S3E06: Scot Wingo - The Evolution of Ecommerce

Welcome to Commerce Famous, a podcast that celebrates the celebrities of our industry. Join us for a conversation with the tastemakers in ecommerce. In their daily lives, these are just normal everyday people like you and me. But to us, they are commerce famous. Alright.

Welcome to episode six of Commerce Famous. I have a really exciting guest, for you this week. I say it all the time, but this time I actually mean it. It is Scott Wingo. Scott, welcome to the pod.

Hey, Jason. Great to be on Commerce Famous. So the, the format's pretty loose. We're basically just gonna have a conversation about you and your history. And, frankly, I know you're Commerce Famous because I've been listening to you and your pod, the Jason and Scott Show, for many, many years.

So couldn't be happier to have you on. I've been a little bit of a fanboy, as it relates to some of the other work you do. Thanks. Thanks. It's always fun to talk to what's fun about ecommerce is we've always always have new people coming in, but then there's the OGs that have been around for a while.

A long, long time. Unfortunately, longer than I would care to admit. But, anyway, why don't you introduce yourself to our audience, a little bit about kinda your background, and and we'll bust into some of those experiences you've had. Yeah. So I'm Scott Wingo.

I am currently CEO of Refibuy, refibuy,um, .ai. And, this is I'm a entrepreneur in the Research Triangle Park area of North Carolina. So this is my fifth company I've started, so a serial founder. And, the thing I'm most well known for well, a couple things in ecommerce. I started a company in 2001 called ChannelAdvisor, and what we did there is we helped brands and retailers sell on marketplaces.

We went public in 2013, and then ultimately in 2022, we were taken private by private equity and merged with Commerce Hub. And then they re I was no longer involved at that point because when you're taken private, they dissolve the board and it's owned by the private equity firm. And then, they rebranded as Rhythm, r I t h u m, and I'm, not responsible for that rebrand. I will typically refer to it as ChannelAdvisor. So, so, and then as you mentioned, I have two podcasts, that I do, in our world.

So the probably the one of the most long running retail ecommerce oriented podcast is the Jason and Scott Show. And, I cohost that with Jason Goldberg. He is the chief digital officer and commerce digital grand poobah of Publicis, one of the biggest, ad agencies, in the world. You know, it's like the big five accounting firms are now the big two or something, and same same in the in the, ad world. So the biggest agencies, you've got Publicis and then, like, one or two others.

It used to be, like, six big ones, and then they've they've kind of slowly but surely coalesced into one or two big firms. So he works at one of those, and he does all the ecommerce things there. So it's kinda fun because he's kind of a store grocery person and payments, and I'm a marketplace ecommerce person. So we end up having a lot of debates about that that end up being, what the audience likes. So, yep, they like they kind of it's a point counterpoint format for ecommerce, if you will, which sounds strange, but it works.

It's not politics. It's like stores are gonna I'm the stores are gonna die, and he's like, you're crazy. So yeah. It's a great it's a great format. And I'll be honest, it is the first podcast that I ever downloaded on my iPhone years ago, and it kinda got me into podcasting and and and subscribing.

And so I think you were ahead of your time. Yeah. Thanks. We're a little, we still do it. We did do it weekly, and then, Jason has become you know, he is as his as his agency has grown, so have his responsibilities, and he he is always on a plane somewhere, which makes it hard to podcast.

So we're kinda down to monthly, quarterly ish right now, but that just makes them that much better because the audience, is constantly anticipating the next one. Commerce famous is proudly presented by Shopware, the leading open source e commerce platform for mid market and lower enterprise merchants. More than 50,000 clients already process over $25,000,000,000 in annual GMV through Shopware. Find out more about Shopware and the best value in ecommerce at shopware.com. Well, so going back to the the channel adviser days, by the way, I I still call it channel adviser just like I call it the Sears Tower.

Some things, you know, it's hard to rebrand some of those things. But but you've had I mean, that was an amazing company back in the day, that you founded and started. What was the idea on why you started ChannelAdvisor? What was the what was the market need that wasn't being met? Yeah.

So so it's kind of a longest story, so I'll try to do the abbreviated version. But, so I had started, I'm a big Star Wars enthusiast. You can't tell it's blurring it, but that's like a Star Wars poster. And I'm I have lots of Star Wars stuff. So I I love Star Wars, and I also collect comic book art.

So I am a I'm a collector as well. If it's nerdy, I'm there. And, so early earliest days of Internet, the way we bought things online was auction sites. Auction sites were more prevalent than fixed price sites. So I started a company based on my own needs.

If I was trying to buy a collectible, I realized there's all these auction sites, and I would have to track inventory across them and bid across them and have these complex strategies. And I thought it would be neat to build a bot or an agent, for that. So in 1999, I started a company called auctionrover.com. And what we did is we could search all the auction sites. There's a bidding agent.

And then the third thing we did is we helped you sell across those auction sites. We got acquired after about six months by and you'll remember this. It used to be called goto.com, but they got in a dispute with Disney. Disney had go.com, and they got in a legal battle, and goto.com had to change their name to Overture. These are the guys that invented paid search, and they acquired us because that's how we were monetizing our auction searches.

We would get the auction sites to pay for for for clicks. And, I was very enamored by this new idea of this thing called paid search and pay per click. I thought that would be that could be kinda big. Turns out, that was right, but the people that made money were Google and not Overture, sadly. So a lot of lessons to be learned in the business world there.

