I had the pleasure to chat with Mike Williams of Everything Marketplaces about the state of marketplaces in the age of AI.

0:04 Intro
‍0:58 Fabrice’s background
‍1:44 What Fabrice looks for in marketplaces when investing
‍4:46 The state of marketplaces with AI
‍17:28 New strategies for AI-native marketplaces
‍19:03 Growth rates for marketplaces with AI
‍22:50 How AI is changing the capital needs for marketplaces
‍25:40 Defensibility for marketplaces in the age of AI
‍28:13 Tips for founders starting marketplaces in 2026
‍31:08 Group Q&A on how investors evaluate managed marketplaces
‍33:44 Group Q&A on tips for transitioning to being AI-native
‍34:55 Group Q&A on calculating LTV
‍36:45 Group Q&A on seller-side CAC
‍38:17 Group Q&A on fundraising benchmarks in 2026
‍40:10 Parting marketplace advice

In addition to the above YouTube video, you can also listen to the podcast on iTunes and Spotify.

Transcript

Mike Williams: Welcome back to Everything Marketplaces, where we talk about founders and leaders on some today’s top marketplaces. So this is episode 2 0 9, which is really a good group chat had with Fabrice Grind, who’s a partner at FJ Labs. FJ Labs is an early stage venture fund that’s backed over 1200 startups, including marketplaces like Alibaba, Flexport, Clutch, and more.

Fabrice is also passed group chat guest. This is great to welcome back on where we started off with a quick overview of FJ Labs Venture Fund and what they look for in marketplaces when investing. We did a deep dive into marketplaces in the age of AI. We discussed topics like how AI is making it easier for marketplaces to do more defensibility, the current state of fundraising.

Fabrice share lots of tips for founders and we also had a great group Q&A. I really enjoyed this conversation. I think you are going to find it a great watch to the end.

So Fabrice, it’s great to have you join us again for the follow up group chat today. I feel like so much as I changed since you’re up past chat, that was only a few years ago, and now we of course have a lot that we’re going to dive into with this. And, marketplace is in the age of AI, but I thought that it’d be great if you could set off with a quick refresher your background for those that might not know and that have just joined us here.

Fabrice Grinda: I’ve been a tech founder and investor since 1998, so 28 years. Built three large venture backed companies. The last one of which I grew to 11,000 employees in 30 countries, over 300 million uniques. It was a Craigslist for the rest of the world, and throughout this entire timing starting in ‘98, I started investing in marketplaces.

And so I, and I professionalize that. In 2016, creating FJ Labs, which is a venture fund specializing in asset-like businesses that work back businesses and marketplaces were large. We’ve invested, I guess in three, yeah, probably should mention a few sets. We’ve invested in 1300 businesses. We’ve had over 300 exits and so far compounding in like whatever, around 30% IRR realize.

Mike Williams: That’s really incredible background, of course. There’s so much that we’re going to get into here. I guess as a quick kind of refresher. Can you give us an overview of FJ Labs, as a venture fund and what you look for in marketplaces when investing?

Fabrice Grinda: FJ Labs is really a reflection of my personality. Like I was a super angel became me before I became venture capitalist. And I never did like top down fund construction. It’s where I see a deal, I meet a founder I like and that convinces me and I invest. And so because I am intellectually curious and I think I see tech as solving the world’s problems. So in the 21st century, I think the three fundamental problems we’re trying to fix: climate change, inequality of opportunity, and the mental and physical well-being crisis. And I think I like using tech, especially asset-light tech to be better, cheaper, faster than alternatives. Now, what I look for hasn’t fundamentally changed in any way, shape or form.

I have four selection criteria that are still the same they are today though there are a few things that I’ve changed in the age of AI, but the four selection criteria are the same, which is, “Do I like the founder?” which for me is someone who’s extremely eloquent; because whether you want it or not, as a founder, you’re a salesperson, you’re selling employees, you’re selling business partner, you’re selling investors, you’re selling the press, whatever.

So it doesn’t matter if you’re introverted, extroverted, you need to be eloquent. And number two, “can you execute on set vision?” And they’re both necessary but insufficient so if they’re separate. So you need both of them together to make a great founder. And the way I tease out in a one hour call if someone can execute is number two, “do I like the business?”. Which is a combination of total addressable market size, but more importantly for a marketplace unit economics. What is your fully loaded customer acquisition cost, both in the supply side and the demand side? What is your net contribution margin per customer? What do your cohorts look like over time?

