Building defensible marketplaces: insights on network effects and AI - Pete Flint (NFX)

Please note: The transcript is automated, meaning that there will be mistakes.
Sjoerd Handgraaf: Hi, Pete. Welcome to the show.

Pete Flint: Thanks so much. Great to be here.

Sjoerd Handgraaf: Yeah. Really, I'm very happy you, took the time to make it. I've been following NFX for a super long time and all the work that that NFX has done around, marketplaces, network effects, etcetera. But before we go very deep down the NFX and marketplace rabbit hole, could you tell us a little bit about, what did you do before you became a marketplace investor?

Pete Flint: Sure. So I guess I've been involved in marketplaces for, oh gosh, several decades now. So I guess I've had sort of three jobs. So the first was was part of the founding team at lastminute.com, which is a big web one point o European travel marketplace.

Sjoerd Handgraaf: Mhmm.

Pete Flint: And so that was just an epic experience in building a marketplace business. And then in 2,003, I moved to The US US to originally attend graduate school at Stanford and then started Trulia while I was there. Mhmm. And Trulia is a big online real estate marketplace, and we ultimately went public and then merged with Zillow. And then I teamed up with James and Gigi who also had a passion for network effects and

Sjoerd Handgraaf: Mhmm.

Pete Flint: All things defensibility, and we started NFX. So, you know, there's a sort of there's magic around network effects that Yeah. Once you kind of, like, get the bug Yeah. You just you just wanna do everything network effects.

Sjoerd Handgraaf: Yeah. No. That is true. I I've talked to several investors now. And, for example, like, Jose Moran, who's also been the marketplace entrepreneur, and and we see it also, you know, like, the company that produces this podcast here at Tribe, we make marketplace software for founders.

And and we see people coming back, you know, idea one. And I was like, well, you know, and then I went here, and I saw another marketplace model in the making. So let's try this. No. I can relate.

So NFX, probably the most network effects specialized, fund, that there is. Is there a typical NFX investment, type that you look for? Like, are there specifics areas that you look for?

Pete Flint: Yeah. I'd say we are a stage specific sector agnostic firm. So, like, what we love to do as a firm is just to be the early, you know, often the first institutional investor with a company. You know, we're all founders and operators, and so we've been in the founders shoes at this sort of really ambiguous time at the beginning. You know, but in term and what we look for is, you know, companies that will be around in ten, twenty, thirty, fifty years time because they have this inherent defensibility.

The defensibility the best form of defensibility we've seen is through network effects. It's the most efficient. It's the most defensible. You know, businesses that have this property just become super resilient. And, you know, I think when you think of network effects, you think marketplaces Mhmm.

And we're super active in that space. But we also just think about network effects in in many different constructs. So whether that's bio platforms, we Omri leads up Mhmm. Biopractice. And so we do a lot in longevity.

And so they may not start with a a network effect, but they build in network effects over time. There's a bunch of AI SaaS companies that way back, which, you know, may not be obvious network effects beginning, but we know that they've got this really incredible v one product that gets a lot of traction. And to stop the myriad of competitors coming off of them and eroding margins and adding more competition, then they need to add a network effect over time. So we love network effects at the beginning. It's often that they add them later.

Mhmm. But we, you know, we off always have a thesis around what is the defensibility, what's the network effect for the company.

Sjoerd Handgraaf: Yeah. Because that's actually interesting. And maybe, like, the ace is it possible to talk about the AI SaaS one as an example? Because often, like, the picture due to outside is like, oh, well, they had some traction and that is what the investor looks for. And then, obviously, this is a thing.

And now you you're basically saying that, well, there is a v one. There might not be network effects here. We expect them to do like, how do you evaluate an opportunity? And and sort of the non obvious one that you mentioned is maybe a really nice one to sort of go through that state, like, through that evaluation.

Pete Flint: Sure. I think, you know, it's look, every company is different, but I think the stuff that, you know, we would look for is, you know, what is the market size? How defensible is the either insight or the capability they've built.

Sjoerd Handgraaf: Mhmm.

Pete Flint: For instance, like, you know, you'd see a bunch of companies that are building AI tools off public data.

Sjoerd Handgraaf: Mhmm.

Pete Flint: And that may be a great way to start because they're kind of pulling all this public data in

Sjoerd Handgraaf: Yeah.