The usually, the many times the inventor, of something is not the one that figures out the better mousetrap. But anyway, so they bought us, and then, like, literally two months after they bought us, the .com bubble burst, and and you you remember that. It was kind of scorched earth, and all the auction sites went out of business except eBay. They made it through. And, fun fact, our our the blue in our color is eBay blue as an homage to the OG, eBay.

And refi buy actually comes from a conversation I had with Pierre Omidar, at a cocktail party one time. So, so it's just kind of a, you know, been in ecommerce for a while. And so when you only have one auction site, you would no longer need a search engine because the eBay search engine is there. And and the bidding the bidding agent doesn't really it worked, but but it wasn't as applicable. But our selling tools became wildly popular on eBay, and this is when the advent of something called a power seller started happening.

So these little eBay mom and pops turned into bigger and bigger sellers. And then then one day, I was looking at a report, and our number one seller on Auction Rover's selling software was, a company called IBM. And they had sold this is back when they owned the ThinkPad brand before they sold it to Lenovo. Oh, yeah. And they sold $2,000,000 worth of computers on eBay.

Well, the eBay computer category was, like, 1,000,000 a month. So so we basically tripled the eBay computer category with one one customer that we didn't even know. They just found our software and started using it. Okay. So then I said, that's crazy.

What if companies view these auction sites as a channel for for selling things? And, I was thinking, well, the they did it to to liquidate because they were just swimming in laptops due to the recession caused by the dot com bubble burst and and all this. And then shortly thereafter, we had Cisco. We got Motorola and Nokia. Once we figured out this use case, we were able to go pitch this idea of instead of liquidating these things to liquidators for 5¢ on the dollar, sell them yourself directly and and get more money.

Almost at the exact same time this started to hit, the CEO of Overture called and said, we're gonna shut down your business because we need that capital we're investing to go buy all these awesome search engines like Dogpile and, AltaVista and whatnot because we're gonna we're gonna own the market and crush these two Stanford guys that started this math sounding company called Google something. I was like, okay. That sounds like a great idea. But can we spin that out? So we're that's how ChannelAdvisor was born from this idea that brands and retailers would sell on we knew they wouldn't be called auction sites because at that point, eBay was starting to pivot and to buy it now, in reaction to Amazon's ascendancy.

And we thought, alright. At some sometime, this is the best format, this marketplace format. There's gonna be a lot of these eventually, and people just like selling across auction sites, people are gonna want to sell across these marketplaces, and we have technology and know how to do that. So that was, like, the kernel of the idea that became ChannelAdvisor. But then what was scary is that was 2001, and we didn't get our second marketplace until 2007.

Now the good news is that one, was Amazon. So if you're gonna if you're gonna get another marketplace, in hindsight, having the Amazon marketplace be the number two, it was worth the wait. But it was a little it was one of those things that, you know, as a founder, sometimes you feel like Don Quixote tilting at windmill. So I was like, you know, I was tilting at the eBay windmill saying or the marketplace windmill saying, there's gonna be more. I know someday this is such a good format, and, you know, eventually, that came true.

So and then then it was, off to the races. So so if you look today, fast forward to today, you know, I don't even know at, you know, during my last days at Challenger, I think we supported 250 marketplaces. And I would say today, that's probably doubled. So you have companies like Miracle and, you know, now there's, like, six companies that look a lot like Miracle, m I r a k l, that will help you add a marketplace. And what's really fueled that is the success of retail media networks.

So that that is this kinda, like, ingredient that really was the accelerant, to make every retailer and many times brands want a third party marketplace. So that has, you know, it's it's easy to make predictions that become right. It's hard to do it in a period of time. So so I was right, but it took twenty four years to kinda, like, get to, you know, my prediction. But, you know, fortunately, I was able to have a company that rode that wave eventually and and did well with that.

Yeah. That is an amazing story. Lots to unpack there. First of all, I think there's a lot of really powerful alumni that came out of ChannelAdvisor and the lessons that you guys learned. I mean, you got a front row seat to the birthplace of marketplace globally.

And and fact frankly, we're we're a catalyst in it. Who are some of those people that that kinda came out of those experiences? Rick Watson, a friend of the pod, is one that comes to mind. Rick Watson, he, was a product manager early on. Dave Spitz, who was our CEO, now works at Vista Equity.

So he kind of went to the dark side, and he's kind of an ecommerce kind of, person there. Michael Jones was, he went to eBay after us, and then he ran some other things, and now he runs a retailer. It's kind of more it's b to b and b to c called CandleScience. Ryan Walsh was our CRO. He has a start up here in town called Repview.

There's yeah. Oh, yeah. So and then, I've tried to hire, Link Walls, our chief product officer. He works for me now. And then our one of our top salespeople, who ultimately became the CRO after Ryan, Derek Conlon works for me now.

And then yeah. So yeah. And then our some of our engineers went to work for MicMac. They bought we had a shoppable media like product and a where to buy widget that MikMak bought. Again, that was after my time, but a lot of them are over there now.

So, yeah, ChannelAdvisor has spread through the the the, ecosystem. A lot of them ended up going to Salesforce to work for their commerce cloud group. Yeah. So they're they're all over the place. That's what I love about some of these gray origin stories is that, you know, you get a front row seat to things that went well, but also things that didn't, and these people end up being kinda outsized impact in other organizations.