What’s your LTV to CAC on a CM2 basis? And whether regardless of the stage you’re at. I expect you to be able to answer this. Even pre-launch, I want you to have thought about it. If you haven’t thought about it, it’s very unlikely you’re gonna be able to execute on the idea.

Number three: what are the deal terms? And I am price sensitive. And I don’t mean I don’t want anything cheap. I want it to be fair. Fair in light of the traction, in light of the opportunity, in light of what you’re solving.

Number four, just as fundamental frankly, is “are you solving a problem that I care about that is making the world a better place from my perspective?” Obviously that is bias, but I’d be pretty. Clear sense of what I, where I see the future of the world heading, the future of mobility, the future of robotics, the future of food, et cetera. And is your idea and it’s good to be aligned with mega waves, right? To be going with the current of history and not against it.

And so is your idea aligned with the direction of history? And so these four criteria are collectively true. I invest and we invest pretty quickly, two one hour meetings over the course a week; we’re in, we’re not. So that hasn’t changed, but yeah. That’s great.

Mike Williams: No, the criteria are gonna be very helpful here for everyone. So I’m sure it’ll lead to quite a few questions when we get to that group Q&A. But, I wanna transition to talking about marketplaces and the age of AI that we’re in. Of course, you’ve seen an impact on marketplaces. So I guess the starting point, where are we at right now with things?

Fabrice Grinda: I would say we’re in peak year from marketplace boundaries and public investors as to the impact of AI on marketplaces. They were like, oh! The top of funnel is going to move all over that to the agents and therefore, you’re not gonna be able to have margin, you’re gonna have massive magic impression and disintermediation. Perhaps the short tail LLMs like, GPTs and Claudes of the world are going to vertically integrate and will, they’ll do everything and there will be no room for it for marketplaces at all. And perhaps the entire sack goes agentic commerce and there’s no even a human in the loop and therefore marketplaces are irrelevant.

And I could not more fundamentally disagree with literally every one of those statements. First of all, when I’m looking at the behavior of users today in marketplaces, I am not seeing a shift to the LLMs that I don’t expect there to be one, because there’s three purchase behaviors on marketplaces.

One is, you’re browsing for entertainment, the equivalent of walking down on, on Broadway and SOHO. And going from store to store. So if you’re in a, one of these types of sites, so think of a Vinted in the used fashion category where it’s really browsing as entertainment and you’re browsing an average 20, 30 pages per visit.

You’re visiting multiple times a month. You’re not really looking for anything, but it’s an a pretty typically low average to value you. Something like you buy an impulse buy. Again, there is no world in which that feed goes to the LLMs. Like it’s not in the top thousand priorities to create a feed of used whatever fashion, a 30 euro average price point that you might wanna buy in any of the LLMs. They’re organized for efficiency, not for someone who’s like trying to brag. In a world where you’re not looking for efficiency, it doesn’t work. Even if the traffic moved there, I didn’t think it would be a big deal.

Number two, if you know exactly what you’re looking for, there’s also zero reason to go to one of the LLMs, right? The current behavior is you go directly to Amazon or eBay or to a luster extent, Google, you type in the model, whatever LG TV C3 65 inch OLED and boom, you get it.

And by the way, even if you want to Google, the amount of value they capture is pretty de minimis because. 43% of the of the results are actually coming to Amazon and eBay. So the liquidity is still provided by concentrating people on the backend. Now there’s a third category of considered purchases where I can make a case that LLM will take more of the value.

Because if you don’t know what car you should buy in light of who you are, where you should live, et cetera, then a conversation that LLM makes sense. And so for considered purchases, but then again is Carvana get or be going to be better positioned to tell you, or is it GPT? It’s unclear.

It’s possible it’s GPT. But even then, let’s say though, so I don’t think, first of all, most of the traffic is moving the LLMs. That said, LLMs now are about a third of the traffic of search and it’s free. So you should absolutely no questions asked in index yourself in the LLMs. And you should do aggressive AEO.

And the early adopters AEO, are gonna be winning in the free traffic war. And so you can get a lot of free traffic there. So absolutely index yourself there. Don’t let them use you for training data, but index, you’re listening such that you show up in the results. Okay, so now it, let’s assume a worst case scenario.

For whatever reason, I’m wrong. A hundred percent of the top of funnel goes to the GPT to GPT. And all the searches start there. Now let’s think through, how much of the value is GPT going to be capturing in this worst case scenario world versus the marketplace? And it comes down to, frankly, the job to be done, right?

