How to build and scale winning marketplaces | Casey Winters (Eventbrite, Pinterest, Grubhub)

Brett: Maybe we could start by talking about some of your thinking on what are the requirements to make a marketplace work?

Casey: There's a bunch of different factors that can help or hurt your ability to build a marketplace successfully, but generally, the first couple you think about are, is there supply fragmentation, meaning are there lots of different suppliers that, you know, a buyer might be interested in purchasing from, if there's only a few, like say in flights, then it's hard to have a, like a high enough take rate to build a successful marketplace.

So you see with the expedience of the world, the bookings of the world, they may actually make most of their money off hotels where there's a lot more variety and a lot more fragmentation. So that's generally the first thing you look at. The second thing you're looking at is either high frequency of need on the demand side or promiscuity for lack of a better word on the demand side where people want to engage with different suppliers over time.

So for Grubhub, you know, where I spent a long time, people don't want to order the same from the same restaurant every day or every week. They want to try Thai and Chinese and pizza. so there's a natural usage of the marketplace across different suppliers. If you're only going to work with one supplier, you know, forever, then the ability for the marketplace to add value and keep those transactions on your platform goes down dramatically.

So those are kind of like the three high levels but for most of them, there's like, okay, well, if that's not going to work, then you have to compensate with some other. if you're going to be low frequency, then you need to be high AOV. And if you can't be high AOV, then you need to be high take rate on the low frequency.

And if you can't get any of those to work, then chances are that a marketplace is not going to be an optimal model. I think the thing that gets talked the least about in marketplaces, that, you know, maybe I'm biased as a growth guy towards, but I think is extremely important is the sustainable advantage that you have in acquiring either supply or demand.

When you look at successful marketplaces, they tend to have a really scalable acquisition loop. So that can be sales on the supply side. but you want like a very quick payback period on it. I think with Grubhub, we had such a, pay performance model, that it was really easy to get restaurants to try, which means our cost of sales was like really low compared to enterprise sales.

and then a lot of times you're using some sort of unique data or content advantage you get from supply to attract the demand. You know, so SEO becomes a really common vector for that of, let me aggregate all the restaurant menu data and show it to people when they search for food delivery, right?

But sometimes it's more viral, right? So Faire is a marketplace that I work with. That's B2B it's between, brands and, boutiques that want to try out their products to sell. Part of their unique advantage is they were able to build a product where, brands can onboard their existing retailers for free and not pay any commission, but they can manage that workflow.

And then that allowed Faire to cross sell those boutiques to other brands, to cheapen the cost of acquisition. So a lot of what I'm looking for is what is that data advantage or that acquisition loop that feels proprietary? so that you're not going to just spend an incredible amount of money acquiring both supply and demand, we've saw companies during the zero interest rate environment, just spend a lot of money and that can work too. But, you know, generally, it's harder to convince investors in 2024 to just give you a billion dollars every quarter anymore. You you need something that's going to be a cheaper way to scale up liquidity in the marketplace.

Brett: On that last point, have you found that the best companies have figured that out very, very early on, even maybe before they started this idea of, compounding or competitive advantages that relates to acquisition on supplier demand, or is it generally figured out and bolted on mid flight?

Casey: Generally what I'm looking for as an angel investor to get involved as an advisor, someone that's figured that out and needs help thinking about scaling it, but they already have that core insight of why it's going to work and why it's going to be defensible, over time. You definitely see some people figure it out a little bit later.

That's okay too. But in today's environment, we're like, paid is pretty hard to get an edge on, any sort of edge you have will generally get competed away by competitors coming into the market. You're looking to see that someone's not going to rely on paid or sales for everything, unless they have way higher frequency, you know, way higher take rate, way higher AOV than whatever you see in the market. There needs to be some sort of either product quality lever that creates an edge on traditional channels or some sort of unique data advantage. And I think we're seeing founders across the board get a lot smarter about okay, when I deliver value to the customer, how is that going to create scalable growth loops to acquire more of them?

And not wait until I have retention to then figure out if I can find new customers. But like, baking both of those into the product market fit hypothesis. You're starting to see a lot more founders do that, not just in marketplaces, but across the board. And I think that's great to see because then as soon as the value gets unlocked, the scalable growth just immediately happens. 

Brett: Yeah, what are good examples of that? 

Casey: The challenge in growth for like all companies is while there are a lot of hacky things you can do at the early stage to grow, which I call like kindle strategies, you just trying to validate your product hypothesis, get an early users to test, There aren't that many that helped you become a billion dollar company.

It's basically sales, virality what I call like content loops, which is we distribute content directly to, you know, Google or to social networks to bring more people in and then those people create content, and then you redistribute that, or paid. And the challenge with paid is it always looks better yesterday than it does today unless you have a network effect business where the product quality is getting better, faster than the quality of the people you're bringing are getting worse right? Cause you always target the best users first in paid. So generally, it degrades in quality over time and eventually becomes unprofitable.

Sales is a little bit more stable in terms of the payback period than say paid, but it's never that cheap unless you just have like this sort of amazing product market fit that's flying off the shelves, like a figma or something in the product design community. Virality, is, you know, mainly how consumer social companies have, have grown, but you get like a really fast curve, but then it asymptotes generally more quickly because all the people that reject the invites the first few times, they're not gonna necessarily accept the value prop the seventh time, you need something else to hook them. So all of these channels have like their pros and cons or what I call like acquisition loops. They all have their pros and cons and generally at massive scale, you're combining a few of them together, but there's generally usually one that you can grow most quickly from at the early product market fit stages. And I like to see founders that have not only identified that, but spend a lot of time baking that into the core product.

Brett: Everybody has reference points of how Airbnb maybe grew or Yelp grew or sort of through SEO. What are examples of like emerging marketplaces where you met or watched what they were doing in the past year or two that's like a fresh example that brings it to life?

Casey: I hope I don't get in trouble for saying some of these. A company I angel invested in is called power. It's a clinical trials marketplace. Part of, what they've done is, you know, a lot of people are searching for alternative methods of care, due to chronic pain or things like that, and they're going to Google and they're searching these really detailed things and they're trying to find things that their doctors aren't even aware of that maybe are going to help, you know, cure them or, you know, mitigate their problems.

So they built just the most informative pages you could possibly build on those topics. A lot of those, treatments are in clinical trial stages, where the people that are running those experiments are trying to find diverse groups of people to actually get into their experiments and it's really difficult to find those people. So they're able to matchmake people that are looking for alternative forms of care with the latest and greatest clinical trials out there. And the pharmaceutical companies and the people running those experiments are desperate for, that sort of audience, which is just really hard to like find, it's needle out of haystack.

So you have a bunch of people, powers being able to raise their hand and say, like, I'm actually looking for this. So that's, working, um, really well. I have a company, in the SaaS space I advise called Fermat Commerce, and, they recently raised their Series A. What they basically do is they take every Facebook ad that you run and create a custom landing page that's going to convert probably 20 percent better than whatever landing page you have.

And they run a ton of experiments to conversion optimize for all of these different clients. Part of how they got traction, mostly selling to direct consumer e commerce companies is they found these nodes in the DTC e-com community that know all the other founders, and they convince those people that it was a great product.

And so it's kind of like almost becomes a B2B influencer model where these nodes in the network go tell everyone else that you should really be spending time meeting with Fremat and trying it out because it works so well. So that's kind of like an emerging channel you're seeing. It's still a lead gen model into sales, of course but the way you're getting the leads feels very different than say, you know, 10 years ago. So those are a couple of examples in, in different, uh, business models, where they're thinking a lot about distribution out of the early gate and unlocking something that looks a little bit different than we've seen before.

Brett: Let's say I came to you and I said, Oh, I'm working on a new idea and I want to create a marketplace for tailors. You know, it's such a pain to go to the local cleaner and get your, pants shortened. I guess there's a startup that does tailor? 

Casey: Uh, no, I just always put Zip in front of any startup idea just for fun.

Brett: Oh, I see because there's the company Zip. And there's ZipRecruiter 

Casey: So I Zip Tailor. 

Brett: Zip Tailor. We're just starting to talk to potential customers and just think about whether this is a thing. Spend more time explaining, like if you were talking to that founder, how do they explore all the potential directions that they could potentially think about building this compounding advantage either on supply or demand acquisition?