Pete Flint: And they can make sense of that. But in unless they have more proprietary data over time, whether that's customer submissions or whether that's something else, then it's gonna be hard for them to have real defensibility. You know, we look for kind of willingness to pay. You know, I think there's a lot of momentum right now in you know, so an example would be that Mhmm. There are certain AI tools that are substituted labor.

Sjoerd Handgraaf: Mhmm.

Pete Flint: And where you've got kind of a labor constrained environment or kind of labor quality problem, then you're seeing a lot of velocity in those areas as well. So those are a couple of things. You know, you could sort of perceive investors to be sort of, you know, sort of singularly focused on kind of like one or two things.

Sjoerd Handgraaf: Mhmm.

Pete Flint: You know, if I was to kind of where investors are probably and this is generalization where there's companies that see a lot of interest is where you see very rapid traction and very rapid growth simply because investors are really struggling to predict the future. And so it could be a company that's in a really crowded space, really mediocre retention, tons of competition, unclear about market size. But if they are growing really quickly

Sjoerd Handgraaf: Uh-huh. Then I

Pete Flint: then investors are like, well, we don't really kinda know what the future is gonna hold, but we know this company is able to scale really, really quickly. So we're gonna Mhmm. Take a bet on it. And that you know, because they know there's gonna be category leaders, which we anoint in this AI phase, just like you see category leaders anointed in the mobile phase and the web one point o phase. Yeah.

And so investors are really backing up the truck to those companies.

Sjoerd Handgraaf: Yeah. Because I think the AI criticism that I hear a lot is that there is no moat. Right? It's like there is because it's all like, they're all LNMs and and but if I understand you're right, you you were saying that, well, there is maybe initially not, but just speaking in very general terms, like, as over time more and more custom data maybe can gets put into the model, it becomes defensible. Right?

Is that is that what you're saying? Am I understanding it right?

Pete Flint: Yeah. I think that's a key part of it for sure. I mean, there's obviously within AI, there's sort of different levels from chips to models to infra to apps. We're really focused on the app Mhmm. Layer, which we think there's gonna be hundreds, if not thousands, of really significant companies.

I think the I mean, where defensibility comes from absolutely is product experience driven by proprietary data or network effects. Very much a kind of embedding or a stickiness of these services just in the same way that the companies that first installed Salesforce are in. Everyone's trained on it. Everyone's using it. The data's there.

And Yeah. Sure, they could switch to something cheaper or, you know, with equivalent for Xfinity, but it's they're there and they're embedded. And so the workflow is meaningful because if you're a tax accountant or a lawyer, like, you go through the change management of installing this piece of software and it's pretty good, you're not gonna rip it out in three years. And then I think there there is gonna be this virtuous cycle of the more successful companies have more customers, and they're able to invest in product experience and marketing and scale and data features and and the ecosystem around it that there is very much gonna be sort of defensibility around that in terms of brand adoption. And some of this stuff is sensitive, and if you're replicating kind of, you know, human processes or dealing with proprietary data, You don't wanna just, like, you know, use any off the shelf, you know, the the cheapest open source thing that you can find.

There will be a ton of defensibility around these businesses.

Sjoerd Handgraaf: Yeah. That makes sense. That makes sense. If I can ask about this because this just came out, but you remember if you were one of the writers, but NFX just put out this new piece about AI agents. Could you tell a little bit about this, especially in the relationship to to marketplaces?

Pete Flint: Yeah. My colleague just put it out. I I have not had a chance to read it. I'm happy to kind of, like, how we think about.

Sjoerd Handgraaf: So, talking about AI. So so now we talked about AI businesses who sort of center around AI. There's also been a lot of talk about how AI will change existing marketplaces or add on to existing marketplace model? Like, what's your view on that?

Pete Flint: Yeah. I think there's you know, AI is both gonna be a disruptor to certain marketplaces, and it's also gonna be an enabler of other marketplaces. And then it's gonna create entirely new marketplaces that we haven't really kind of envisaged today. So, I mean, the disruption will happen where you have somewhat digital commodity services. So you could think of, companies like Fiverr Mhmm.