You know, I I don't, I don't know some of these people, but it looks like you did bring a handful of people from channel advisor over to Refi buy, and that probably accelerated a lot of your your speed of trust, essentially, with these employees. Yeah. There's people that aren't in ecommerce underestimate the complexity because it looks so easy on the surface. Right? Because we present these easy user interfaces to people, it's like, oh, one click shop, you know, one click shopping.

How hard could ecommerce be? But ecommerce is deceptively complicated and, you know, the we we all know that because we're in the industry, but there's, you know, for for every 10% you make it easier for the consumer, there's, like, a 100% more effort and stuff that can go wrong on the back end. So so it's very hard for a new person. It's not impossible, but the learning the learning curve is a mountain, and it it is getting bigger and bigger over time. So what we do, I like to do things that are kind of into the the plumbing, you know, area of ecommerce.

They they get into, like, the product catalog and integrating with platforms like y'alls and, this kind of thing. And when you get into that layer, stuff gets complicated real quick and you have to have, you know, what I would call really deep domain expertise or subject matter expertise. And it is a again, it like is a way to go way faster to be able to hire people and bring them onto the team that that, you know, we don't have to under explain what's a variation, what's a, okay. When this transaction is happening, what are some things that can go wrong? How do we how do we make sure that doesn't happy happen?

Think things of this nature. Yep. Well, so you said you're a serial entrepreneur. You've got a bunch of successful startups, some exits. One of those, which, you talked about a lot on the Jason and Scott show, was get spiffy.

You wanna kinda tell people what that is, why you started it, and, you know, obviously, you're no longer involved, but maybe kinda talk a little bit about what that business is. Yeah. The the, you know, the idea there was and and I'd say a lot of my ideas so as an entrepreneur, I've tried to, like, go up the the the TAM ladder, the total addressable market ladder. My my first company was this little developer tool idea for Visual c plus plus developers. So the the TAM was probably, you know, maybe 20,000,000 or 30,000,000, which sounds like a lot, but a VC would never fund a business like that.

And in fact, we couldn't raise VC because it was too big of a too small of a TAM. So so but that was just kinda like my first business, and I I have an engineering degree, so I didn't even know what any of this stuff was. I've had to kinda learn it all on the job as it were. And, so I've tried to, like and and then one of the, you know, I'm a big study of all these business case studies, and one of my favorite CEOs, founders is Jeff Bezos as as you probably aren't surprised to hear. And, there's this one interview with Bezos where someone said, when he'd started Amazon in, like, you know, '95, '97, they said, what is the trend you're you know, what's some new trend you're betting on?

And he's he always said, I'm gonna bet on trends that aren't gonna change. And if you go read his, like, first shareholder letter, and it's 1997, it it is appropriate today, you know, twenty eight years later. It's pretty crazy. And it's because he's he said people won't get tired of low prices, selection, and fast free shipping. He He didn't even know to call it Prime.

Prime wasn't even a glimmer in his eye yet, but you can see the the the the the string he pulled to get to Prime. So so I've always thought, you know, what is something that is gonna be a decade long thing that you can kinda ride that and just and I like this idea of these these waves of disruption sweep through, and that creates this vacuum for startups to fill. Okay. So that's the theme. So after starting ChannelAdvisor in 2014, we were on the roadshow to go public, and I had my first Uber experience.

The guy that run ran, this is a funny ChannelAdvisor connection. We had this customer named Josh, and, what was fun about ChannelAdvisor is we worked with the Nikes and Underarmers of the world, kind of the intrapreneurs and the chief you know, what are now the chief digital officers, but they were like the, you know, I'm the ecommerce rogue person inside these orgs in that time frame. And but then we also worked with all these random entrepreneurs that were just starting businesses. And, this guy had he was selling n c double a licensed stuff on eBay and Amazon, and he was doing, like, he was in he was in high school, I think, or just going to college, and he was doing, like, 10,000,000 a year and probably netting two or three. You know, he was just, like, tonning it as, like, a 17 year old.

And, those are my favorite. Like, I love those stories in ecommerce. And I I got to know him really well. His name was Josh. And I was like, he's like, I'm closing down.

He called me one day and was like, I need to terminate. I'm closing down my business. I was like, oh my god. What happened? And he said, oh, nothing happened.

I I got this job offer to run New York for this company called UberTaxi. And I was like, I've heard of these guys. This is the worst. Don't do it. You're making us too many here.

Yeah. And, he did not, listen to me, and he was, like, employee number six at Uber. So he's probably, like, a bazillionaire at this point. But he got me into the beta. So I was on the beta of Uber when I was in New York at the at, you know, at at this Goldman Sachs roadshow thing.

And once I had my first Uber experience, that at that part, it was just black cars. But I was like, oh my gosh. We've seen and if you look at big picture, if you look at GDP, you know, 2020% of what we do is is consumer goods and but then, like, 40 or 60% of what we do is consumer services. We consume way more services than as a economy than we do products. So I was thinking, what if these services go digital?

What's that look like? What kinda looks like the Uber experience on the front end? It doesn't have to be the back end Uber experience. So I thought, how can what's an area that no one's really thinking about doing this that I could get into? And I'd previously invested in some car washes, so so I thought we could start with car wash and then add in other car care things because no one loves having their car dealt with.