Casey: There's basically two approaches you can take in the early days, which is you can say I can get supply and supply can bring me demand, which is not really a marketplace but maybe you can use that to, to scale demand later. Or there's kind of, how do I think about, I'm bringing tailors new business they otherwise would not have found so that they're making more money. But in order to do that, I need to like have tailors to send to people, you know, when they come to my website or app or whatever.

Brett: You said the first didn't fit your definition of a marketplace because?

Casey: Oh yeah. So if supply brings all the demands, then really what you are is just fulfillment for their own customer base. You're just a tool. You still work with buyers and sellers, but the difference between a marketplace and what I would call like a SaaS network is that a marketplace is primary value prop to the supply is that you bring the demand. Both of these models can work. Obviously Square is a very large company and it doesn't bring the demand, the consumer is not aware that they're purchasing from Shopify. Most of the time now it's Shop Pay. That's starting to change, but for most of their history, you know, it's just a website and, Shopify is white label, right?

Square was always like, you're using a Square terminal and all this stuff, I didn't bring the customer. It's the customer is already there. So yeah, that's like a SaaS network. It's fulfillment, generally lower take rate. If you have a take rate model, it's gonna be more like 5 percent versus 15%.

And then of course, you're paying the payment processing. It's harder to build a massive scale business. so you see fewer venture wins in that category. Whereas if you own demand, which is generally the most important goal of every marketplace, then you're going to see a lot more, you know, venture scale wins.

So if you're looking to actually build a true marketplace, because I usually have to explain this differentiation to founders, they don't know how big a difference in the scale of company they can build between the two is, okay, well. How do I skillfully acquire tailors? What are my options?

can I use the tailors to bring some of their supply online, some of their demand online, but then also how do I find people that are looking for tailoring services? How are they currently serviced? How do I create some experience that's 10 X better. Uber was just so much better than a taxi, right?

Airbnb was in a cooler part of town and cheaper than a hotel, how are you going to get something that's better? Maybe it's delivery, the tailor comes to you, but then, you know, how do we make that worth the tailor's time? Is there higher price we can charge that we can give more money to the tailor to compensate for travel?

We need to think through what the core product value is and how the economics will work. And then does the money you make allow you to more profitably acquire tailors or to acquire, people who want to get tailoring, or is there some way that you can dramatically drop the costs? I know there was like a laundry marketplace I was talking to recently And they have like some technology that allows them to use the laundry machines in like all these big properties that have a, you know, a bunch on the first floor and that allows them to, you know, more cheaply handle washing than some of these existing laundry marketplaces that have failed scale.

Right? So that's going building some hardware, right? Which is a bigger investment. But a lot of times you're trying to think about Hey, how does the marketplace really add a ton of value? Sometimes it's just trust is all you need. Right? I need a way to rate buyers and sellers. And then I know I'm going to get the best tailor.

And that's enough. Sometimes that's not enough because the tailor doesn't want to come visit your home, or they're not in a good part of town, so it's hard to get a bunch of people to visit them, and they need to think about different, machinations of the product that are going to enable that experience that's going to attract both of them, and then, what networks you can find them, cheap enough to generate a return.

Brett: In the context of selling a widget, there's a lot of thinking and best practice around engage in customers, prototype, bring that prototype to customers, iterate, et cetera, et cetera. how does that map onto what really high quality work looks like in the context of building a marketplace from scratch? What's the scaffolding that they should think about building the company around or doing that early work around?

Casey: Well, I think it's similar to enterprise in a lot of ways, but one thing that's different is in, you know, SaaS, right? You're like, here's the software and the value is what the software does. It saves me time. It makes me money, whatever. In a marketplace it's a bit different because the only thing supply cares about is more demand. And the only thing demand generally cares about is selection of supply, more selection than they previously knew is available, right? And high quality, of course. Marketplaces are not really usually winning on the software itself. Only if the software helps you find more selection on the demand side, or, brings you more money on the supply side, and those can be interchanged depending on the exact marketplace we're talking about. You can do things incredibly manually, as long as supply is getting more business, and as long as demand is seeing more stuff, right? 

So, Grubhub, we just collected all the menus that we could find of every restaurant and put it online and got it indexed on Google and that drove demand. Was it a good product? No, like it was an embarrassingly bad product, but it showed that there were more restaurants that delivered than people thought. So that brought consumers and then once we had consumers, we then went to restaurants and said, Hey, if you enable online ordering, you'll show up ahead of all your competitors. and, oh, if you don't get any orders from us, you don't pay us anything. So it was like risk free return for them. The product itself was like bad for a long time, but you know, the restaurants were getting more demand and the people who are in food were seeing way more options than they had in the past, right?

Because Yelp wasn't really good at telling you who actually delivered to you. They didn't know that. You know, it's the same as like enterprise software. A lot of times you're just kind of manually doing consulting work until you figure out like the scalable piece of software, just the customer know that maybe they do, maybe they don't, maybe they don't really care.

Right? But it's the same in mark places. I mean, Tony at DoorDash was delivering the meals himself for like the first year, right? Cause he didn't know how to acquire drivers yet. and he didn't want to work on that until he validated that there was a real product there. So there's a lot of manual, work up front it's just more so in marketplaces. The only thing that matters is did I get enough supply? Did I get enough demand? Are they converting with each other, at a high enough rates? and then are we getting them enough value that they're coming back right? So you have all these slightly bespoke product market fit metrics, which is usually called liquidity in marketplaces. It's usually around conversion and then are there enough conversions where supply is going to stick around and it's worth it to them? And are there enough selections of high enough quality that the demand comes back and purchases, you know, again and again. And then once you get that, you're like, okay, all we do is bring in more demand and bring in more supply and that's kind of all you think about for a long period of time, right? You don't need to think about new products. You just kind of scale that cross site network effect, is wide and large as possible. So you want founders to be aware that that's the prize and then be willing to do a bunch of non scalable stuff to unlock that initial liquidity. I remember shortly after I left Grubhub, which was the end of 2013, I was talking to, the marketing team they were like, ah, man, Postmates and DoorDash are spending so much money on AdWords, it's not sustainable, you know, we've historically had six month payback periods, we're thinking about raising them to a year just so we can stay competitive. What they're doing doesn't make any sense. And I had to like remind them, they're not even playing the same game you're playing. They're playing a liquidity game because they have drivers sitting idle and you're playing a, we're about to go public and we want to show really profitable LTV they don't even have a concept of LTV yet.

They're like, I got restaurants and drivers sitting idle. I need to get them food or, I need to get their food out to someone or else I just, I'm wasting even more money. So you're trying to get founders at the early stages of marketplace to understand what game are you playing? We can do non scalable stuff to get to liquidity, assuming it opens up more scalable practices, you know, post that.

Brett: But so does that mean there's not as much customer development work to do on this on the supply side If I bring you more customers, will that make you happy? That's basically the job to be done. And most of the work is figuring out what has to happen to generate demand? Because, you know, that if I'm bringing you a marginal customer in most cases, maybe the top end of any marketplace, the number 1 10 or 100 don't need more customers.

The top 20 restaurants in San Francisco, if you were to go to them and say, do you want more customers? They say, no, we're basically filled. But once you get outside of that, is there a little work to be done on the supply side in that thinking?

Casey: Well, I think what's interesting about that example is for OpenTable, that might be true, but for Grubhub, that's basically never true because, the kitchen is essentially always an underutilized fixed asset. They can always pump out more food, 99. 9 percent of restaurants can pump out more food.

It's the front of the house that gets constrained because they're filled with reservations. They think of that as really high margin revenue because they're going to pay the rent no matter what, right? So if they can pump out more orders and pay the same amount of rent, yeah, there's food costs and stuff, but that's generally a smaller part of, you know, their expenses.

even the top end are like, yeah, I'll take more catering or delivery orders for sure. I think we were seeing with the first generation of marketplaces was basically exactly as you said, all I care about is demand so a lot of the product development is iterating on the sales value props that I can convince you I can bring demands.

What you're seeing with little, uh, A little bit more of the next generation of marketplaces is it requires more products to be in a position to drive demand scalably, maybe you need to build out some workflow products, or maybe you need to build out, some sort of free management software to kind of unlock the ability to drive demand. And that was the case with Faire of, we kind of need you to start using this free workflow tool so that we can bring more demand and cross sell it so that everyone gets, more business and that business moves online instead of, you know, at trade shows and things like that.

And you see that more commonly in a B2B marketplaces where it's like, ah, well, I kind of need the bundle demand with some tools so that you can actually act on the demand safely. So a lot of the customer development is like, yes, I know demand is the thing. What hoops do I need to jump through so that I can actually scale a way to deliver demand?