In one extent. And while they're probably pushing back on the disruption, it's sort of this inevitable there's a degree of disruption in the area or coding marketplaces Mhmm. Like Upwork. And so, you know, where really the AI substitutes the human in that regard. And then you've got, you know, perhaps the other extreme is like in the in here in San Francisco, Lyft, which is the kind of number two behind Uber in terms of ride share marketplaces, and has equivalent market share to Waymo.

And, you know, Waymo is, you know, is part robotics, part AI, but it clearly wouldn't exist without AI. And so it's like that's very disruptive, to portions of that marketplace. Not Mhmm. Perhaps not everything, but, like, portions of that marketplace. So that's where there's disruption now.

There's another piece which is enablement, which is look. When you think about the kind of marketplaces themselves, you've got supply acquisition, you've got matching, and you've got demand acquisition and retention. Yeah. And you can see how AI can do incremental or significant impact to kind of to each of those kind of three areas. So, you know, you can sort of either sort of create synthetic supply

Sjoerd Handgraaf: or you

Pete Flint: can engage supply more effectively or build AI SaaS tools that kind of, you know, create the tools that kind of keep supply engaging, create, you know, highly personalized matching experiences. Yeah. You can create, you know, similarly demand engagement and acquisition as well. So you've got you should be able to kind of deepen liquidity to the incumbents with these existing network effects, you know, should, in many situations, see AI as a accelerant to the business. And then finally, you've got, you know, AI which is gonna be create entirely new marketplaces.

And so it's

Sjoerd Handgraaf: Yeah.

Pete Flint: You know, we'll see kind of where that goes, but, like, you know, hugging faces is essentially a marketplace of models. You can see it agent to agent communication. There's gonna be, you know, platforms that facilitate kind of privacy and authentication across different services. And so that's gonna be entirely new companies are gonna be created out of this.

Sjoerd Handgraaf: Yeah. When you say those things, make me glad I'm not an investor because I find it really hard to sort of imagine. I would be invested because you said there's trouble envisioning the future because that is, you know, like, the marketplaces the AI marketplaces, we haven't thought up yet. I I find that really hard to just envision, like, to to to see where this would be going.

Pete Flint: There's certain markets which have historically been just sort of off limits because of their complexity or because of a degree of kind of market adoption. So and I think if you look at you speak to any CIO now, they're like their board is demanding what their AI strategy is.

Sjoerd Handgraaf: Yeah. Yeah.

Pete Flint: So the CIOs are like, okay. We you know, previously, we use pen and paper to do something or spreadsheets. And now we're like, okay. We're using AI to do things. So you've got this Trump's world right now, but, like, government procurement and government processes is, like, most sane founders would avoid that.

But you have the combination of AI plus Mhmm. The kind of regulatory environment where it's like, oh, this is game on. And there's a bunch of those industries which, you know, they should have digital platforms there, and AI will create this sort of momentum in those sectors that we haven't seen before. And that's for sure gonna happen.

Sjoerd Handgraaf: Yeah. I think that that is, like I think that's the case I can still sort of imagine, especially, like, a few of the investors I talked to have a sort of a b to b focus where you have indeed, like, these sort of very complex procurement processes, which are you know, that always have had some kind of human interface to just get all the information in the right order or to ask the right question. And I can definitely see that sort of empowerment or even maybe even a completely new model of doing those things through some AI agent if that is the right term. I can see that. But some of the, you know, AI marketplaces that are, well, that are creating data that wasn't there before and somehow make it usable.

But, yeah, speaking of things that are coming, if we leave the AI on the side for a moment, I'm wary of using the term trends. But do you see in the portfolio or in recent slide decks that are being sent to you, do you see some other interesting developments in marketplaces or or two sided platforms or network effects businesses?

Pete Flint: The AI thing is prevalent in in really every kind of software company that's starting over the last couple of years right now and even the scale one. So I think it is a really big story for every incumbent or or new marketplace. You know, I think there is you know, you've seen this, perhaps evolution of marketplaces to go from pure marketplace to vertical marketplaces to fintech enabled marketplaces. So the sort of fintech component is is very much there. And then adding the kind of services component on top of that

Sjoerd Handgraaf: Mhmm.

Pete Flint: In terms of sort of agentic experiences to kind of create these more kind of full stack. Marketplace is very much there. And we're seeing a lot of marketplaces that are really coming up through this SaaS tool approach. Yeah. They're kind of building they're solving a sort of a niche problem, which is interesting and enough to get sort of traction.