If you live in a big city like New York or something, you may not even have a car. But most people, you know, are are not in the big, you know, like, 30% of people are in big cities and have that, but, like, most people aren't. So, you know, where I am in North Carolina, we we're in our cars, like, you know, four or five times a day. We live in our cars and drive everywhere, and it's a total drag to have your car done. So that was the idea of Spiffy was, let's let's go, create a digital experience for that service of how you get your car taken care of.

So we had an app. We had a consumer piece, and then, you know, the fun the fun thing about these start ups is you really know, where they're gonna go. So, you know, we launched this, and what we quickly realized is, that, you know, the, the big opportunity was fleets. So all these big rental car companies started calling us in, and, it's kinda funny. Amazon ended up being one of our biggest customers because, you know, they have this DSP program, and DSPs it's interesting.

They really value, the van uptime because an Amazon van delivers 500 to $1,000 of product a day. And if it's down, it's very hard for them to fill that capacity because they run at almost, like, 99% capacity. So so they will pay a lot for, the vans to keep rolling. So so for example, at at Spiffy, we do midnight oil changes and tire changes and things. We we kinda work a third shift there for them, but they're willing to pay very handsomely for that.

So that that was kind of the Spiffy story. So grew that from zero to 30 stories, 30 cities, and then, you know, in the range of $6,070,000,000, ARR. Incredible. Incredible. What what I love about the story, number one, is that you saw the opportunity.

You found a parallel use case around Uber and how you could apply it to a different industry. And then it feels like there was a pivot where you started to get market demand from a place you didn't expect and you built the services around it. Is that fair to say? Yep. Mhmm.

Yeah. That is amazing. And so now serial entrepreneur gotta figure out what to go do next, what new problem to go solve. Tell us about refi buy. Yeah.

So so in August of, '24 last year, I decided, I had a cofounder that wanted to take over and, you know, I I like to get things up to a certain scale, and then and then, it's off, you know, to someone that's more capable of running things to that to that next level. And, also, I have to admit the siren song of of, you know, AI, and when I'm I'm gonna use Gen AI just to kinda like I like to draw a distinction between, you know I took an AI class when I was, like, in the in the, you know, nineties when I was taking a master's. So, their AI has been around forever. Gen AI is kind of the new one. And, so and Gen AI was born in fourth quarter of two thousand twenty two, kinda when ChatChippity came out is is when I draw that one.

So Oh, okay. I was really enamored with that. And, and then what happened is so so to kinda set the stage, and I'm I'm sure you guys deal with this as well, and you've at at your previous gigs, you've seen this problem. The the weird thing about ecommerce as someone kinda coming into it like I did from from, you know, kind of an engineering perspective is in in engineering world, we like things to be, like, really well specified. So we have this group called I triple e.

I can't even remember what it stands for. And, you know, so so in the electrical world, for example, every little thing has a specification for it. So your your panel in your house, everything, you know, amps and watts, and they're all rated, and we all know how that works, and every part has a spec. And and ecommerce grew up, so fast and kinda haphazardly, the product catalog has no spec. So so it's totally random, and and it is it's basically, you know, a hot mess, for lack from an engineering perspective.

So Every company every company has their own Everyone. And it's nothing to be embarrassed about. It's just like how it and in fact, it's there's incentives are aligned so that they actually diverge. So, like, Nike and Under Armour or anyone that's competing, they don't even wanna use the same names for their, the word size. Right?

This is like it's crazy. Or colors. You know, everyone's trying to trademark a color, so we can't just say red red, yellow, green, or whatever. It's always like, you know, firebrand magenta or something, and it's just like everyone's creating new colors. That makes our job so so at ChannelAdvisor, our job was everyone would send us their product catalog and say, here's my product catalog.

I would like you to gusset it up and send these, you know, this portion of SKUs to Amazon and these ones to eBay and these to Walmart and these to Target and so and and, you know, we wanna send these out to Mercado Libre in Brazil and translate it to Portuguese and blah blah blah. And, you know, the the the challenge there was it's kinda like getting everyone's dirty laundry dumped on you at once, and you're kinda, like, trying to sort through it. And and, you know, and then none of the none of the socks ever match up. So it was kinda like that. So, it was it was a huge engineering problem, and we threw a lot of smart people at it, and we could we could kinda knock it down half.

So we could kinda, like, get this problem down to about half. And then the other half is in this category in computer science, we would call non deterministic. So so an algorithm or heuristic can't really figure it out. And, and at that moment, AI wasn't even near to being able to figure it out. And, so and data science couldn't figure it out and all that kind of stuff.

So, you basically had to kick it out to humans. So so we had a team starting in North Carolina that would look at these things and and kind of, like, figure out what what's going on. And then ultimately, that team got to be so big, we had to outsource it and outsource it again, and it ended up in Bulgaria. And by the time we were processing, you know, 15,000,000,000 a year in GMV, we had, like, 300 people on that team, that, you know, were were adjudicating all these things that had to be decided. And this this matters in our world, because let's say you you're selling on eBay or Amazon or a marketplace, they all have these rules.

And one of the most strict rules is there's always performance rules and, you know, catching fraud and that. None of our customers were in that bucket, but then there's this one over which is called significantly not as described. So if we did something and messed up a color, the the customer gets really mad, right, because they're expecting the Under Armour shoes to be in, I don't know, whatever Texas orange. And, you know, we said we got that confused with standard orange. And they will not only will they return that product, but frequently, they will file this ticket called a SNAD, significantly not as described.