And that might be actually building some more real software, but you know, generally the top level problem is customer numbers go up, supply's happy.

Brett: The last point you made is an interesting one. Is there at times a risk that you can get too focused on the supply side tool that ultimately, if it's a domino rally, where the first domino that has to go is I have to build the tool that you find value in as a standalone product and then I bring demand, that you at times run the risk of spending so much time on this tool,

you wake up 12 months later and you built maybe an interesting tool, but you've done no de risking on the demand side and you end up with a relatively uninteresting kind of tool business, not a true marketplace.

Casey: I think you are absolutely correct. I generally advise founders against marketplaces phase two of the company. So there are a lot, a bunch of people were like, ah, SaaS to marketplace transitions, that's the model. And I'm like, Name an example in the last 10 years where that's actually worked because running a SaaS business is not simple and if you build, you know, a SaaS tool that helps aggregate supply, cool.

I get how the next possible phase is now go figure out how to acquire demand, but you've now built a company around people just servicing supply. You've built a company around people that know how to build SaaS tools, not how to build demand. and generally making that transition is very difficult.

So part of my job at Eventbrite was trying to make that transition. Eventbrite was driving some demand, but the majority of the demand was coming from the event creators going and doing their own marketing, right. To bring people to Eventbrite to transact. And we had to do a lot of culture shifting, to even have a chance of building more of a marketplace product. And I would say still, we're only partially there. I spent three and a half years as chief product officer. There were some distractions, like a global pandemic, which we could talk about, but, it takes a lot of effort to figure out how to reorient the core customer from supply to demand. 

Brett: It's very counterintuitive. Eventbrite is a good example. You would say, okay, I have all of these supply objects, which is so hard to get all these events in all sorts of different geographies. In this case, let's just say across the U. S. And not focus internationally. It should be easy to sort of transition all those supply objects that I'm servicing into a true marketplace. But to the point that you just made it, definitely was not the case. 

Casey: there's a few things that happen that make this hard in practice. there is the SaaS business you're running, which in the case of, Eventbrite was making hundreds of millions of dollars in revenue at that point, about to go public, right? there's, I can spend a bunch of time going 0 to 1, building out a consumer destination that shows all of our cool stuff, integrating that into SEO, getting really good at emails and notifications, and, you know, get 5 percent more orders short term, or I can make this tiny shift In a SaaS business and it will show up as higher revenue short term. So it's, a marshmallow test, right? are you going to stay, and eat that marshmallow in front of you or wait and get the second marshmallow later on, it's hard. Even if the founders will pass the marshmallow test, it'll be a lot hard for all the way down the organization to pass the marshmallow test. 

So I think that's one two. The culture and the type of talent you bring in to scale a SaaS business is not the same type of talent that builds a marketplace business, the types of things they know how to build, the types of customer development they're used to doing are sometimes

Eventbrite to building a marketplace.

So. What we did at Eventbrite for a few years was we actually created like a separate business unit for the consumer team so they would not get sucked into the day to day trials and tribulations of the SaaS business. we had another variant to this problem at Pinterest, which was, Pinterest got really big in the U.

S. very quickly. it was always easy to run an experiment to move U. S. growth 1%, 2%. Rather than get Brazil from zero to one, So that's what everyone was doing and international was going nowhere. So, you know, Ben, our CEO was basically like, I don't care. All of the metrics for every team at the company only need to be gold on the ability to impact these five non U S countries. Yes, the U. S. traffic growth will go down. I don't care. I only care about growing these five countries. And of course, if we work in those five countries, we can scale it to more. And that actually did work, but it takes this really forcing function from the top to say, like, I don't care what you've been working on.

The thing you've been working on is actually going to stop growing or grow a lot slower because we need to make this bigger, zero to one investment. That's a simpler example than Eventbrite, because it's still the same product, we're just expanding it geographically. but you run into kind of, yeah, short term optimization versus long term step changes.

And it's really hard to get a team working on a product that has product market fit to then go shift to something else that's not going to show any return. And a lot of times they just don't have the skills, right? They have scaling skills, they know how to build features, they know how to do growth. They don't know how to build a product from scratch, right? So, there's a lot of these different factors that I've learned some the hard way, around the Eventbrite transition. And you know, Eventbrite's basically gone from driving zero demand to like 30 percent of the demand. I don't think cross site network effects really kick in until you're like past 50 percent of the demand, probably even more than that. So I'm in price on this journey and it's like slowly hacking away at becoming more of a true marketplace. but it takes time and you can see even today, cause I still keep in touch with the team. They're still trying to build some more consumer oriented instincts. because just culturally it's very different. 

Brett: The point that you were just making is that it's surprisingly hard to get someone that came to sign up for the local baking class to stick around and sign up for the cycling tournament or something?

Casey: while discovery of events is a real problem, being able to understand people who have come to you a couple times a year and understand what all their interests are and how to recommend them great content when all the content is ephemeral, The events happen and then they go away. It's not like a restaurant that's there, hopefully for a long period of time. And, you know, it's similar to this other restaurant. you have kind of a femoral inventory, generally not a lot of data on the consumer side as to what consumer preferences are because historically, you haven't been collecting it. so then your ability to match. And drive true discovery is initially very bad. And then you got to build that skillset generally from scratch, hire a search team, hire ML teams, and then they're going to tell you, I don't have the data on the consumer side to train a great model. So then you need to figure out how you get more engagement from consumers.

And, you know, then price at the scale where it's got over a hundred million people have bought tickets. It starts to accumulate the data in a very, very slow way. And it takes a long time to figure out how to build the skills to use it effectively.

Whereas a marketplace that started from day one, that would be a skill they're building essentially from day one of starting the company. 

Brett: A couple things to bounce around on. One is you gave this really interesting sort of anecdote when you were leaving Grubhub. And I think you said DoorDash was coming up at the time, and or maybe Uber Eats or something 

Casey: It was Postmates and DoorDash at the time. Uber Eats was a little bit later. 

Brett: They essentially were playing a different game.

I'm really interested, knowing what you know now, what was the correct thing that Grubhub should have done in that moment?

Casey: let's talk about what it means for them playing a different game. Grubhub was an asset light marketplace where the restaurants were in charge of doing their own delivery, and we basically had every restaurant across most of the U. S. that did their own delivery. DoorDash was building out its own delivery network to enable any restaurant to be a delivery restaurant, and that meant that it was not an asset light model at all, because in addition to acquiring restaurants into acquiring consumers want to order food, you also had to acquire the drivers that we're going to deliver back and forth early on.

The economics of that model were incredibly negative, which meant DoorDash had to raise billions of dollars of capital. So you're a company Grubhub that's raised 80 million dollars before you go public. Pretty asset light compared to most, you know, IPOs, then you go public and you're telling the public markets, you're going to be this extremely profitable high growth marketplace. And then this competitor comes along, that's losing 2 per order and raising hundreds of millions of dollars fairly quickly in Silicon Valley. And, you know, to Silicon Valley investor like Grubhub's validated that it's a big market. So you can, you know, bet bigger. and of course capital was pretty cheap during that period of time. So you're like, okay, they're building a delivery network. They're not just matching restaurants that already do delivery with consumers who are ordering food, they're basically spending their way to growth. Whereas, you know, Grubhub was primarily growing through SEO. And whenever we typed in food delivery, we would show up number one in any sort of context. So I think, you can't really go to the public markets and say like, Hey, we're going to adopt this incredibly high operational model. We have no skill in, that's going to burn hundreds of millions, if not billions of dollars after we just got you to buy a stock, convincing you that we were going to be an extremely profitable asset, marketplace. 

Brett: reminds me slightly a little bit of what was happening with Zillow and Opendoor when they were trying to do something.

Casey: Totally. Right. So in both cases, both Zillow and Grubhub much later, after those companies got a lot more traction, we're like, Oh crap, we're going to have to do this. Right. So Grubhub did eventually build a delivery network. Zillow tried and failed to replicate the Opendoor model, but they brought Rich Barton back to try to build that because you kind of can't change the culture that dramatically without the founder coming back, I think, was part of the calculus there.

So Grubhub for a long period of time was like, They're losing multiple dollars per order. This is eventually going to go to zero, right? This makes no sense. Grubhub was like, let's wait it out. and then the pandemic hits and all the economics of DoorDash flip positive because everyone's all of a sudden ordering online and all the drivers are much more busy.