So if you’re telling me. You’re booking an airline ticket and they’re five airlines and they have 99% market share, and they’re oligopolistic. Okay, you’re screwed. But then again, in today’s world, how much money is Expedia making from airline tickets? Nothing. Like they, they’re not really a marketplace. They’re a distributor for agents. And by the way, this is true. Have any one of you building a marketplace out there? If your supply is concentrated, you are not a marketplace. You’re a distributor for it, whomever you’re selling, and they were gonna have pricing power over you, and you’re not gonna be able to have reasonable margin.

Now think of an Airbnb with millions of listings or DoorDash with hundreds of thousands of restaurants. That supply, which has been aggregated, is extremely hard to replicate. There is no world in which GPT is going to go do that. And by the way, the more work that you then do to match them with couriers.

DoorDash has the three-sided marketplaces, like the buyers, the restaurants, and the couriers. And the more you do in payments, so logistics, et cetera, the more value you capture. And let’s not forget, most marketplaces are winner takes most and most of these categories, and you know there’s basically just Airbnb and then there’s DoorDash and Uber Eats, but that’s it.

There’s Uber and Lyft, and that’s it. The more there are the  small recurring transactions with massively aggregate disaggregated supply, the more value you’re gonna capture. So the most I can see A GPT capturing would be the equivalent of a Google SEM at the top of the funnel, and that’s assuming a 100% of the traffic went there, and I don’t think the traffic will go there.

So, I would not worry about “what is the impact of like  eugenic commerce?” And by the way, eugenic commerce today is zero. It is non-existent from a volume perspective. Like there a non-human d directed purchases of items by agents is like de minimis. And will it grow over time? Yes. Do I expect it’ll be more than 10% of commerce in the next five years? No.

So number two, so I would actually not worry that much about the threats I just mentioned, but instead worry on the opportunity. To me, that’s where people are underestimating is how much you can do today as a marketplace founder using AI. And it is mind boggling. Lemme give you a few examples like maybe six categories of things you can do.

One is cross border commerce, that’s maybe less relevant if you’re selling in the us. But in India, for instance, where there’s so many languages, Bengali and Tamil, and Hindi, et cetera. It used to be disaggregated marketplaces. Selling locally like CDC did, could not work. You could not sell from one consumer in one region in another because they didn’t speak the language.

And that was true in Europe. Europe was never Europe. Europe was France and Germany and the UK and the French people don’t speak English or German, and you could not be shipping across borders. But now Vinted, which is the fashion marketplace I mentioned earlier with over 10 billion in GMV. A billion in net revenues, hundreds of millions of free cash flow.

Their geniuses, they’re auto translating the listings. They’re auto translating the conversations between buyers and sellers. And so they’re using their French liquidity when they’re, where they won to then go to Spain to go to Italy, and they immediately have liquidity and they immediately become a dominant force. We’re investors in marketplaces like even B2B marketplaces in used car like CarOnSale in Germany, which is now selling in France. And like within a year 30% of sales. So you can basically go cross border and in collect your business 30% in a year using AI.

Number two, listing on marketplaces historically in most marketplaces, it’s 1% seller, 99% buyers. And it doesn’t matter what, whether it’s services, whether it’s products, et cetera. Hardly because the barrier to listing is reasonably high.

If I need to take 20 pictures and write a title and write a description in the selected category and select the price, it’s a fair amount of work. And most people are lazy and are not willing to do it. And that’s even especially true for low average order value items. And it’s also true for categories where the transactions are complicated.

But now with AI, you can just take a photo of boom. Category description, boom. And you’re done. And like even improved listing with higher conversion rate on the selling. So you absolutely should be using AI with the proper training and the data sets for your category to simplify the listing process to the maximum to improve the percentage of visitors that are sellers and generally improve the user experience, that improves the buy rate.

Number three, you can fundamentally inflect your cost structure. We’re investors in a company called Ace Waves, which does customer service AI for marketplaces. The marketplaces that have embedded it have lowered their customer care costs by 50% in six months, 50 while improving NPS.

At the same time, by using Vibe coding, you can improve your program and productivity dramatically. So you should be able to improve programmer productivity, lower cost dramatically.

