot of this is still designed with the human in the loop.
Mm-hmm
Right? Making the final decision, assisting with the search, discovery, and transaction. Certainly every e-commerce or in general, every website or app is still built around like optimizing for that human experience.
Yeah.
Do you sort of fast forward and take a longer term view on agentic, where things start to become more automated? What do you envision sort of that to look like for wayfair.com or whatever? Like what needs to be built to enable a more agentic search, discovery, and checkout flow that doesn't exist today?
It's interesting because again, the building blocks need to be there. A lot of it's around how is your e-commerce journey architected, right? Your tech stack architected so that you can expose out APIs for different parts of the e-commerce journey. How do you make sure that, you know, once that's available, authentication, authorization, but now you have the extra layer of if it's going to be agent ready, what does that mean, right? Some of that's around, again, just making sure that, you know, APIs are available, but then also, you know, making sure they're idempotent. That you need to make sure that now if it's gonna be an agent that's gonna be operating against it, you know, what does that mean, right?
I think that's gonna be the evolution of it, and I think there's something around like making sure that your APIs are, say, agent ready for an external platform to be able to sort of represent it on an external platform. Then also, you know, we need to, you know, one of the things that we're also really focused on is how do I make sure that, you know, I want customers to still come to Wayfair, and I can also provide a you know, agentic shopping experience on our platform. Again, go back to how do I provide that assisted shopping experience or, you know, over time, a more autonomous shopping experience. I think you need the building blocks, right? Where you're still confident.
I don't know, I think maybe there was probably a thought about a year ago when OpenAI first talked about taking on commerce as a platform. I think the initial maybe thesis was that it was people don't want to shop anymore, and it's gonna be all automated and that was gonna be the thing. You know, you've seen both companies really step back from that and just say, "Hey, look, we still wanna be the merchant still important." You know, at NRF earlier this year, you had Google and Sundar Pichai talking about how, you know, launching UCP but still being very focused on, hey, look, you're still the merchant of record. Retailers are still the merchant of record.
There's still a lot of fulfillment, post-purchase, even transaction information that needs to be there. I think that, you know, we wanna be there, we wanna evolve as customers become more, you know, comfortable as their shopping journeys evolve. I think we have got the building blocks. More will have to be done because I think the bar for an agent to be able to take actions is higher, right, than a human being able to take action. We wanna make sure that we are building towards that.
Every category is a little bit different as well, right?
Mm-hmm.
Home goods related stuff tends to be more visual.
Mm-hmm.
Maybe a bit more emotional in nature when you're buying something.
Yeah.
I guess maybe, you know, what data have you seen in your consumer base in terms of anything you could share on the extent of sort of browsing and visual discovery that goes on before someone actually makes a decision, which then, you know, needs them to be involved actively in a transaction?
Yeah, I mean, you know, I think again, it's for our category just first of all, it's high consideration, right? It's visual. I think also because it's home, it's very emotive, it's very personal. I think, you know, again, for our category especially, I think we still imagine that humans would wanna still be involved, right? I mean, it's something that you're building a nursery for your baby. You know, it's just hard to imagine that this will be completely agent enabled away. But we wanna make sure that it's really easy and pleasurable, right?
That's the other thing is I think, you know, I don't know, I think there's a lot of us that still feel like I think shopping should be fun. It should be delightful. We wanna make sure that we are allowing you to do that and being able to do it differently and feel more empowered, right, when you shop on Wayfair.
Ryan, maybe a question I'll direct towards you. You know, one of the things everyone in this room is thinking about is sort of what happens to traffic between direct traffic and traffic through these other avenues. You know, one of the concerns for marketplace businesses at large is that you just start to see an erosion of direct traffic, is sort of the fear. When you put your strategy hat on, when you listen to sort of what management has been talking about, like how does Wayfair think about and assess the risk reward of being there on all these agentic sort of storefronts?
Right.
Making sure that you retain your direct traffic.
Yeah. It's something that, you know, we wanna be very thoughtful about, and obviously you've seen some retailers take different approaches where they say, "Hey, we don't wanna be involved with the big LLMs. We're gonna just own it all ourselves." You know, our view is that we wanna be present there and make sure that we show up well, just in the same way that we have on, you know, Google Search for 20 years. But we also wanna make sure that the experience that shoppers have on Wayfair remains differentiated and quite unique and very much built for this category. And obviously, that has always been kind of the through line of our operating and will continue.
I think that spread will continue to widen given a lot of those stuff we're working on, much of which Fiona talked about, which we think will really help personalize the shopping experience in the home category for each unique shopper and functionally help drive curation, right? It's a large catalog, you know, over 40 million products that we have to offer, which for a lot of folks can be somewhat intimidating to navigate. We think a lot of these tools will really help find the item that's exactly right for you that much more quickly. That's an experience you're only gonna really be able to have on a platform where we're custom building that shopping journey for the home category that, you know, the customer is looking forward to.