And the way it is is you kinda get three strikes and you're out. So we're helping you know, our software is running, you know, a $100,000,000 businesses on Amazon that are three mistakes away from being sent to zero. So that's why it's important. So, you know, it's not like, this is why I like ecommerce versus health care, you know. We're no one's gonna die in this scenario, but it is gonna you know, they're not gonna be a customer for long if that happens on our watch.

But it is it is a problem. Then the other one is if you're like, okay, that's scary. So we're gonna, like, send a minimal amount to those guys so we don't get anything wrong. Well, then it won't sell at all. So that's also the bookends of this are, you know, if you if you kinda get risky, you get a you get a strike.

And then if you're not aggressive, then you don't sell anything. So so you had there's, like, this this this Goldilocks balance in the middle there where you had to kinda, like, do that. Okay. So, any any questions on that? Well, I a couple of thoughts.

So Well, this is a known problem in ecommerce. Everyone and it's kinda like we're we're like, oh my god. We're, like, stupid. We can't figure this out. And then, you know, you have these little backroom conversations with ecommerce platforms or retailers or, even competitive vendors, and they're like, you guys can do 50%.

How'd you figure that out? We're like, woah. We thought we were the dummies in the room, and we were, like, actually kind of ahead for a while. So yeah. Go ahead.

Oh, yeah. So I was just gonna say, I I just to put context for the listener, this significantly not as described, the reason that these marketplaces hate it so bad is because all the externalities of what happens from it. It's returns, expensive. Chargebacks, really expensive. Cancellations, negative feedbacks and reviews, calls to call centers, all of these things basically wipe out any margin for any any category that has relatively thin margins at which the retailers generally do.

So, this do you agree with that point? Is that why this is such a red flag issue for for these retailers? Absolutely. And and, you know, talking to the eBay team, they're they're, you know, to their credit, they're really transparent about this. They have a trust and safety organ.

This goes back this is a long time ago. This may not be the case today. But, you know, we'd have these conversations be like, you know, a $100,000,000 store can't live on three strikes, and their, you know, their argument was every time we get a snag this is like the slang of of ecommerce. Every time we get a snag, that buyer has a 99% chance of just churning from the plat we never see that buyer again. And some of these are are, you know and then they would say, you guys are correlated with our best buyers.

So so, you know, you're actually, like, you know, we we should actually have a a one striker out. And we're like, well, that's not where we wanted the conversation to go. Okay. We're we're okay with the three strikes. But, anyway, you know, and that's a valid, you know, that's a valid, And they use this term on their side of the marketplace operators.

We have GMV, which is what the seller gross merchandise value or volume, which is what the seller is producing. They had a term which was GMB, gross merchandise per buyer. So they would have these huge GMB, these whales, basically, these people buying a lot on the platform. They would hit us they would hit us significantly not as described, and then they would get so mad, they would just leave the platform. And that's that was their justification for having such a high bar.

Wow. It's the whales. It's like how Las Vegas treats big gamblers, these, mountain places. These big gamblers The gaming platforms all have their whales, you know, and, like, Clash of the Clans or whatever that is. They all have people that spend an ordinary amount on there.

Amazing. So now, obviously so now you have the basis of understanding. There is a problem that's pretty universal. The TAM is enormous. Anybody who has a catalog is probably has a problem.

And so you use that kernel of an idea, and and is that how we got refi buy? It is. Yes. So back to 2024. And, so I was reading up on a bunch of stuff, and, a lot of innovations have happened since ChatChibi Tea.

And what's really funny fun about this space is if you like to learn and you like to be challenged, it's growing exponentially and and there that is a a wave that is, you know, a fun challenge. I'm not a surfer, but that's what the crazy surfers you know, that's why they go to, like, like, when we have hurricanes here in North Carolina, we have all these inbound surfers that go to the coast because that's the best waves, and you're like, yeah, whatever. So that's what I'm doing. I'm like, this is the best wave, from a mental exercise. And, so I was I was learning about retrieval augmented generation, which is where you can kinda take subject matter expertise and put it into an LM to reduce hallucinations and give it some domain expertise that maybe it couldn't find crawling the web.

And then in you know, around November and December, Anthropic came out with some key papers. The first one was they started talking about Agintiq, where they and this existed before, but theirs was, like, a new generation of this. So so they early on, they would take an ALM and an API and let it do things, and it was kinda bad. So so it didn't really do much because it was kinda like, you know, it it was giving a, you know, someone that's never had a gun on Uzi or whatever. So it was like a bad combination.

And, but then what they did is they solved all that. So they they took an LM, so it's like the brain, and then they have a retrieval log mini generation, so it focuses on something you want it to do, and then they have memory, so it learns over time, and then they gave it tools, and they invented a new tool for it, called MCP. Because they're not really good at APIs, they can't figure out the context of an API. MCP, think of it as a way to for, an LLM to get an API wrapped with the context so it knows how to use it better and do things with it. And then I was reading this paper, and they said if you if you believe that, and it's kinda like this little Lego block or a a comp a composable component, but we'll call it a Lego block, same same thing basically.