Number four. Marketplaces these days have more and more revenue streams coming to them like it used to be you take a commission and that’s it, but these days it can be, you have a SaaS subscription fee, plus you take a small commission.And one of the big ones is selling advertising. Your own sellers, buying advertising. By the way, most people don’t realize Amazon is a marketplace. The vast majority of items sold on Amazon are third party items. Amazon is totally a marketplace which manages payments in pick and packing and shipping. So logistics, returns and customer care, but they’re a marketplace.

Their sellers, their merchants buying ads on Amazon for sponsored listings is now a multi-billion dollar category. And the beauty of that of advertising is it’s a 95% margin product. And so companies like Instacart, 5% of GMV comes from self-service ads, and it’s like the vast majority of the profits.

And so marketplaces as they scale, should totally implement advertising, especially sellers of the marketplace, buying ads to promote themselves using companies like Topsort. And obviously AI, right? The thing is, doing it is pretty hard. Like you can’t just be selling to the highest CPC because actually what you’re optimizing for is not CPC, it’s CCM.

You’re optimizing for the CPC times the click-through rate. And so you need to have really smart AI to figure out what ad to your display where that is gonna lead to the highest CPC times the highest CTR. And so it’s totally AI driven on the revenue generation side.

There are millions of other things you could do tracking of the items. I mean that basically your company should be built AI forward and AI first with integrated AI from the get go, and this can inflect the business. So I would look at the opportunities of AI over the go.

Actually no, one more thing. There are many categories where marketplaces could not exist in the past because too much human labor had to go in making the transaction happen. And there are many categories where, imagine you’re a general contractor and you’re doing a build somewhere.

The number of sub subcontractors you’re working with is mind boggling, and everything is in iMessage or WhatsApp and there’s no Gantt chart, et cetera. Now you could use AI and agents to basically replace the job of humans and like actually streamline workflows, lower costs. I know construction companies that are now, their tech companies, not construction companies, they’re selling tech to allow people that are like bidding for RFPs for building commercial or warehousing or replying to RFPs from cities. They basically build the RFP where the agent, they can get through the entire licensing process for the city.

They can press the timeframe from years or month to weeks, basically. And so there are many businesses that could not be done before that you can build that at the age of AI.

Mike Williams: That’s a great breakdown. How AI is impacting marketplaces and I’m glad we also spent some time addressing the how marketplaces can do more and also the why now, right now for marketplaces as well, those are topics that we’re often discussing in the community.

And another topic that I wanted to transition into talking about is some of the new kind of strategies. ‘Cause these AI native marketplaces, sometimes not taking the more kind of traditional approach to starting a marketplace. So what are the, some of these new strategies that you might be seeing that, that are marketplaces are starting to use today, that are playing out?

Fabrice Grinda: Yes, but I’d say they’re tactical. So I know many marketplaces that right now, they’re using. I guess you could use Claude Co-work but before Claude Co-work came out, they’re using OpenClaw. That found out all the leads of the supply of the demand on LinkedIn. Created a LinkedIn InMail account, and reach out. And basically with essentially zero CAC or underlying credits, so maybe not exactly zero.

And was able to build both the supply and the demand of the marketplace in a category. I’ve seen examples where people coded their open class to basically do telesales where they were coding a voice interface slide, like with Whisper or whatever. And then they would call where they’d get a Twilio number and they’d call and essentially could be cold calling at reasonably low costs, either supply or demand that they’re trying to do.

So there are many things you can do to hack the liquidity using AI in, in today’s world. So I, but again, to me it’s tactical. So here you’re using the tools in order to do interesting things or you’re using agents to basically fill roles that would’ve not been uneconomic and you could not create a marketplace for in the past where now you can make a market least work.

Mike Williams: Yeah, definitely great. It’s tactical and nuanced as well. And another thing with these kind of new AI native marketplaces starting, we’re seeing them get off the ground and getting liquidity and growing faster or never. So how’s this potentially change on how you evaluate you know them when investing and maybe like what some of those kind of benchmarks are for growth.

Fabrice Grinda: It’s part of raise for marketplaces today. And the reason it’s hard to raise for marketplaces today and the A and the B is because AI has captured all the oxygen in the room, and they’ve had these examples of companies going from zero to a hundred million in AR and 90% margin and the Lovable and Cursors et cetera in the world when you’re measly marketplace goes from zero to 3 million in GMV and then 3 million. 3 million GMV to 15 million in GMB to 50 or whatever, over three years. It’s hard to get, the bigger VCs that are writing the bigger checks, excited.

That said, you could do a lot more with a lot lots. And so what I’m seeing more and more of is essentially marketplaces bypassing the pre-seed, right? It used to be your pre-seed stage was a million and whatever, six free, again.