That's where we think that, you know, the offering that we're gonna have on our surfaces is gonna remain distinct and quite attractive in that landscape to both, you know, retain the customer as they think about their place on the LLM versus on the dedicated surface, but also as they think about navigating other pure play retailers in the space and why they come to Wayfair versus a competitor, you know, that also focuses on the home but doesn't have the scale and sophistication to be able to offer that same, you know, highly personalized shopping journey that we can bring.
I think one of the things that has been talked about is AI helping with advertising off platform or improving the quality of ads that are being run on Google.
Yeah.
Our social channels. Can you talk a little bit about where you're seeing benefits and improvements on that dimension?
Yeah, I mean, you know, Fiona, please chime in too. I think that one of the earliest areas that we've been able to leverage some of the generative technology has been on creative. You know, just in terms of being able to build ad formats and develop ads, especially kinda short form video, these types of things where you can iterate much more quickly and do things that, you know, would just not be cost-effective from a production standpoint to try to replicate in real life. That resonates well with shoppers. That's, like, one piece of it. There's other areas where obviously the technology can help optimize the actual bidding and auctioning, you know, as part of the process as you think about the more traditional ad platforms.
It is an area where we have driven, you know, a very high degree of sophistication for a long time, right? We have kind of been on the bleeding edge of ad technology as we think about integrating with, you know, the major ad platforms that have been out there for some time. You know, this is gonna be an area where we see ourselves as always kind of one of the leaders of the pack, so to speak. I don't know, Fiona, if there's anything you wanna add to that.
Yeah, no. I mean, 100%, right? That continues to be more sort of predictive ML focused. Being able to now, our thing has always been, like, we wanna be able to run experiments. The other is now that we have the generative capabilities, we can explore, exploit much more quickly, right? You can figure out, hey, look, which content actually resonates best, right? It allows us to iterate a lot faster. It's sort of in conjunction with a very completely sophisticated martech stack, right.
Right
... we've built on our own. A lot of it was a necessity. Again, being digitally native, that was an area of focus and really an area of research and development for us, from the very early days.
I think one aspect of e-comm on the consumer experience is always sort of returns when things go wrong.
Yeah
Customer service. Anything you can quantify there on degrees of improvement, that these tools are helping you with?
Yeah. I mean, I think certainly, you know, again back to the, you know, in as much as we can make sure that the product descriptions are accurate and we can help even customers visualize that, you know, this couch is not gonna fit in your room, et cetera, I mean, those all help with the returns, you know, sort of the post-purchase experience. Then we've also, you know, obviously leveraged GenAI on the customer service side. A lot of it's around, like, you know, we wanna still make sure that customer service is something that we pride ourselves very, very much so on. I think it's the company was built on that as a very, very important thesis. So we wanna make sure that we meet customers where they are, where they're at.
With the, you know, again, with the advancements and the capabilities, you know, we can handle a lot more customer interactions autonomously, where customers actually appreciate that is actually the medium or the channel how they wanna have their issues handled. What we do is we always wanna make sure that we're, you know, looking at containment, but then also customer satisfaction and how quickly we can help service the customers. Then there's another host of customer service calls where a super powered human assisting them is still absolutely still the way to go. We don't wanna ever feel like or have our customers feel like that is something that we're not going to offer them.
That's always gonna be something we're gonna lean into. We're just gonna make our customer service agents that much more capable of handling their more complicated requests. We are, again, in our category, we're gonna have much more complicated, I think, customer service interactions than the typical, "Hey, it doesn't fit. You know, I need a different size or different color," right? Ours are a lot more complicated.
Sticking with the theme of e-expenses and maybe switching gears a little bit to productivity, how is software engineering changing now with-
Oh, yeah. Goodness gracious. It is changing quite a lot, and I would say in the last six months probably even more so. You know, we started having the coding agents, right? So it was really around, like, how do you, as an engineer, it was mostly focused around engineering, on software engineering, and you would have sort of a co-pilot with you and, you know, it would generate code with you and you became more productive, but you were still sort of the driver from that perspective in terms of the code, right? I think the recent advancements, two things. One is that it's now at a point where it can think about the software development life cycle as a workflow, you know, engineering or you know, the coding part being one of it.