There's some ways to put these together in interesting configurations to achieve different things. And one of the ones they talked about was an evaluation network. So you can take, like, four agents and give them all different LLMs. So you can have a Chad CPT, a Grock, and a Perplexi one or whatever. And then you can have them vote on something, and it's a way to solve these non deterministic problems.

And that really hit me because one of the ways some of the problems were so hairy, we would have a panel of Bulgarians look at it and vote. You know, so if it was three and it came out zero three or three zero, you knew you had an answer. But if it was two one or one two, you knew this probably needed further, you know, it was still risky to to to publish it. So I was like, wow. I wonder if we could use something kinda like this, this agentic thing to solve that problem everyone has in ecommerce.

Called some of my ChannelAdvisor buddies, and, one of them one of the engineers, interestingly enough, we we, we'd opened an Irish office, an office in Ireland, in Limerick because, we as we start we grew, and you guys have done this the reverse. Right? But, you know, at some point, your software company that's in one country wants to go into others, and you have to do this what's called internationalization, I 16 z and or I 16 n. And, it's, it's complicated, and this group in Ireland is really good at this. And so we had them do that, but they also ended up being really good at working on this catalog problem because there's a lot of, oddly, there's a lot kind of similar to, internationalization and what we call canonicalization, which is the the product catalog mapping problem and all the all the subsequent things that come from it.

So one of the guys there had gone to Micmac and was was actually just leaving, and he said, I am actually curious about this. I just kinda had a similar idea. I wanna work on this. And so he built a prototype in sixty days, and and interestingly, we just randomly picked pickleball paddles. I did because it's like a new category and kinda one that we could get our hands around.

And in sixty days, we had the perfect pickleball catalog. And, you know, so we had crawled the Internet, found all 2,000 offers for pickleball paddles, canonicalized them into 200 paddles, and found an all, you know, 100 attributes per. So you could ask it any question and could answer about any pickleball paddle. So I thought, well, that's kinda magical. And, you know, to the to the layperson, it would just look like you had and anyone could manually do this, but this was just like, you know, agents doing it in, like, an hour.

So it was like, that's in our world, that's pretty magical. I've never seen anything like it. And then serendipity happened again. Perplexity came out with Bion Pro. And I was like, wait a minute.

If the if the GenAI answer engines get all these users, right, so they're they're they're caught lightning in a bottle, especially ChatGibt. ChatGibt has 800,000,000 people using it, and they don't pay for them to they paid nothing to acquire them, and in fact, most of them are subscribers. They pay to use ChatGibt. That is very, very, very hard to do. In fact, the only the only companies that have done that have become trillion dollar companies, Facebook slash Meta, Google, etcetera.

So they're well on their way to being like a you know? And then when you have all these users, you're gonna need to figure out how to monetize them. We live in a world where we are a great way to monetize things. So if people start shopping on these GenAI engines, they're gonna start buying things, and then the engines are gonna, like, make money from that. They will probably become marketplaces is my thesis.

And so this is a thesis I had kind of like in q four of two thousand twenty four. Now perplexity was more agentic at that time, so it it actually sends an agent down to the checkout to buy. But I thought, ultimately, if they wanna make money from that, they're gonna have to charge a vig or a percentage of sales, what we call in marketplace is a take rate. And, you know, and they're gonna want the payment. How if you're if you're trying to keep the user up at Chatt GPT as an answer engine, it's opposite of Google.

Google sends users away. They're they're like a search click go is what they want because they get paid on that c that click. Yeah. Yeah. If you're an answer engine, you want people to stay and and stay up at this level and get answers and do things.

So I thought, ultimately, they're gonna have to bring that checkout up there because if you research, find, and buy, if you research, find, and then leave to buy, that's no for for those guys. They want them to stay there, and it's also more convenient. If you're trying to get people what they want fast, you want all that together. Research fine by research. So, so we we then it turns out that to help people do that, you have to have our piece because the engines need that.

The engines the answer engines, are not keyword based. They're context and content based. So if you look at your average product detail page, it has five bullets of text on it, and that's because that's what a human wants. And SEO likes that too because it's keyword based. You can put all the keywords in the in the bullets, and everyone's happy.

So Google's happy and human's happy, human buyer. But LLM isn't happy because LLM's like, that's not enough context. I need to know, like, where you should wear this dress. Is this good for cocktail parties? Is this a work thing, or is this, like, athleisure and I'm gonna wear it to my yoga class?

I can't tell. So the the answer engines want you know, in a perfect world, they love an 80 page PDF about every product you have. So now we have this really weird world where you kinda have three three use cases for your product detail pages. You've got SEO slash Google, you've got, human, and now you've got GEO or or Gen AI engine optimization or or, you know, what do the LMS want? And that's what our technology does is we basically, can the way it works is you you give us your product catalog, and and we, you know, we look at it, we evaluate it, and we say, for your existing attributes, here's ways to make them better.

And then we say, we think you're gonna need to add forty, fifty, 60 attributes. And then the the customer is always in the middle. We we know our our brand or retailer is always gonna want that. And then they go in and they they kind of massage this, and they add the attributes. And And then, you know, we do not advocate those going on the PDP.

Again, it's like not not this is not for human consumption, but we then distribute those that enhanced catalog and enriched to the engines. Then we have a monitoring capability. So there's a lot of companies that do just this piece where they just kinda run sentiment analysis. We actually go so they're up at the they're actually above research. Before even people research, they're kinda like, what do people feel about Nike today, and do they like our running shoes?