If you’re in YC it’s gonna be 30 or 40, but so there’s a range but maybe it’s eight, these days or whatever. And you needed a million to get a retraction where you could get your seed round and your seed round. And by the way, the seed for marketplace businesses, the valuations haven’t changed all that much.

They’ve inch up a little bit, but frankly. We’re still seeing a lot of seeds where the company’s doing 150 K a month at GMV, 15% take rate with the 70% margin, and they’re raising pre at 12 pre, right? And maybe they’re raising five and 15 or whatever, but it’s like not, and maybe the median series A now in, with a 15% take grade, right? It is maybe 10 at 30 pre with a seven 50 K million a month in GMV. So we’re, I guess the traction expectation has gone up because you should be able to get more. And so we don’t do that much idea anymore, it’s people come to us and it’s kind, even though they’re two, it’s already live, they’re already at subtraction, they’re ready to pay unit economics, et cetera, because you can build anything with nothing. And in fact, if you are not capable to launch, you get traction with very, with like peanuts money.

It actually speaks to your inability to potentially execute in the category you’re in. And I’d say that a big trend is bypassing of the pre-seed ran, going straight for seed with traction. Now look, if you have, there’s a few exceptions where people have been able to scale, like dramatically, and they go from not the old school was you were at 150K a month.

You raise three, you go to seven 50K a month, you raise 10 or seven, you go to 2.5 to 5 million a month, you raise 15 or 20 or 25. Sometimes you can accelerate that. The thing is, I haven’t seen that much acceleration in these businesses versus pure AI native type SAS subscription companies these have been able to scale it insanely fast.

I have not seen much of that. The only places where GMV scales like crazy is in the B2B marketplaces, but then again, if your take rate is 1%, I don’t care. I care about the net revenue, not the growth, the GMV. And so if your take rate is, so a million with 15% take rate is 150 K net. And so if you’re at 1% intake rate, I need you to be at 10 million, not 1 million to be at the same series A level traction.

Mike Williams: Certainly and you did mention, some kinda like skipping, pre-seed and just, being able to get much further. And so I guess related to that, how do you think AI will start to change the capital needs for startups and marketplaces specifically in the future?

Fabrice Grinda: AI itself has been extremely capital intensive, right? So, part of the reason it’s hard to reach for marketplaces right now is all the oxygen in the room is being captured by AI related, like 95% of YC companies, AI companies, 75% of the funding in 25 was going AI and I mean AI LLM type companies and almost all the capital went to five companies.

So not only is it all AI all the time, it’s also, Claw or Anthropic and ChatGPT capturing a massive percentage of that capital, followed by the whatever Lovable and Cursors of the world and ElevenLabs. And then defense tech, like things like Anduril, which is also AI adjacent with a big chunk of a AI or whatever figure, which I like big AI components.

The answer is pretty typical. It depends, right? I think it depends what your customer is like. At the end of the day, marketplaces, as I mentioned earlier are unit economic businesses. And so it’s a question of how quickly can you make the unit economic work and often the unit economics are not really in the early days. In fact, a clear signal that you have product market fit and that you have, you’re building an actual marketplace now effects is that your CACs are going down. You know, every more buyers bring more sellers. Every, more sellers bring more buyers.

A clear sign that you don’t is as time goes by, you’re spending more and more to bring the marginal users and then you’re more of a sales driven type of organization. And so to the extent that you have. Positive economics, lowering CACs and you’re able to automate more of the processes. Do I think you could go a lot further with a lot more, a lot less capital?

Absolutely. But again, have to the estate and I’ve been, I’m still investing in whatever, 150 marketplaces a year. I haven’t seen many examples of marketplaces that I’ve scaled, extra to extraordinarily scaled with very limited capital. I’ve seen them scale with very few people. Because they’re now using agents to be much more productive, but they’re still using capital essentially to scale supply demand, and you always wanna scale both in parallel. And it doesn’t matter whether it’s sales driven, it’s marketing driven, it’s agent driven. They all have costs in involved or associated with them. And so I have not seen, something go from zero to billion in GMV with no capital raised, essentially. Maybe OnlyFans rather than OnlyFans.

Mike Williams: Yeah, definitely. And you briefly mentioned earlier the kind of market sentiment and then, when it comes to, fundraising. But I think, a topic that right now that has been, coming to play a little bit more is the know around defensibility, right?