And then that creates a kind of a wedge tool that can create an entirely new marketplaces. So my partner James wrote a piece, gosh, maybe a decade ago, which was called Market Networks, which was really the identification that there's, you know, significant value to be created, particularly on the b to b side where you've got market networks, which really combine great workflow, so typical sales tool, combined with a marketplace where you can participate in transactions and, you know, incremental revenue and utility, and then the identity layer. So how do you make it a multiplayer experience? And you'd you know, you actually see these sort of components in various different weightings across, like, so many SaaS tools. Right?

Yeah. That sort of identity layer is sort of prevalent in so many things that are kinda happening right now. And I think the marketplace perhaps started off as a sort of transactional marketplace, and I think it's gonna get That won't just be like, okay. Where can I find suppliers or where can I No? Purchase goods.

But I think you'll start to see where can I get services? Where are the agentic services I can get? And so if you're in, finance space, okay, like, how do I get rather than introduce me to an accountant, okay, who's gonna file my taxes or through a kind of AI agentic experience. And so there's a sort of high degree of complexity around this, which may be intimidating for some startups. But the cost of building software has gone down so quickly

Sjoerd Handgraaf: Yeah.

Pete Flint: That you're seeing small startups build, you know, these fully functioning experiences with agentic type tools or services or marketplaces and, you know, a high usability kind of workflow solution.

Sjoerd Handgraaf: Mhmm.

Pete Flint: And identity layer wrapped in. And you can see how this once this thing kinda gets established, you're like, woah. There's so much defensibility

Sjoerd Handgraaf: Yeah.

Pete Flint: You know, and also virality within the organization and internally and potentially externally that you can see a lot of adoption defensibility.

Sjoerd Handgraaf: Does that also so so that changes, you know, like, you you mentioned now that, like, you're bringing these things together, it'll change all the things, like, individually. Do you see those changes also if you think, for example, about, like, revenue models, you know, like, coming back to what you said earlier, like, there used to be a demand, supply, transaction, and then, you know, the the monetization happened on the transaction side. Now, of course, with all these added services, like, are you seeing any new developments in in that part? So so the the revenue collection or the the business model?

Pete Flint: I think marketplaces that are purely providing matching are generally insufficient to provide enough kind of utility because matching is sort of a discovery problem. You know, I'd say the big opportunities that have been kind of built over the last twenty years have already been taken, many of them, for simple matching. But it's actually a lot of the other, whether that's trust and safety, whether that is, you know, tools on the supply side to keep them deeply engaged, tools on the demand side to kind of really provide personalized experiences. And so, you know, simple matching kind of in the vein of kind of eBay is sort of, you know, eroded in these kind of, like, highly focused and great experiences and vertical solutions. It doesn't mean your business model can only be transactions, but I think there's, entrepreneurs are seeing tons of other opportunities.

And the SaaS business model itself is, you know, an incredible business model as well. And so we're seeing a lot of SaaS enabled marketplaces naturally. Yeah. And that's been the case for for many years. Obviously, a lot of secondary revenue streams in the form of fintech where companies are providing, you know, whether that's payments or financing

Sjoerd Handgraaf: Mhmm.

Pete Flint: Experiences on top of that. And then, you know, the the agentic business model is really the sort of pay for the outcome or pay for the work, which is transactional related, but it's the margins on those businesses can be very significant. And you could debate whether they're, you know, they're marketplaces themselves, but really there's a ton of money to be made for startups which are providing there are labor components that would be traditionally done associated with the transaction. Mhmm. By automating that process, the market is placed and able to tap into that particular transaction.

Sjoerd Handgraaf: Yeah. Could you maybe give an example of that? Like, when you say, like, an agentic workflow, like, do you have a a real real world example, like, from maybe from the portfolio or from a known a better known example?

Pete Flint: I mean, I'm just taking some ideas from the real estate space. And so you can see in, you know, real estate transactions that you see you know, there's a transaction coordination piece, which is kind of managing the documents and Mhmm. Getting everyone online, which is the outsource to a human, that would be thousands of dollars. We can also see a case in kind of architecture where similar kind of like perhaps consulting or kind of permitting process would be replaced by agents who would provide a bunch of those those services. And and if those are kind of components within existing marketplace, we kind of often see there's sort of two modes.