Which means all the latency in the model that was creating the losses went away. And at that point, Grubhub is basically cooked. And people have asked me like, okay, well, what do you do differently? how did Grubhub screw this up? And I'm like, Grubhub didn't really screw anything up.

the only thing they could have done was buy DoorDash as early as possible. And then basically let Tony run the company and build a delivery network. That's like the only thing you could have done. Sure you could have copied DoorDash more quickly. Probably wouldn't have mattered. I'm sure you could have gone into the suburbs more quickly.

Probably wouldn't have mattered. you built a company with a very different DNA and, you know, operationally, I just don't think there's a way that a public company that raised like 100 million in IPO is going to outcompete companies that have raised 4 billion each. I think in Uber and DoorDash like to crush you.

It's a classic disruptive, you know, scenario. So I think, they should acquire DoorDash relatively early. That was the clear winner in terms of building out the sophistication of the delivery network and then basically let DoorDash kind of end up running the company. I think that's the only play and GrubHub did have an M& A skill.

It had bought Seamless. It had bought a couple other competitors. It knew how to integrate companies successfully. So that wasn't risky. I just think the price never really lined up, especially considering how heavy the losses were at DoorDash. So it never made sense until it was the thing that really crushed the company.

Obviously Grubhub is still around, but I'm, you know, it got acquired, but I'm assuming if it was public, the market cap would be smaller than the acquisition price today.

Brett: What do you make of Uber versus Lyft in the United States? 

Casey: It's fascinating. Right. Because, Uber of course comes out first and it's only black cars. And I remember, you know, getting a survey on like, how often would I plan to use Uber? they didn't even have an option for how infrequent I wanted to pick. so I told them that cause I knew the GMs in Chicago at the time and they're like, well, what would make you use it more often?

And I'm like, it has to be a lot cheaper. there's good public transit in Chicago, right? I'm only going to use it when I'm stranded or the airport or whatever. and then Lyft comes along, and I'm in San Francisco, and I'm seeing all the pink mustaches. while Uber early on was optimizing for luxury as a form of trust, Lyft was optimizing for, like, friendly neighbor as, a form of trust.

There were basically the scariness of going into a stranger's car. So Uber early on was all about legitimacy. These are licensed drivers, it's a black car, all this kind of stuff. And Lyft was like, oh yeah, it's like your neighbor, and it's fun, and it's friendly, and it's safe. And then I think Uber was smart to say actually their positioning is not what matters, what just matters is the price.

So they copied that aggressively, only focused on price, didn't focus on all the brand things around trying to make it feel safe and fun, just focused on price. And then scaled operationally a lot better, than Lyft. I think they were much more ruthless, much more aggressive, and raised more capital more quickly, and just took their model and, and scaled it faster.

Now Uber obviously made, played an international game as well, which didn't work out super well for them. So there was a lot of distractions there, but even Lyft not playing the international still got outcompeted in the U S on the model they were first or debatably second with sidecar to launch because I just think largely Uber out executed the model. it's easy for us to say in retrospect, all this stuff you were doing around brand, why that was such a waste of time. why'd you bother? And then it's like, well, we didn't know in 2013 that people were this quickly going to get used to jumping in stranger's car. They thought they needed to invest a ton in brand to make it feel safe. And Uber proved that you didn't, you just need to make it cheap, So, like, I understood the bet that Lyft made, but I think they could have reacted to that bet being incorrect earlier. I also just don't know if you could compete with the founder ruthlessness of Travis and his team.

They're so aggressive, raised so much capital, got to most markets before Lyft, even though Lyft launched this model earlier in San Francisco. It's a tough one, right? I mean, Lyft is, you know, struggled a bit, you know, on the public markets. and then the other thing that happened is they were diversified by the time the pandemic came.

So, they launched Uber eats and look the early Uber eats model. I don't know if you remember it. It was not good. we've got a bunch of sandwiches in the trunk that we'll deliver to you in 10 minutes. It was copying like the Sprig SpoonRocket model, which I think is correct in the environment they were in, but maybe not would be correct today, which is they basically were like in Atlanta. We're going to try convenience delivery. You know, in Chicago, we're going to try grocery delivery in Phoenix. We're going to try, you know, restaurant delivery and restaurant delivery works the best and it allowed them to use their supply that that was signing up for Uber, but maybe you didn't have a good enough car or maybe the person was too young. So allowed them to get some more efficiency on their supply acquisition as well and then they just scaled up the appropriate model after they tried like this more Sprig SpoonRocket thing, which didn't work.

Then they're like, art, we should just copy DoorDash and Postmates, right? So then they did that and that worked and then they just, they canceled all the rest of it and focused on the food delivery part, but that made them a little bit less fragile to the pandemic because while you know rides shrank during the pandemic food delivery demand went up and Lyft didn't really have that counterbalance only being focused on rides.

So, you know, some of that is like, you know, it's just hard to compete with a company with that type of culture that's raised that much capital. But Lyft had raised a decent amount of capital. I think it's largely boils down to some inappropriate focus on what builds the trust in the network early on.

Not as quickly reacting to that and then getting out executed.

Brett: it's impressive if you're in the shoes of Uber in the very early days. And when they pitched us, it was clearly, this is just sort of a new take on black cars and it was a vision for that. Like, it was very well thought through, but then Lyft came along. It would be very easy to not update your priors and say, no, no, no, our whole thing is trust through black cars and licensed drivers.

Casey: Maybe that was a bigger pivot for Grubhub to pull off versus Uber and Lyft but I think one thing marketplace founders need to be careful about is if you are aggregating supply and demand in the market and, someone else is aggregating the same supply and offering it at lower cost or with a higher frequency type of transaction you need to copy them immediately or you are going to get disrupted. Rover is a good example of this also. So Rover was a marketplace for dog boarding. It went public, but I think it got acquired. I think got taken private recently. So originally there this dog boarding marketplace, which is low frequency, it's travel, right?

When you travel, have someone take care of your dog and then Wag comes along and they're like, no, we're doing dog walking marketplace. So all the people who love dogs on the supply side who would board are also going to walk, but the demand is going to be a lot more frequent and Wag just shot out like a rocket compared to Rover. So Rover copied the model and Wag ran into some operational execution issues at scale and Rover eventually took the lead on dog walking as well. But if they hadn't, we'd forget they even existed. I think you have a few examples where, the thing that comes along is disruptive in either higher frequency or lower cost and you just got to go for it immediately as a marketplace founder. You know, DoorDash was way more expensive than GrubHub because they were upcharging a bunch of fees and the menu item prices were higher, so it wasn't really as obvious. What was obvious is that they dramatically expanded supply. And if you dramatically expand supply, generally what wins in the marketplace is who has the most selection for demand.

And that was something that was probably more immediately obvious to GrubHub, but it would just kind of felt hamstrung on what to do about it.

Brett: So I wanted to go back to, the conversation we're having a moment ago about jumpstarting and their early days of building a marketplace from scratch. I'm really interested in hearing more about what are the signs of really strong product market fit early on in marketplaces? Very, very early on, you have your first 100 users, 1, 000 users on the supply side, demand side, etc. What are the key indications that, not that, like, there's something interesting here, but, things are really starting to work?

Casey: A lot of times they won't have product market fit at nine months old. Marketplaces tend to take a little bit longer cause you're effectively building for two customers at the same time, which can take a little bit longer to get it fully to unlock. But you know, there's a couple of things that you're looking for.

You're looking for what is the qualitative data that supply is happy with the service? what are the testimonials? How are they scoring that? And then quantitatively, what does that look like for the frequency with which the supply is interacting, you know, with the marketplace?

And then you're doing the same for demand. Who qualitatively is the demand side? How many of them are coming back and purchasing again or, or whatever. What qualitatively are they saying is the value prop that they're excited? If you're measuring NPS, of course, we'll use that.

And then if you're like, okay, well, I understand why supply and demand are excited. And I see some early data that on early cohort sizes, they're sticking around, That's great. That's the first sign. Then you go down to like, okay, well, how are they reaching, the supply and demand? How costly is that looking?

Is it really hacky? Something that won't scale? Or is it something that will seem to scale pretty well as they get to a thousand, ten thousand, hundred thousand, you know, on the supply side or the demand side, right? And then TAM is kind of a misleading metric for marketplaces.