And then, so how do you think about defensibility, in the age of AI and specifically with marketplaces now.

Fabrice Grinda: So first of all you should all go to my blog and watch episode 52 Marketplaces in the Age of AI. And I have a few slides there that look at what are the things that make you defensible is AI and long term.

So first of all, whenever you launch your marketplace, you have zero defensibility, right? So your moat is your liquidity. Say at the end of the day, it’s liquidity. And once you have liquidity, you’re basically non displaceable. Just look at Craigslist, which is still shockingly sell relevant in 2026.

And from a volume perspective in certain categories, like blue collar jobs, despite being, having the worst UX UI, it’s intentionally destroying the liquidity, doing things like charging, et cetera, right? The way I would think about it is A, how much work are you doing as a marketplace, right?

Are you purely a lead gen marketplace, whatever, Zillow, Angie, and you’re not really doing anything or you heavily managed a lot, DoorDash, Amazon, where you have like warehouses and customer care and returns and payments and managing third party delivery.

The more managed you are. And the more fragmented your supply is the less the less amount of impact, the more defensible you are. The other thing I would think about is how frequently are people buying the products? And how much time do they spend thinking about what they’re buying? And again, the less time they think about think what they’re buying, meaning it doesn’t require research.

And the more frequently they’re buying the product, the more defensible you are. Instacart, DoorDash, Uber, Amazon. Etsy, eh! Not really worried. Now if you’re a high consideration purchase. So people are spending a lot of time thinking about whatever they’re buying or somewhere they’re hiring. And it’s not very recurring, probably a lot riskier.

And so I think there are a lot of things you can do to defend yourselves in market from the LLM by just doing more and picking the categories where this is true. But ultimately, once you have liquidity. It doesn’t matter. Even if the top of funnel went to the LLMs, they, because the transaction would be happening through you, even if it’s an agent exactly and you would be capturing most of the value.

Mike Williams: That’s gonna be a really helpful outbreak down for us here.  Cool. So, I’m getting some of my questions from founders and I promise I’m gonna save time for the group Q&A. Right before we jump into it, do you have a few tips for founders, starting on marketplaces today?

Fabrice Grinda: Use AI tools in the ways I mentioned earlier. Be super capital efficient because capital is going to be the hard to get to or to get by. Make sure you go as far further. Basically make yourself default alive in default and investable. If your traction is indisputable and the business so good that no, that some VC is going to have to fund it no matter what.

If you went from zero to whatever, 10 million a month at GMV with very little, the capital degree economics, it doesn’t matter the next, that stage is 30 million a month and not, a hundred, it’s okay. Someone is gonna in invest in this. And there’s still VCs that like marketplaces because they are defensible.

What people are not realizing is like 46% I think of AI companies ever created already are dead. You have companies that go to 4 billion in value, like all of AI, and they die. So your companies are a lot more defensible, but in order to make it, use the tools, be capital efficient, have great unit economics, and really find product liquidity and the way I would find liquidity and prove your product market fit, I’ll share the biggest mistake I think marketplace founders make. The biggest mistake marketplace founders make is because sellers on the marketplace. It doesn’t matter where they’re selling: products, services, whatever you use, et cetera. Or financially motivated be by the market marketplace, it’s very easy to flood your marketplace with supply, right? Anyone you go to, even though you say you have no traffic, you’re like, Hey, I’m launching your marketplace. You want a list here? It’s free and roll. The decommission sells. Everyone will say yes. The problem is that if you have infinite supply and you have no demands.

You’re gonna have zero engagement with the sellers. They’re not gonna be engaged. If someone actually by mistake, buy something, they’re not gonna reply. They’re gonna have bad experience. It’s way better to have the best, highest quality, most curated supply. You treat them well, you make them happy, you bring them.

Depending on the category, if it’s items they’re selling, maybe a use item, they’re selling like 25% probably of selling. So like low end of liquidity. If it’s a service, you wanna represent at least 25% of their income. Ideally a hundred percent. Like Uber, but maybe 25%. Then you start scaling the supply, then you add more demand and you scale them in parallel.

You never wanna flood your marketplace so much supply that your liquidity goes down. It’s liquidity, liquidity, liquidity. And when in that more liquidity.

Mike Williams: Yeah. That’s gonna be great. So I’m glad we’re highlighting and important to that. And then, I can speak from my own experience myself, grew kind of supply early on at the cost of demand, cool. We’re gonna get into the group Q&A. Hey Lisa, I saw you raise your hand. Did you wanna jump on here?