It's kinda like the things you prompt for for traditional GEO. Where we are is we're actually looking all the engines now have product cards that come up, and then they they can show you who has that item. We look at that to see where our customer is on there, and then our system grades that and feeds it back in. So it's this continuous optimization loop through the product, and it always starts at the product catalog because everything in ecommerce has to start at the product catalog or else you're never gonna capture it right and figure it out and optimize it right if you're not going through there. So that that's what we do.

Amazing. I mean, that was a very, very clear explanation, Scott. I I've spent some time on your website. I think you should play this video on your website. It is a very, very good explanation of what you guys do and why you do it.

My initial Only only an insider would understand the, the whole story, but yep. So Yeah. I that's phenomenal. That's the that's the Jason explanation because I know I know people that are inside the industry kinda get that. But if you're, you know, if you're relatively new, that may not make a lot of sense, but but that's that's what we do.

And that's that's where the idea came from. So a couple of questions following on. So first of all, I love how you see a what I love about founders is they see something before others see it. And and later, over time, these things become incredibly obvious. Like, yeah, marketplace is incredibly obvious.

Servicing, you know, downtime for avoiding downtime for Amazon trucks, obvious, but nobody did either. And so the fact that you're already kinda thinking about GEO and how to optimize for those, and that's it's amazing. Really, really interesting. Two two core questions. The the first one is, how do you make money?

What's the business model on RefiBuy? So the business model is we're, just like, many companies in the space, we decided to go pretty traditional with this. And it's a it's a b to b software company, so our customers are retailers, brands, and their agencies that support them. And, you know, we we we price it by, because we're we're built we're Agentic ourselves. Everything is agents all the way down.

You know, it is relatively expensive. It seems like it it's cheaper than humans, but it's not a because it's built on GPUs and all the underlying infrastructure, it's more expensive than compute. So, traditional CPU compute. So but it's orders of magnitude cheaper than if you were to hire people to do this. So we do have some limits.

So we basically say, if you have this many SKUs you wanna put in the system and it starts at 5,000, so it's kinda 5,000 and under, and then there's tiers from 5,000 all the way up to you know, we have some customers we're talking to now that are up into the hundreds of thousands, which were, you know at ChannelAdvisor, we had sellers that had, like, 8,000,000, 10,000,000, 12,000,000 SKUs. We're we're we know how to build systems that can scale to that level. But there's a there's a tiering of the pricing by SKUs. And then also the monitor, function is relatively expensive. So we we we and it's hard to kinda digest it all because we always have a human in the middle.

We don't want them to get overloaded. So we usually say whatever SKUs you're gonna put in, five to 10%, which is is what will be monitored at a single time, but then we'll rotate those through. So so let's say you have 5,000 SKUs, we're gonna monitor 500, and the goal is to own the product card is what we call it. So when you see a product card, you're number one. So in those first 500 we look at, you may already own a 150.

So so we're gonna we're gonna mark those as done and then sub at another one fifty, and then as we work on the remaining three fifty and knock those out, we'll we'll constantly we'll eventually chew through your whole inventory, and it starts to go faster over time because we as we start to see the common patterns between the answer engines and your catalog, you know, by the sixth time we've seen this variation tweak we need to make, the system just starts there and and in fact can proactively do it on the next. So so so let's say you're a shoe retailer. So so one of the ones that we have been able to announce is Steve Madden. So so, you know, it took a so let's say there's some nuance of selling shoes on Perplexi, Attach, BT, Grok, etcetera, all all eight engines, Gemini, so on and so forth. You You know, the system learns that.

So this is the other thing that's really weird from a software perspective is our system surprises us every day because it's learning things, and I've never had that experience as a software person before. I've had, you know, the stuff engineers build to be surprising, but I've never had the software, kinda freak me out, if if for lack of a better word, where it's kinda like, you know, you you know, we think you should do this, and that kind of thing. So yes. Phenomenal. Okay.

Thank you for that. The second question is around, how people access Refi Buy. Do they get it, direct from you? Do they get it from their agencies? Do they get it from a a ecommerce platform app store?

Is it compatible with all kind of systems? Does it does it disintermediate them? Yeah. So we're not, you know, our job and and, again, we've had a lot of experience at this. Our job is to not displace other things.

So if you have a PIM, we don't wanna get involved in the PIM wars. We're certainly not an e commerce platform. We're very e commerce friendly. Whatever platform you're on, this is this is a a teeny tiny project with a p, with a little baby p. This is not a this is not a multiyear journey that that many things in retail tend to be.

And, you know, and, you know, this ends up being good because as I've talked to a lot of, retailers and brands, the big roadblock is the legal team, because once you say AI, the legal team throws up the Chinese wall. And, and, you know, the good news is our our argument is the only thing we're gonna only data in your organization we're gonna touch is your product catalog. And, then sometimes lawyers don't like to hear this, but the the the overall fact is your product catalog is out there already. I can run every one of your SKUs. Yeah.

I can run every one of your SKUs through any of the answer engines, and they know as much about it as you do. So, so this is our not only is it not proprietary, it is already basically open source or or in the common domain. In the common domain. Yeah. Yeah.

Now they would argue it's copy yeah. So the lawyers wanna have an argument over that, and I'm like, you know, I'm not gonna get in this. If you wanna sue any of those guys, that's okay, but we are not the ones leaking your catalog to them because it's already out there. That's my whole point is the vendor in the middle. We don't need your sales data.