Cause in all the biggest marketplaces, like Uber and Airbnb, it looked like there was no TAM. but you're kind of trying to understand, well, what is that initial market? If I'm trying to describe the supply in one word, what is it trying to describe demand in one word? What is it? And then what's the theoretical expansion from there?

If they get it working here, is there category expansion that logically makes sense where the ways they've gotten these two people to work together can work in a similar category. Whatnot is one of the companies I work with as an advisor, and their first category for their live streaming marketplace was Funko Pops, which is an incredibly tiny category, right?

It showed that they could make obsessive collector communities hang out and transact, you know, over live stream models. And they've expanded that to sports cards and women's handbags and sneakers now as they become a multi billion dollar company. So you're looking for like, okay, well is the way they unlocked liquidity in this usually small market early on going to most likely be able to jump categories or jump geographies easily and in some cases the answer is I don't immediately see where it would jump to, which means the TAM of that initial category needs to be larger. Obviously, GrubHub never jumped away from food delivery. It was just like, uh, one of the things I really valued about the founders is we're focused on food delivery, only food delivery.

And we're going to do that for a very long period of time. And that, you know, proved to be a massive multi billion dollar category, but there wasn't TAM at the time that suggested it was worth multi billions. no one knew. we just had conviction, that it was big. But if it's not that big, or if it needs to expand dramatically, then I'm trying to evaluate it.

What theoretical category or geographic expansions can make this look larger over time? Sometimes the answer is also that, the value prop needs to change. Now, most of the time we talk about with, marketplaces that you just connect supply and demand and you scale it up for 10 years, no new product developments, don't get distracted by a second product, just scale it up as far as possible, and consumer social is similar in that regard. In some of these next generational marketplaces, you're like, okay, I see the market. I understand exactly what the market is, but I now have a bigger understanding of why that market is fixed, but did the thing you use to unlock that market eventually help you build a second product that's going to expand use cases like Uber, going from, you know, driving people to driving food around.

and I think, you know, you could easily see that potentially working out, you know, with Uber, did they need it to build a multi billion dollar company? No, but to build a hundred billion dollar company, they, probably needed, you know, both of them. In order, trying to understand like the value prop, the customer, make sure the customer is really happy and make sure they're really sticking around. What are the acquisition loops? Do I believe in those acquisition loops? Do I see other ones they're not leveraging that potentially could be using? How big can that go on its own? And if it can't go big enough on its own, then if they get to 10, 000 suppliers, does it open up like a second thing?

Or you do, is that open up a new category or a new, geo that can make it venture scale? And sometimes the answer is no, and that's just a pass. They might build a good business, but they're not going to build a good venture return. And sometimes the answer is obviously, yes. There's an angel investment I made, which is a service technician marketplace called Heave and, they, basically allow independent technicians to go service big contractor equipment, which turns out way cheaper than going to the dealer, to have it serviced. And there's new right to repair laws in the United States that means like Caterpillar, whatever, can't force you to go to the dealer to get it repaired anymore.

So they're like, okay, well, this is new form of supply that hasn't existed, but you look at The price, they're getting a better price and a lot of times better service. Could this eat a lot of the service category? and then the size of that, category is just incredibly large. So are they going to do that in the first year?

No, but theoretically, you could see how it easily expands to eat a lot of the repair work over time.

Brett: How often do you see when a marketplace works exceptionally well in one category, they end up not being able to get big because they can't crack another category? So the Whatnot example would be, they had incredibly strong product market fit in Funko Pops, and that was it. They could never expand. Is that an often configuration or the mistake is made at least in this case, potentially as an investor, maybe an employee thinking about joining the company that they think it will always be a Funko Pops company. But more often than not, if you have incredibly strong product product market fit in a category like Funko Pops, it is highly likely that you will then actually be able to parlay that in the next category, and then the next category.

Casey: There's obviously been examples of both, right? Because you have Whatnot, which has scaled into multiple categories successfully, and then you have Goat, which really hasn't, or it's always been a sneaker marketplace. It probably always will be a sneaker marketplace and that will limit ultimately how big that company could get.

 And you're trying to look at, what are the attributes that enable it to hop very easily? I think as investors, I find it best to assume that if it's working really well in one, it's got the potential to get in multiple. And I should try to get in to all companies that look like that, and if only 7 out of 10, you know, are able to make the hop that's going to be a pretty good portfolio, right? Obviously if you're a later stage investor, it becomes more important to say, is this the right one that's going to make all the categories work, but generally a lot of times you already have the proof point or they're already working in two to three categories.

There isn't like, an incredible amount of big data we can analyze on thousands of marketplaces. It's a small end problem. And I think as we look at most of the winners in marketplaces, You know, when I would say small winners from like 100 million to a billion or bigger winners, which are a billion or more, most of them are single category, but then there's some really big outliers that are multi category, or maybe they're just working into you know, hotels and flights, the Expedia is and bookings of the world.

I think there are some attributes we could probably discuss on what makes something look more like an eBay versus what makes something look more like, a Grubhub or, or DoorDash where it's going to be really focused on one thing. DoorDash is trying to do some expansions right now but, you know, I think a lesson for Marketplace founders is, remember when DoorDash was founded now, 10 years in after they finally won the food delivery market, they're starting to expand and see what will work. I remember having conversations with Tony much earlier about making these expansions and I was like, don't you haven't won yet. You gotta focus like there is no expansion business to worry about yet.

Brett: But why in the case of DoorDash would it be a bad idea to expand earlier in the sense of you can then deliver more value to the demand objects, the end consumer, you can increase, frequency of the app? 

Casey: Well, food deliveries already high frequency, but I think this is a good point, right? A lot of times you're going to feel forced to expand categories to get frequency up like an eBay example, right? If you're just in Beanie Babies or whatever, you're not going to get a lot of high frequency purchases.

So you need to category expand a little bit faster. And this is a conversation you have a lot with founders early on of, oh, we're doing this really low frequency thing. And I was okay, well, if you're a low frequency marketplace, that's fine. There are examples of big companies being built that way.

You need to have incredibly low cost of acquisition or an incredibly high take rate, or you need to have a non transactional product that keeps people engaged even when they're not purchasing so that you don't have to go pay to acquire them when they actually have the need. Or you just need to be so much better than what the market has enabled that they're going to remember you directly, which is what Airbnb had for a long time, right?

Being better and cheaper in a cooler part of town. Well, if you're not going to do one of those, then you need to increase frequency, which means maybe category expand, you know, more, quickly. I think there's a lot of debates about when to do that. But one of the things I'm evaluating is like, okay, if you're low frequency, is the thing we're working on grinding down cost of acquisition or is the thing we're working on trying to get frequency up by expanding, you know, more quickly? could be either or depending on the scenario.

Brett: Are your favorite low frequency marketplaces?

Casey: I think Zillow has always been a really interesting one in that, you know, real estate's incredibly low frequency, but this estimate means everyone feels like they have a relationship with Zillow all the time if you are a homeowner. So it's a way for them to stay top of mind in a way that low frequency marketplaces are generally difficult to do. You know, my first job, which I'm still partial to is apartments. com and what the brand should tell you is that everything is about ranking number one for apartments on Google and it drilled into me a lot of the key SEO learnings that I've leveraged, you know, at Grubhub, Pintrest and Eventbrite since, but everything about that brand is about being intuitive, being quick on conversion.

And I think that's normally what a low frequency marketplace looks is you're just grinding on SEO and conversion all day, every day. And there isn't really that opportunity to build that non transactional product like Zillow has to create engagement. So essentially Google's always your front door, which is scary, but that's kind of the typical, scenario of a low frequency marketplace.

And I've learned like, Hey. Don't hate on it. it is what it is. There's a lot of great businesses that have been built that way. You know, learned my chops at that particular one, but you know, pick any of the hotel marketplaces, their trip advisor, whatever, they're running the same game, right?

Which is just it's very simple to run a business where you know exactly what the key constraints are all the time and they never really change. So I'm like happy to grind on SEO all day, every day, if that's what matters. Some people would find that boring. but I, I do like the models where there's a way that, Hey, even if the purchase is low frequency, we have ways to keep the engagement high so that it doesn't feel like Google controls how fast we grow this month.

And Zillow, I think is the best at doing it.

Brett: One of the unique dynamics of software marketplaces is you kind of have these two specific parts of the experience. One is the software wrapper that you're delivering around the good or service. And then the other piece is the good or service.