Lisa: Yeah. Hi Fabrice. Good to meet you. I’m Lisa. I’m building buddy at B2B Recycled Metals Marketplace. I know you are an index fund, get a huge amount of opportunities crossed your way and managed marketplaces and true marketplaces, right?

So both of those exist, but I think you’ll seen like revenue means can mean completely different things in a true versus managed. For example, we are take rate based. Everyone else in our space is buying and reselling its principle.

Fabrice Grinda: Yeah, but we’re not dumb. Meaning if someone buys and resells and they count that as revenue, of course, I still look at the margin that they have and also the capital deficiency if they have inventory.

So at the end of the day I’m going to be comparing apples to apples. So if your revenue is your take rate. I will look at what the margin structure of that take green. I’ll compare it obviously to what they did. So, I would not overthink. I would do whatever is right for the business from KPI OKR perspective and would not overthink this. So investors are not, and marketplace investors understand the difference.

Lisa: Thank you. And I know you’re a marketplace I investor, so you’re very familiar with that. But our experience has been non marketplace investors have, not maybe evaluated in the same way as you. And I was wondering if you had any advice around the framing, around being something that has a potential to become a transaction layer because it’s the nature of a true marketplace versus, only being able to take the revenue off the transactions that it captures.

Fabrice Grinda: You should to the extent you track it, mention all the GMV that’s going through the platform, right? Like then it becomes a lot more comparable to whatever the top line of the guys that are buying and selling.

Lisa: Okay.

Fabrice Grinda: Try to index them to GMVI. And then you can talk about what a percentage of the transactions you’re able to monetize today versus you’ll be able to monetize in the future and why, what the current take rate is and why you think that can tread upwards in B2B marketplaces,the take radio hat can take varies on the elasticity of supply and demands. And there’s some categories where frankly you can’t take anything. It’s zero. But there’s some categories where, you start one, two, or three, but over time as you have more liquidity and you provide more value, you can actually start training it up to seven, 6, 7, 8 or 9.

And then you can add things like financing you can add things like insurance, you can add things like whatever other value added insurances to get your blended effective take read or whatever, 10 to 15.

Mike Williams: Hey, Fabrice, I wanted to ask a question for a founder that I could not join, and this is somewhat common as well in the community, and it was around transitioning to being a AI native or, considering like how AI is impacting your business.

So we have a lot of marketplaces that have previously, say a year and a half ago, kinda raise like a proceed round. So there any, are there maybe any kinda like best practices or tips that you have for, businesses that have, previously raised pre AI kind of era and then, today thinking about like how it impacts their business and potentially transition it.

Fabrice Grinda: To me it’s the exact same advice as you’re building to start and new startup today. Use the tools, integrate AI, like there, there is be AI forward. And if you need to rewrite your stack, rewrite your stack. But then again, often you don’t need to. And by the way, like a lot of our marketplaces, they’re using off the shop tools, right?

Using Shopify because at the end of the day it’s less the tech even though you are doing things like single photo listing, et cetera it’s much more about unit economics and liquidity. That is the differentiator than the underlying tech platform. But yeah, be tool forward and do a AEO and index yourself in the LLMs.

Mike Williams: Hey, I’m Molly. Sorry. You raise your hand. Do you wanna come on?

Molly: Sure. Yeah. Hi. So I’m the founder of Recess. It’s like the booking.com of kids classes and camps. We’ve. Raised 4 million to date. We have less than a year of revenue, but we’re growing really fast, so we’re going out for our next round soon.

One question that’s come up in discussion with investors is around LTV, which our LTV is pretty long, given sort of the age of child that we serve. Our ICP realistically have zero data to back up that LTV. Is there anything other than, user research that’s out there about when the purchase cycle normally is for your product that investors would accept for LTV?

Fabrice Grinda: I didn’t actually hear what you were selling, but to the extent churn is low and people were paying monthly, for instance they’re willing to buy your LTV because the churn is low, right? So net revenue retention after a six month, 12 month. 18 month, et cetera, actually speaks to, even though, so most startups that we look at the seed have not been live for more than 18 months. And yet we can project an LTTV to CAC over five or 10 years based on churn.

So to the extent you have negative net revenue retention or, so you have 150%, sorry net revenue retention after 18 month or whatever you can actually do and you know what your monthly, annual churn rates look like, you can actually project that and we will buy it.

