The Exchanges

Q advising many startups like FAIR. Whether you're a founder in the early stages of building a marketplace startup or a leader looking to reignite growth in a more mature business, I think you'll find tons of insights in this episode. With that, I bring you my conversation with Casey. Maybe we could start by talking about some of your thinking on what are the requirements to make a marketplace work?

A 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, like, a high enough take rate to build a successful marketplace. So you see with the Expedia's of the world, the bookings of the world, they 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, the…

AI assessment note: “generally the first couple you think about are, is there supply fragmentation?”

Q customer, in most cases, maybe the top end of any marketplace, the number one or 10 or hundred 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 little work to be done on the supply side in that thinking?

A 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% 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 their expenses. Even the top end are like, yeah, I'll take more catering or delivery orders for sure. I think what you're 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 prop that I can convince you I can bring demand. What you're seeing with a little, 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 fair of we kind of need you to start using this free w…

AI assessment note: “next generation of marketplaces is it requires more products to be in a position”

Q 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.

A I hope I don't get in trouble for saying some of these. A company 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've 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 its needle out of haystack. So you have a bunch of people Power is being able to raise their hand and say, like, I'm actually looking for this. So that's working really well. I have a company in the SaaS space I advise called Fermat Commerce, and they re…

AI assessment note: “A company Angel invested in is called Power. It's a clinical trials marketplace.”

Q 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?

A 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, you know, 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. Let's 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, 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 turn. So you can generally figure out How long is the supply going to …

AI assessment note: “marketplace businesses need to get sophisticated around data, usually a lot more quickly”

Q the 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 hundred users, thousand users, On the supply side, demand side, et cetera. What are the key indications that not that like there's something interesting here, but things are really starting to work?

A A lot of times they won't have product market fit at nine months old. Marketplaces tend to take a little bit longer because 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? Where 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 with the marketplace. And then you're doing the same for the demand. Who qualitatively is the demand side? How many of them are coming back and purchasing again 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.

AI assessment note: “there's a couple of things that you're looking for. You're looking for what is the qualitative”

Q 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?

A Let's talk about what it means to For them playing a different game. Grubhub was an asset-like 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-like model at all, because in addition to acquiring restaurants and to acquiring consumers who wanted to order food, you also had to acquire the drivers that were going to deliver back and forth. Um, early on the unit 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 eighty million dollars before you go public. Pretty asset-like 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 two dollars per order and raising hundreds of millions of dollars fairly quickly in Silicon Valley. And, you know, to a 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 de…

AI assessment note: “you can't really go to the public markets and say like, Hey, we're going to adopt”