 One of the things that I've often reflected on in so many companies, if you think about this construction company, we're talking about repair company and you have a tractor trailer that's broken and you need someone to fix it, just as a thought exercise, you use the product all this, you know, somebody comes, they fix it. And if I'm asking you what your experience was with Joe's repair marketplace, a disproportionate amount of that experience is going to be on the end human coming to your truck and doing something on there.

Brett: And that seems like it has all sorts of inherent challenges when you think about creating a company.

Casey: Another way to say it is the product in a marketplace is the supply, not your software. The software has to work to connect you to supply, but if supply doesn't do a good job, they're going to blame the marketplace because you recommended that person. So you generally have to really get good and raising the standards of supply over time.

So what happens when you start a marketplace is you kind of let supply do their thing. They're like, you know, your business, knock yourself out. And then over time you find as the marketplace team, you're smarter about how to run the supplier's business than even the best suppliers, you know, what the man wants, you know, how to acquire customers better than supply.

So then you do 2 things, you either raise standards of what it takes to be part of the marketplace on the supply side. Or you incentivize the behaviors you know that get them more demand, better quality, you know, all that kind of stuff. So what we did at Grubhub is if your ratings were low enough, we just kicked you off.

If you couldn't reliably deliver food of good quality and we got enough complaints or enough low reviews, we just kicked you off because we know you're not going to deliver a good service. Now, basically no restaurant always delivers a good service, right? There's always mishaps with the driver, sometimes the food isn't, you know, prepare the best. And early on, we were basically like, look, if there's an issue with the driver, if there's an issue with the order being late, we'll take it on as the marketplace, because we know if we make up for a bad experience by going above and beyond, we'll actually have better retention then if you didn't have a bad experience at all, so we were able to prove that in our data, so we went pretty hard and fast into great customer support, which I advise for most startups. Is this going to work on your public company? Maybe not, so, you know, that's, one element. If there's legitimately an issue, just take care of it for the customer, and maybe you don't even charge the supply for it later, and maybe you do, you can kind of figure it out. And then there's a bunch of data that you can aggregate for the demand side to help them make the right choice, which is basically usually ratings and reviews, sometimes a lot more information than that.

And then on the supply side, you're kind of trying to train them what demand expects. So part of what you can do is you can say if you do these things that we know demand likes, you're going to rank higher, which means you're going to get more orders. And then sometimes you can eventually say, Well, you now need to do this to participate in the marketplace.

So a famous example is like Airbnb with Instant Book, request a book, meant you were waiting many times, hours or days to understand if you're going to have the place. There was some discrimination issues with who they accepted for some host. So eventually they just made Instant Book the standard.

But before they made Instant Book the standard, if you had Instant Book, you ranked higher than everyone else. You're kind of going through this process of on all things you learn as you scale up the marketplace, what do I now know better than even the best supplier and how do I use standards or incentives to get more of the suppliers to do the obvious thing. 

Like for Eventbrite, we know event marketing better than even the best event creators, so we started building an email tool and, performance marketing tool for our creators. And we're like, if you use this, you will sell more tickets. we built a freemium model around it. Right? that's something that you see a lot of the good marketplaces do is they're like, ratcheting up the quality of supply as they start to understand what matters to demand. The flip side of that is it creates a dynamic where only power sellers remain that can do all the standards. So the more, you know, individuals, a lot of times can't keep up with the demands of the marketplace.

And you certainly eBay is a prominent example of that, but it's common in most marketplaces.

Brett: There's two things that come to mind. One is that in most cases when the marketplace is successful, people will professionalize on the supply side, and you'll end up with much more of a power law type dynamic on the supply side.

Casey: The level of power law, I think, is different for for different businesses. It's not like there's any restaurant at Grubhub that makes up more than 5 percent of sales so it's still reasonably fragmented. , In a lot of these markets, there's kind of a clearing price over time that gets raised on how sophisticated you have to be, to get real demand from the marketplace.

And, you know, eBay was one of the early examples where it's like, no, if you're on eBay, you're running an eBay store and that's a full time job. And that's what you do. But if you and I want to just sell something on eBay as a random person, the chances of us getting distribution are kind of low. It doesn't mean it's impossible, but it's just harder because we don't have a thousand reviews that are all a plus plus five star.

And, it's kind of riskier to purchase from us because they don't know if we're going to ship quickly or our shipping fee might not be free. All this kind of stuff.

Brett: On sort of a similar thread, do you have thoughts on unique considerations when you are getting supply to do the thing for the 1st time versus the behavior already exists and you're trying to capture it? So in the case of Lyft, Most of those people maybe never were drivers. It's not like they were black car drivers and they're coming over to Lyft. eBay is an interesting example when there was probably a mix of some people who were just selling a pair of shoes that they wanted to get rid of versus sort of people who had a consignment store and brought it over eBay.

But there's lots of examples of kind of new behaviors on the supply side getting created. Airbnb is a good example. I think it's obviously become much more professionalized, but I think in the early days, a lot of people that were providing the supply objects there were not doing this already.

Casey: There's kind of two things that end up happening in these examples. One is you're going to do a lot more training and handholding to make sure they deliver the service as, intended, right? So Uber would manually onboard every driver and Lyft in the early days, right?

Make sure the car was up to spec, make sure they understood how to do all of these different things. And of course there's a lot of waste in that it's expensive, but they had a ideal standard they wanted to administer, and they needed to make sure, you know, that happened every time. So then there's the example of, okay, you actually have an asset, that can be leveraged for the marketplace. And generally the way those models win, whether it's, you know, Airbnb or like Hipcamp is, is a company that I spent a lot of time working with, you're asking that supplier to generally do very little at all to start.

Have the keys available somewhere, but otherwise, you don't need to do anything. and then over time, once the revenue is coming in, where it feels like free money, where you're doing nothing, then you can kind of ask them to raise the standards a little bit more. But it starts off a little bit more like, we actually kind of want the supply to do basically nothing and just make sure the assets there, whether it's Turo and Getaround and car sharing or, you know, Airbnb and Hipcamp and you know, land or home sharing, they're kind of like, yeah, you just kind of get out the way and let us print money from your asset and that's obviously the extreme opposite of Uber or Lyft, where it's here's every step you need to do very detailed to get this money. And then they'll start to look similar over time where they're kind of raising the standards and the level of effort, but then the proof point is there of now I'm making thousands of dollars from Airbnb. It's worth it for me to invest to make that 1500 a month. So those are kind of the two different models at the start. So that's, once the underutilized fixed asset model, I happen to have a car I'm not using, I happen to have, you know, a home I'm not using, I happen to have land I'm not using, whatever, time spent by me as the supplier is directly proportional to the value created.

Follow this very clear set of instructions if you want to stay making this, you know, hourly rate, and until eventually that becomes more commonplace.

Brett: The comment about you have an asset that's sort of underutilized is the thing that you have to get right there is just to make sure that the demand is willing to pay at a rate that is even worth the while of the end asset owner. And maybe this is taking the wrong lesson from it, I think one of the interesting things that I've noticed in the car rental companies, is they've all struggled to really compound value.

One of my theories is just even though someone has an asset that is not being utilized, it's hard to pay them enough to get them to actually move and care and give someone their keys and let someone take the car because they're getting nine dollars an hour or fifteen dollars a day or forty five dollars a day, versus even though it may be more intimate to have somebody stay at your house, the dollar value is just so night and day that i'm willing to even get over small amount of friction that it might be and that there's some counterintuitive things is, in terms of the way that people think about getting paid for these remnant assets, basically, or unused assets.

Casey: Yeah, I agree with that point. I think of it, maybe it's not the right analogy, but I think of it as the bid ask spread in finance and. Car, the area of transportation, there is so many skews at every different price point the bid ask spreads are pretty low in some of these models. So at the low end, you have extremely low end is public transit. And then you have Uber and Lyft, which have dramatically grinded down costs over time. so then you're like, okay, well, if I look at Getaround or Turo how much cheaper is it versus, Uber and Lyft versus the convenience of finding the car. Cause there is lower liquidity than Uber and Lyft, obviously. and then there's, oh, well, I think with Turo, you still have to meet the driver with, I think with Getaround it remote unlocks, but with Turo, you still have to meet the driver or sometimes you still have to meet the driver. When you combine all of this, you're like, wait, it's a very small amount of use cases where car rental feels the optimal decision in the stack of all the different options that I've got, right? Buying a car, taking a self driving car, you know, Uber and Lyft, public transit, bike share, scooters, there's just so many options that it starts to, well, theoretically your TAM is incredibly large. Each one has to take this very thin vertical slice and I feel like in that market, the car sharing companies have ended up with the smallest slice.