We will buy what your long-term LTV to CAC looks like because it is driven by your cohorts. And as long the cohorts keep looking good and the trim, remains the same, it remains good. It’ll be fine. So I don’t need you to have been live for 10 years to buy that. You have users that will be paying for 10 years as long as the data’s there to support it.

Mike Williams: Hey Julius saw you raise hand. Do you wanna don’t jump on here.

Julius: Wanted to ask on CAC on the seller side. What would be like benchmarks you would have maybe, or say where you’re saying CAC is actually quite good. And also conversion rate, meaning that we onboard a seller and then he basically that lists something on the platform.

We have 16,000 monthly active users. That’s like where we’re currently at. I’m just curious on the CAC side, on the seller side.

Fabrice Grinda: The CAC here support is driven by the, that you’re willing to support is really driven by the LTV, that the user that they have.

And so I can’t answer the question not understanding average order value, how many items they’re selling et cetera, et cetera. You want good economics, right? And usually the seller economics are much better than the buyer economics because one seller ends up selling a lot of items over a long period of time.

So typically, ’cause it’s easier to get sellers than buyers. The un, the unit economics and the sell side or the seller side are much, are like 20 to one or 50 to one or a hundred to one versus the buyer. The unit economics on the buyer side, were only three to one or four to one because you’re spending a lot of money on Google and these guys buy every, depends the category, but like once or twice versus the sellers keep selling over and over again.

And I would make sure that your seller economics or whatever, 10 to one, 20 to one, like on unit economic basis.

Mike Williams: Hey got Godfrey. Sorry. Raise your hand. Do you wanna jump on? We’ll try to get out. One last question in here.

Godfrey: So Godfrey here, founder of a cross-border shipping marketplace that help used car the last submiss missionaries are brought to find vetted freight brokers for shipping their cars from auctions in US and Canada overseas.

So you mentioned one of your four criteria for funding is good deals, and I was looking at the FJ Labs fundraising metrics. My question is how have these fundraising metrics changed?

Fabrice Grinda: Has it’s really changed honestly, in marketplaces? We’re, so the maybe an inch up a little bit. Both ramp size evaluation, but not all that much. The things that have moved the needle have really been the AI companies where you have like a company, like a AMI labs that was raising a billion seed at like 3.5 billion pre or whatever.

And then when you look at the averages, the medians in the industry, nothing makes sense because you have these stupid deals in AI that are completely inflating or conflating where the statistics are right. Other than the fact that people were skipping pre-seed, for the most part, we haven’t seen a very big change.

But what we have seen is in a difficulty in raising. And so, you need to have a growth story at good economics. That’s makes it basically so compelling that the VCs have no choice but to take you seriously and invest.

Mike Williams: That’s a great way to wrap things up here, Fabrice. I know we’re out of time here, but, I really appreciate taking the time to join us here today for the follow up group chat.

And I know we really packed it in and, some of the topics and questions are of course, like really nuanced, and it depends right? But I think this is great and it’s gonna be super helpful for everyone. And I had one last question and this is my typical kind of parting our closing question, so I’ll slightly reframe it this time, which is that if you could go right back to, pre 2026 and this kind of new era that we’re in with AI, what would you tell yourself about a marketplace specifically?

Fabrice Grinda: Honestly, if there’s one type of business that I think is gonna be the most insulated from the AI revolutionist marketplaces. And by the way if there’s a category where I think people are still underestimating how big it could be, it’s marketplaces. Like right now, yes in our consumer lives, we’re actually very well served, between Lime and Instacart and Amazon and eBay and Uber and Airbnb. We have marketplaces covering 15, 20, 20 5% of commerce in each of the major categories.

But then you’re looking B2B and. And nothing’s been done. We’re like sub 1% penetration in most categories. Definitely sub 5% in these categories that are trillions and trillions of dollars. And where I think because you could do much more work and because they’re also slow moving, et cetera, you’re gonna have amazing defensability. So I still think marketplaces are the way to go. They’re asset light, they’re gonna, you’re gonna build a boat, they’re gonna be defensible.

Just use all the tools don’t index yourself, but don’t let them use you for trading data to be clear. And yeah you’re still the right place, doing the right thing.

Mike Williams: That’s great. That’s great parting advice, and you got lots of head nods with that. We once again, really appreciate, taking time and join us and of course, I share a link to your blog and to FJ so everyone can keep updating and follow along if they don’t and reach out as well.

So really appreciate, joining us and thanks everyone for the great questions today.