One is that they discount back to the consumer or the user where they say, okay, we're gonna offer this product, this service at a much cheaper price and that's

Sjoerd Handgraaf: you know,

Pete Flint: disruptive experience. Or they the other approach is like, okay, we're gonna keep the margin and then triple down on sales and marketing or other product experience areas. It often depends on the willingness to pay. There are many and I think a lot of the a lot of these sort of traditional functions which were done by humans is actually quite high willingness to pay because, you know, less commoditized, they're more complex, and so Yeah. People are willing to pay.

And so, you know, the margins on some of these AI, you know, add ons to marketplaces or pure air companies Yeah. Can be really significant. It's you know, it could debate how defensible they are, but I think if they're associated with a highly defensible marketplace, they're they're gonna be pretty defensible.

Sjoerd Handgraaf: Yeah. You mean, like, defensible as in, like, oh, over time, of course, their value will diminish. Right? Like, as they become new table stakes to the market. Maybe, speaking, about, like, over time, like, you mentioned, you've been in marketplaces over some decade.

The audience of this podcast is a lot like aspiring marketplace founders, early stage marketplace founders. Are there still some things that you feel like that you learned, for example, in your experience that you see and that you regularly find yourself, like, advising founders? You know, like, this is actually like, in a two sided platform, this is what you should do or this is what you absolutely should not do.

Pete Flint: There's a bunch of different things. I think number one is, like, find what is the really hard side of the marketplace and figure out how to solve that first. You know, usually it's like sure supply and demand are hard but one side is particularly hard and focusing in on like how you can solve that side of the marketplace first. It's gonna be critical. You know, there's a lot of marketplace founders which I think are not really attuned to the kind of, like, the level of sort of fragmentation within a category.

And fragmentation can be, like, super scary, but it's necessary to kind of to build a really compelling marketplace. And so, you know, you see that just it's sort of supply side homogeneity. So, like, you look at, you know, Airbnb versus Uber. And so looking for that fragmentation is really critical. And then often finding a really clever hack to aggregate supply and or demand.

Sjoerd Handgraaf: Mhmm. And

Pete Flint: there's, you know and I think there's always a hack whether that's like companies were building in the mobile era that they could attract supply and demand or demand in a unique way. Yeah. Whether that was sort of the Facebook social network era, how can you kind of attract supply and demand in a clever cost effective way? Yeah. And then AI creates, like, such a myriad of solutions.

So it's quite possible that there are marketplaces that just don't exist today because it's been too hard to get liquidity to aggregate the supply demand. But if you can identify ways to aggregate using some new technology or some new economic trend then that can be the that can can be super critical. Perhaps the last thing is like, you know, I I sort of for better for worse, I kind of my experience in online travel was during September and my experience in online real estate was in 02/2008 in the global financial crisis, which which were sort of awful at the time. But that change created the opportunity. Because if you take a step back, what our marketplaces do, they connect supply and demand.

Where you have a mismatch in supply demand through you know, because of some sort of economic change

Sjoerd Handgraaf: Uh-huh.

Pete Flint: Or sort of some environmental or regulatory change, then that absolutely creates an opportunity for marketplace is to provide much more utility over the incumbency. You can see folks moving over from perhaps an incumbent marketplace to a new marketplace if they're able to really navigate some of this economic or cultural change. You know, some founders, they're faced with a a sea change in their market. Mhmm. And that's exactly the opportunity to strike.

And often those sea changes are kind of are in periods of distress

Sjoerd Handgraaf: Mhmm.

Pete Flint: In a market. And you may think you're kind of like the market opportunity is much less interesting today, but that's exactly the time that you should should be moving forward, moving quickly, trying to gain market share when there's all this, like, volatility going on. And then come as typically, you know, just unable to react with the speed, the new demands on the supply and demand side.

Sjoerd Handgraaf: Alright. Hey. I think that's a great way to end this podcast. Thanks a lot for your time, Pete. Where can people check out Pete and or NFX?

Pete Flint: Thanks. It's great chatting. So nfx.com is a website. You can find more info about me, the team, and then sign up for email newsletter that's packed full of cover by aspiring and existing entrepreneurs.

Sjoerd Handgraaf: Alright. Thanks a lot for your time.

Pete Flint: My pleasure.