Is there a use case where they make sense? Yes, and I think in the case study, I think Turo has done better than Getaround if I've paid attention correctly and part of the reason why is Getaround built a bunch of infrastructure to make it easy for hourly rentals and Turo was wait a minute, there's no money in that, we should optimize everything around daily so at least their AOV is higher when they do get, picked. I could be misunderstanding that, but, you have to kind of see if it's not really competitors, but if the alternatives are like good enough, can you really build a big enough marketplace there?

So I think car sharing is really suffered, suffered from that. And look, we'll see how they do in the longterm. I know Getaround got a very, very low market cap on the public markets right now. And Turo hasn't gone public yet. So it's harder to tell. But, yeah, it seems like that's an area that's gotten squeezed out. 

Brett: When you think about how a marketplace functions in the first couple of years in terms of buyer or seller behavior, how easy is that to shift over time and improve over time? Or do you need to take the resting starting point behavior, maybe more seriously than founders would imagine? Sort of a related topic is one of the things that I've noticed when I've talked to a lot of more consumer oriented founders, not consumer marketplace, but just consumer products, if you think about how retentive a product is, for example. How frequently used it is, in year 1, it is surprisingly difficult to move that, you know, they think, well, I'm going to make this easier. I'm going to use push notifications and I'm going to do it. But like, kind of like the shape and the way that the engagement looks, maybe outside of a step function, like if you introduced reels or some like wildly new content type that maybe bring someone in, you can eke out percentage points, but kind of like the way that people are behaving around the product is give or take what in year five, it's going to look like in terms of how people are behaving around the product.

Does that corollary exist in marketplaces or actually when you look at the behavior in year 18 month, you know, 18 months post starting, no, in 36 months later, you can actually radically sort of shape how either side of the marketplace is behaving?

Casey: I think you can in both models. I do believe it's easier in marketplaces and I'll talk about why. Essentially the phenomenon you're getting at is you tend to bring in the best users first. On the demand side, if it's a marketplace on a consumer app, you bring in, you know, the early adopters first, of course, and they're going to put up with the most flaws.

They're going to have the most egregious need for the thing. And then inevitably, as you want to expand your number of DAUs in a consumer app or the number of transaction, transacting users in a marketplace, you're going to need to expand to people who have the need less, which by default means they're going to be harder to acquire.

They're going to attain worse, so higher CAC, lower LTV, right. Which is probably the worst thing you could possibly see in a consumer business, right? Over time. And we see a bunch of examples of companies just completely failing as they scale up like a blue apron, because they were not able to rectify that issue.

So what needs to happen is that the product needs to get better faster than the users you acquire, get worse and network effects are really the only scalable solution that can make that happen. So for Grubhub, We were acquiring people that would order food regularly in the early days, which wasn't that large of a market.

Only in a few cities like New York and Chicago, do people do that. We went from having 10 online ordering options to a thousand. And then that brings in a whole new crew of people that weren't ordering food as regularly or even at all before. So what we found at Grubhub is that our retention got better every cohort, because the product did get faster on the supply side, faster than the people were targeting, you know, got worse.

So you can see the January, 2019 cohort performing better than in January, 2018 cohort, because they activated better because there were more restaurants, right? So if you look at the cohort retention curves, the year later actually is higher than the year before, which is great and isn't normally the trend in most businesses.

The other thing that can happen is for the people that stick around in the older cohort, they start reactivating or ordering more frequently because there's a lot more options than when, you know, they started, that can also happen in a consumer business, direct network effect business.

at Pinterest, we were able to double the activation rates, after a year of work. But, we had to dramatically simplify onboarding, we had to make sure it worked a lot better in all these international use cases that we weren't that good in, we had to delete a bunch of features, there were like five or six major milestones that got us to double that activation rate because the default is it's going to go down year over year.

So, and a lot of times there just isn't really, there aren't big wins like that left available, right? So it doesn't mean that every time I go to a startup and I advise, I'm like, I can promise I could double your activation rate. Like I don't do that because I'm not sure, right? In the case of Pinterest, it grew very quickly and then flattened out.

And then we got it growing again on, and there were all these different use cases and all these different things that we had built before we were sophisticated enough to understand what maximized activation. And as we got sophisticated, we very much streamlined that experience to maximize, you know, that number.

and I believe in most consumer businesses, that is the most important number, which is how many you activate, meaning, and for those of you who are like, what does that mean? Of the people who signed up and try the product, how many have built a habit where they're going to continue to use it over time?

Meaning like the cohort flattens out. These people like never turn out of the cohort. So, that was the main number we focused on at the growth team at Pinterest and we were able to move that dramatically. And at Grubhub, also that number moved up dramatically, but it mainly moved up not because we iterated on software, it mainly moved up because we got really good at acquiring restaurants and really good at showing great restaurants to people, you know, who are coming in from Google or, or other channels. I do think the default is the users get worse a lot faster than the product gets better. Sometimes you have that key insight on how to simplify the experience or connect to the value a lot better like we did at Pinterest, or sometimes you just have the cross site network effects taking care of it for you and as long as you're able to continue scaling supply, it will solve the problem.

Brett: Lot has been written about pure play, sort of consumer, consumer, social companies, as it relates to the idea of activation. And there's the famous sort of Facebook story, which is probably not true that if you get to X number of friends, you'll become X times more likely to be a daily active user. So the entire company is organized around how do you connect with 10 friends or whatever it was. Does the identical thinking translate to pure marketplace businesses or are there kind of implementation details or things that people need to think about as it relates to the idea of magic moments? 

Casey: So I think the problem with the Facebook analogy, besides the fact that it's probably not true, is that Facebook is such a pure direct network effect business that they can focus on one thing and a bunch of these other dominoes fall for engagement, where for a lot of businesses, it's three different things.

So I break it up into not a magic moment, but a setup, aha, and habit moment. So the setup is what work does the user have to do to have a chance of experiencing the real value of the product for the first time? And for grubhub that's like you got to give us your address, we just can't tell you which restaurants are gonna deliver to you unless we know that and for Pinterest It's we need to know what topics you care about we're gonna bring you in and we're gonna stop you from seeing cool content even if you were browsing it before and saying hey what are your interests?

You're going to pick five of those then we're going to show you a feed that combines all of them. Cause if you only are into one interest, you'll churn like we have all the data, that's the setup moment. You have to get that right. Usually it's some sort of user friction you need to put in place that either the product needs to take care of, or the user needs to do some work to unlock the value.

Then the second is that aha moment. The first time they experience the product value. So for Pinterest, oh, they see cool content related to their interests they didn't know exist. For Grubhub it's like, ah, they see a restaurant that they didn't know delivered and they can, you know, order it, right now.

but then just because you've experienced the value prop once doesn't mean you're going to build a habit of using it all the time. The work to get to the habit moment is how much do we need to get that aha moment repeated before it's become a habit and we can reliably predict they're going to come again and again and again. For Pinterest, that was like, okay, we have to get them pending at least once in the first 4 weeks and after 4 weeks, if they're doing it by the 4th week, we know week 5, we got them like, they're going to stick around basically forever. Right?

for Grubhub, we had to get you a 2nd order in the 1st 30 days to reliably predict you building that habit. So what I try to work with with the companies I work with is try to map out quantitatively and qualitatively where are the set up aha and habit moments and then see how do we see what influences those and then start running experiments to see if we can move each of those higher because then that's when you have the opportunity to really move your activation rate.

A lot of companies don't really know what they are, sometimes it's obvious. Grubhub, you can't really do much besides order food, but for Pinterest, it was kind of like, ah, well, they could just be clicking to the content to go where it came from on the internet. They could be saving it.

They could just be browsing a lot of stuff. So we had to do the analysis and say like, no, it's actually saving a piece of content we recommended that drives the activation rate. So a lot of what I work with is trying to figure that out for different companies and then optimize the hell out of it.

Brett: In this sort of framework that you outlined, could you give some other examples of companies that you've worked at or examples in the past that you thought illustrated this sort of more clearly?