It's now, I would say, most of our software engineers are now moving towards not still being in the flow of coding, but more in the flow of orchestrating, right? Then also, which is, you know, I, you know, having coded before, certainly I don't do it as much now, it is a shift, right, from being sort of, like, involved in the actual generation of code with some help to I'm not really gonna generate much code at all, maybe not at all, right? Just be able to set up the system architecture, set up the instructions, provide the context, and then have these worker agents go do stuff, right? At the end of it, they're still making sure, yes, it's good, and actually deploying it.
If you take that one step further, it's also, like, our product managers, our designers also being able to leverage the capabilities. 'Cause if coding just gets faster, but the other you know, the other functions are not able to leverage it, then you still have a bottleneck. 'Cause at the end of the day, you wanna make sure it's measured, it's not just how many lines of code that you wrote. It's really measured around, like, what is the new feature capacity and what is the output of capabilities, right? We wanna make sure that we are able to do that much more holistically. It is changing. It's
I would say, you know, there's a good amount of change management involved in that process, but also, you know, there's a lot to be gained from it.
As productivity goes up, I'm sure there are other departments at the company that are also seeing benefits on how people work. You know, you start to generate savings, you start to generate sort of better output. Like, how do you think about as a company managing that sort of trade-off? Do you think about these savings, you know, flowing through to the bottom line? Do they become source of more aggressive reinvestment to grow the business quicker? What is the Wayfair philosophy on the productivity boost and the savings and how you think about allocating those dollars?
Yeah. Well, obviously, you know, we have for years executed a stringent regimen of cost efficiency. Obviously, you know, we've taken our headcount and our fixed cost base down for, you know, multiple consecutive years here to get us to a really strong place where we have a team today that can drive the business as it sits today, but also drive a much larger business. You know, the way we generally think about it is that as this business, which, you know, we did roughly $12 billion of sales in 2025, you know, grows from $12 billion to $13 billion and beyond, we don't need to expand the team to be able to facilitate that. We have the folks in place today that are driving the initiatives that will get us there.
Really, it gives you that much leverage, that much more leverage on the fixed cost base. Certainly with the, you know, with the implementation of a lot of these tools, each human is that much more productive. The bar to then add an additional human is that much higher. We can keep that fixed cost base very much static. As the top line grows and the business gets back to a traditional cadence of growth, you're gonna just see that much more powerful flow-through on the profitability side from it.
As you think about maybe just to push on that a little bit, so we've talked about revenue initiatives that are enhanced by AI tools. We've talked about different cost departments, whether that's advertising or headcount, that get some help and benefit.
Mm-hmm.
If you're an investor looking at this P&L over the next one to three years, where should you sort of mentally think about the benefits showing up? Is it more so top line? Is it more so through the expense base? How would you characterize that?
I would characterize it mostly on the top line. These initiatives are gonna help us accelerate our share spread. Obviously, the unknowable is a category that remains somewhat volatile today, but will help us accelerate our share spread, grow that much more quickly. The benefit then will be that as that growth comes in, the flow-through on that is quite attractive because you're holding your fixed cost base more or less static. That's been the operating model that we executed against in 2025 quite successfully. That's the model you'll see us execute against here this year and then going forward.
I think one of the main sort of messages or comments that came out of the last earnings calls was this idea of share spread
Right
widening and continuing to widen as the year goes on. We talked about these initiatives that contribute. As you're sitting here today and you're looking at the balance of the year, Ryan, how are you thinking about additional factors that are contributing to that and your just overall view on kind of share spread as we progress through 2026?
Yeah, I mean, obviously we had a number of initiatives and to Fiona's point, many of these were enabled by the completion of our re-platforming work to really be able to dedicate the bulk of our, you know, engineering and technology team to this notion of product-led growth that played to our success and to our benefit in 2025. We talked at length about things like the loyalty program.
Programs like Wayfair Verified. You know, a lot of the work we did on the marketing side when it came to influencer-based marketing and the like. You know, all of the initiatives that drove that success in 2025 continue to compound and many at an accelerating rate here in 2026. We see that as a nice driver of share spread widening. We're gonna layer on top of it all the things we're doing in the customer experience, many based around agentic technology, generative technology is probably a better way to frame it, that, you know, we think again will differentiate the shopping experience folks have on Wayfair services versus what they can find elsewhere.
We see all of that helping us continue to drive just a wider and wider share spread versus the category wherever the category may go. That's really where, you know, we want folks to focus is that, you know, at the end of the day, we are in the driver's seat around a lot of this. We haven't been able to be in that place for some time because of the kind of cost of the re-platforming work, the time and attention that that took. This really harkens back to a Wayfair of old, where we are really focused on driving improvement in the customer and the supplier experience in a way that just continues to win both the customer and the supplier order day in and day out.
Great. With that, we are right on time, so we'll leave it there. Thank you both so much for joining.
Thank you.
Thank you for having us.