We don't we don't want any other data, because we don't want any risk. So so this is actually a pretty low risk thing for you to start AI in your retailer brand. That that's a positive because we're only doing the product catalog. And also it's relatively as you know, it's relatively easy to get the product catalog. So many, almost every brand and retailer generates a Google Shopping feed, and we can start there.

We're also experts at, web data extraction, which is also known as crawling or spidering. There's a million names for it. The politically correct is web data extraction. So so we also we like to do that as well as get a feed because we wanna look at your site in the lens that the, all the answer engines are building a web index just like Google has. And we wanna we wanna look at it like they do because we we frequently find structural issues that would stop all the secondary and tertiary optimizations we wanna do.

So we need to look at the structural stuff first, and that's best done as a web crawler. So yeah. So that you know, so we integrate with whatever platform, we don't have to do that upfront. We can also get started without integrating with the ecommerce platform because, you know, we can start making our optimizations, and then they're in our system. And then whenever you wanna extract them out, it's your data.

So so this catalog enrichment we've done is what you've paid us for. It is your enrichment. You you then can put it in a PIM if you want. You can keep it in our system however you want. We're not gonna delete it.

It is your data. So so we have a lot of those mindsets that are kind of in the mock alliance in these things of, you know, composability, it is your data, it's not a walled garden. All these kinds of things are just core beliefs that we have. And and Yeah. Yeah.

Yeah. I love it. You're taking a very smart strategy around not being anybody's enemy, and and the ability to just help merchants. That's all you're doing. You're just trying to help merchants sell more, which everybody who's in the value chain should care about.

Yeah. The part we're still figuring out is, like, how do we make sure we fit in with your workflow? So so what's neat is there's, this is, like, an interesting problem you would probably understand. So our customers, mine and yours, are used to a soccer as a service interface. Right?

So they're used to, like, a menu and nav and pushing buttons to make actions, and we we have that. But when our engineers work around their system, they basically chat with some agents. And they they have a they have they have a chat interface with the agents, and the agents all have they're like people. Right? They have their task, and and they they actually have a an org structure.

They have, like, boss level and super boss and stuff. And that's how they interact with it. And and and then part of it is a chat, and then part of it is almost like a people management platform, like a monday.com, where you can kinda see projects and how the progress is going and which agents are doing what. That's how we look at it. And eventually, I think those worlds collide, but I know our customer isn't quite ready for that yet, because it's it's head explode emoji, because, you know, you you you have to basically personify the agents.

And we think of them almost like people. They're like digital workers. And sometimes you'll see the CEO of NVIDIA talk about this, and you're like, that's really weird, but now we're doing it. And it's kinda like, like, I guess it makes sense. So yeah.

So it's really interesting to think through how do we make sure we plug into that workflow and whatnot. We we we are we should probably talk offline. We're dealing with exactly that use case. We've got Yeah. A bunch of agents that we are deciding what to call them.

We're calling them Codys as an internal code name, but they're but they're for developers to play with and interact with like humans. So I completely appreciate the point. And then the marketer wants the interface and the developer wants the code. So it yeah. It's a serve a lot of masters kind of kind of scenario.

Yep. Well, I think we are as you always say on the Jason and Scott Show, we have used up all the time again. My favorite part of having you on the pod is that I get to kinda use the the Timoo version of the Jason and Scott Show. I'm just the, you know, the the lesser Jason. You talk a lot less than Jason.

Yeah. Well, sometimes. Yep. Listen. I really enjoyed having you on the pod.

Really, really insightful. We'll publish and share this out, and and then I the people who are trying to figure out what to go do with AI and how to have really near end turn impacts really need to check out Refibuy. You guys are built for this is what I love the most, to solve today's problems with today's product as opposed to taking old tech saying you're an AI company and and trying to kinda squeeze it in. I I just absolutely thrilled with with what you guys are doing. Yeah.

There was a company at, this retail club show that's been around for a long time, and they're trying to be agentic. And they said, humanizing the product catalog for real people, and then they added and agents. And it's like, how do you humanize the product catalog for agents? Like, it like, it was just funny. You could tell they were like, oh, let's just throw and agents on there on the end.

It was like, we call them and agentic companies. So yeah. Yeah. Exactly. Yeah.

And then for your audience that's interested in this topic and, you know, not even, you know, not even on the refi buy side, but we also found that there's no one really talking about all these issues. So we started a a a podcast and a substat called Retail Gentic, and it's the intersection of a Gentic and retail. So so I think that they would enjoy that. So we have some news. We keep up with just the news on this topic.

And then also, lots of real world examples of, so for example, there's all these browsers coming out that have agents. What what's it like to shop on those versus shopping on ChatGPT? What's going on with Rufus and all those things kinda like behind the home page? So yep. So we're trying to cover a fair amount of that to keep people up to speed on what's going on.

Amazing. Well, we'll put all of the links to all the things you're involved with, the Jason and Scott Show, RetailGentic, RefiBuy, GetSpiffy. How are you? I mean, geez. You must not see your friends and family very much.

You're a busy guy. I try to be I spend a lot of time, trying to be as efficient as I can. So yeah. Love it. Well, thanks for spending about an hour with us.

Scott, you're a star. Appreciate you. Thanks, Jason. I I really enjoyed the conversation.