Casey: So, you know, the Eventbrite example, right, is if you come in as an event creator, you have to build an event listing that is robust enough that you can market it and sell tickets. Generally, the, set up moment is you have to tell us a bunch of stuff about what events you're putting on and then the aha moment is when you first see the event listing and the buy button and you're like, ah, this is great and then you go share it. And then there's all these mini moments first purchase or three tickets sold that all correlate better to you having a good event experience but ultimately you're not going to retain back to our, chat earlier about a lot of this stuff happens online, you're not going to actually retain unless the event goes on and you put it on and enough people show up and they had a good time. And we don't even have visibility into all of that. So we try to like kind of get some survey data, Oh, the people who showed up, how did they rate the event? How many people bought tickets?

We have the ticket scanning app, so we know how many people actually showed up, so we have all of these metrics how many people bought? How many people showed up? Did those people have a good time? And all of that correlates to them putting on future events and becoming a habitual event creator.

 So the challenge coming in compared to like a Pinterest or a Grubhub was, if you want to map people building habits, well, some people's idea of the habit is putting on events every week, and some people's idea of the habit is putting on events every quarter. And we found that that's not super influenceable, the frequency maps to their business model, not to how good our software is, right?

The music festival person's not going to start putting on music festivals every week. Like, it's just not possible. They actually, Eventbrite when I joined was not really in the habit of measuring cohort retention, it just measured net dollar retention. And I'm like, no, we need to measure cohort usage retention, but for like 20 different groups.

So the music festival is we need to map like not the frequency of events, but the frequency of them engaging with other parts of the product to make that music festival happen. But for the smaller frequent creators, we need to map their frequency of hosting events at least once a month, or it was actually 35 days because of variances with like holidays and weekends.

So we had to basically build custom cohorts for different sub segments of our supplier base. which I hope you don't have to do in your business because it's much more complicated, but it was the right thing for us because we're extremely horizontal product. There's just so many different types of creators of different sophistication, different frequency.

If you put them all on one cohort, it kind of looks like gobbledygook, right? And we would say all the festival businesses churned, which obviously they didn't, they put on and then once a year. So that's a more complicated example and you'll see this in wide horizontal products that service a lot of different types of users either.

Types of businesses, types of categories or types of sophistication.

Brett: When you think about all the marketplaces that you've studied and so many of the marketplaces that you've actually worked on are there specific patterns of errors that you keep coming across?

Casey: So one that comes off the top of my head is that marketplace businesses need to get sophisticated around data, usually a lot more quickly than other models. And part of the reason for that is once you're doing more than one category or more than one city or neighborhood, the aggregate data doesn't really tell you anything anymore.

The only trends that are interesting are mapping by specific geo, by specific category, by specific acquisition channel. So you need to ramp up your data sophistication a lot more quickly. Or you end up seeing stuff in aggregate that's actually conflating different trends in different cities or, different categories.

So a lot of what I end up working on when I work with marketplaces after like series A or series B is like, okay, let's go build up your data capability. Let's go build up the dashboards that your team should be paying attention to. Lets really make sure you're looking at the appropriate slice of data.

And then I have, you know, all the examples I've used from other companies to help them, you know, do that, more quickly. And then related to that, which we talked about earlier, is what's that custom activation metric on supply and demand? A lot of times, people are acquiring a lot of supply that doesn't really get any transactions.

And then that supply will churn. So you can generally figure out how long is the supply going to give you to start showing them value and then you need to make sure by that time, everyone's getting the appropriate amount of transactions, right? So at Grubhub, that was like, you needed to get them two orders a day to stick around as a restaurant, and then they would never, you know, churn. That's a problem that, you know, most people aren't familiar with. I'm trying to think of just mistakes I made, cause everyone else makes them too.

One is, discounting to product market fit does not get you to product market fit. So. In a marketplace, you are selling people on the value of the selection on the value of, you know, the service and if you then give a 5 discount then you start changing the value prop to saving money on service instead of the value of the service.

So we were very anti all promotions at Grubhub, but we acquired campus food and I was in charge of integrating it. So we're basically Grubhub, but for college towns. They were discounting to first order and then it meant it took to a third order instead of a second order to see if people are going to retain.

So then they started discounting the third order and then it pushed people to a fifth order to see if they retain. Then they started discounting the fifth order. And it's like, no, this is like never going to end. You have to be convincing people up front to pay full price. The flip side of that is if the promo is used to get someone to try something, that they wouldn't have tried that, you know, in your data predicts a dramatic improvement in LTV.

That's where promotions can be quite useful. So when our mobile app found product market fit at Grubhub, we decided to give everyone 10 off their first mobile order because LTV doubled if you got the app. And it was a little bit of a gambit, but a bunch of people use that promo and then they started using the app and their frequency doubled.

It worked out for us. And then we later did stuff like, Oh, well, if you order sushi for the first time, we'll discount that because if you order sushi for the first time, you're going to order sushi again, because sushi is awesome and it's a higher AOV. So there's a bunch of things that we said, Hey, these correlate to hire LTVs and higher frequency.

Maybe we can discount to see if people can try that for the first time and then build a habit, but they were already habitual Grub abusers was more of an LTV game sort of thing. So, but yes, the promo stuff is a very dangerous game, especially for new users. And I would generally shy away from it. So that's a, that's another mistake I see a lot of marketplaces make. 

The other thing is trying to do too much expansion without really refining what the playbook looks like. So, obviously we talked about geo and category expansions generally very important for marketplaces to reach venture scale. But it's very important for you to say what really mattered in getting our first market to work? And which of those do we need to replicate to get the second market to work or the second category to work?

And really refining, what are the steps that really matter? And what are the steps that don't matter? and that will be a bit bespoke to every company, right? A bunch of people tried to copy Uber's launcher strategy and then a bunch of people like almost went bankrupt doing it cause it's really expensive and they weren't raising billions. You didn't really understand what your playbook required. You were just kind of carbon copying a very different, you know, marketplace model. So I see a lot of them, they haven't really written down, here's exactly what matters, and here's how I have someone in charge of making sure all these things happen to unlock liquidity in a new category in a new market. And what that means is until you're good at that, you should only be doing one at a time. You shouldn't be trying to launch 10 markets at a time if you don't have a refined market expansion playbook, you should be doing one and validating your hypotheses or same for like category.

There are, counter examples to this, right? Like Thumbtack basically got in all categories because they found it was just as easy to rank an SEO for wedding DJs as it was for roofing. So they just did all of it. And then they would kind of manually send those leads to like pros they found on Yelp until they could acquire the pros.

So, Hey, if you got an unfair arbitrage like that, cool, knock yourself out. But most of the time you want to have that really refined playbook on how to make it increasingly efficient to launch new markets and get them to liquidity. And some people skip out on that part.

Brett: So to wrap up, I'm curious, when you think about building and scaling marketplaces, is there anyone who's kind of imparted more on you or lessons learned around marketplaces than anyone else? 

Casey: There's, you know, a few that kind of stand out. Obviously, Bill Gurley has written a bunch of great content and was on our board at Grubhub and I think just really instilled the value of owning demand being kind of the only thing that matters. And if you have control of the customers, then supply is going to be fine.

And famously Uber, like does not believe this. They believe supply is all that matters. And I fundamentally disagree. Supply is a lot of times the tool to get demand, but owning demand is what keeps the supply happy. So I think, you know, he's written a bunch of great content. Simon Rothman was someone I worked with at Greylock who is, the, GM of, eBay US and eBay motors, and very prominent angel investor and a bunch of great marketplaces.

He helped really refine a lot of these different factors and then what their sensitivities are, by being able to evaluate, you know, marketplaces with him. Where I learned a lot more from him is, so there's marketplaces that unlock liquidity, which we could say is product market fit, right? My mental model at the time was okay, now you're in my land, you're in growth land, and I will get this as many users as possible as efficiently as possible, right? And I think what Simon opened up my eyes to is yes, but is there a stronger form of product market fit possible? is there a more frequent use case we can do?

Is there an optimized version of the core product experience versus just going and getting more supply and going and getting more demand, which wasn't previously part of my mental model and I now have counseled a lot of companies I've advised or angel invested being like, yes, this works. there's a even stronger form of product market fit out there for us.

If we're a little bit more patient, if we invest a little bit more on the core product experience generally that always will pay back 10 X more. Uh, and that's not something, I really had thought a lot about until I got to work with him closely on a bunch of different marketplaces. So, those are two that like really stand out and just, their writing on marketplaces, their thinking, their board experience, you know, in, in Simon's case, the operating experience was top notch.

Brett: Cool. Great place to end. Thanks so much for spending the time.

Casey: Thanks for having me.