Latest / Marketplace Businesses with Fexingo: Two-Sided Networks, Liquidity, and Take Rates / How the Best Marketplaces Decide Which Side to Subsidize
Transcript
- Lucas: If you're building a marketplace, the single most expensive strategic decision you make isn't your logo or your app design — it's which side you subsidize. And most founders get it wrong. Luna: Subsidize meaning you pay one side to join, or you charge them less than it costs to serve them. Lucas: Exactly. And the classic mistake is subsidizing the side that would have shown up anyway. Let's start with a concrete case: OpenTable in the early 2000s. Luna: OpenTable — the restaurant reservation system. I remember they gave restaurants free hardware and software to install the booking terminal. Lucas: Right. They spent something like $100 million deploying those terminals before they had a single diner booking through the platform. That's a massive supply-side subsidy. The bet was that restaurants wouldn't pay upfront, but once the terminals were in, the value to diners was obvious — and diners would pay nothing, while restaurants paid a per-cover fee. Luna: So they subsidized supply because supply was the bottleneck. Restaurants were hard to convince; diners were easy. Lucas: Bingo. Compare that to Uber. In Uber's early days, drivers were the bottleneck — they needed to see demand before they'd sign up. So Uber subsidized riders. First rides were free or deeply discounted. That created demand, which attracted drivers, which made the service better, which attracted more riders. The subsidy flipped sides depending on which side was harder to get. Luna: But not every marketplace follows that pattern. Airbnb didn't subsidize either side heavily at first. They let supply grow organically from people who already had spare rooms. Lucas: That's the key insight — the right subsidy strategy depends on the elasticity of each side. Elasticity meaning: how much does a price change affect how many people join? If supply is highly elastic — meaning a small subsidy brings a lot of new listings — subsidize supply. If demand is more elastic, subsidize demand. Luna: So the rule of thumb is: subsidize the side that's more price-sensitive and harder to acquire at scale. Lucas: Exactly. And you can measure this. Look at the cost to acquire a supplier versus a customer. In the early days of Etsy, it was cheaper to acquire a buyer through Google ads than to recruit a seller. So Etsy subsidized sellers — they kept listing fees low, offered free shops for a trial period, and invested in seller tools. That made supply abundant, which attracted buyers. Luna: But there's a trap, right? If you over-subsidize one side, you attract low-quality participants who churn as soon as the subsidy ends. Lucas: That's the danger. The best marketplaces use subsidies as a temporary bridge to liquidity, not a permanent crutch. Once you have enough activity on both sides, you gradually reduce the subsidy and let the network effects take over. The data from your first thousand transactions tells you which side is stickier. Luna: I remember a great example from Thumbtack. They originally charged service professionals to bid on jobs, but they found that pros were price-sensitive and would only bid if the lead cost was low. So they shifted to a subscription model — pros paid a flat monthly fee for unlimited leads. That effectively subsidized the supply side by lowering the marginal cost per lead, which kept pros on the platform. Lucas: Smart. And the subscription model also reduced the risk for pros — they knew their maximum cost per month. That's a form of subsidy that's actually a pricing innovation. The point is, the decision isn't just 'which side gets the discount,' it's 'how do we structure the subsidy so it aligns with long-term behavior?' Luna: You know, this is exactly the kind of thinking that makes these shows useful for people building or running businesses. And if you're finding value in them, a couple of dollars a month is genuinely what keeps these going — buy me a coffee dot com slash fexingo, if you've gotten something out of them. Lucas: Yeah, it's a small thing that makes a real difference. Keeps us independent and ad-free. Back to subsidies: one of the most counterintuitive findings from the academic literature is that sometimes you should subsidize the side that's already growing. Luna: Wait — that sounds backwards. If one side is already growing organically, why spend money on it? Lucas: Because that side might be the accelerator. Take a marketplace for freelance graphic designers. If designers are joining on their own because of word of mouth, but clients are slow to arrive, you might subsidize clients — give them a discount on the first project. That creates demand, which makes designers happier, which attracts more designers. The subsidy on the growing side is actually a catalyst for the other side. Luna: So you're not subsidizing to fix a bottleneck; you're subsidizing to amplify a flywheel. Lucas: Exactly. The classic mistake is subsidizing the side that has low retention. If you pay people to join but they leave after one transaction, you're burning cash. The metric to watch is repeat usage. Subsidize the side that shows high repeat rates once they're in. Luna: What about the timing? When should you start subsidizing? At launch? After you have some liquidity? Lucas: Most successful marketplaces start subsidizing before they have any liquidity — that's the chicken and egg problem. But they do it surgically. They don't subsidize everyone. They identify a small cohort of high-quality suppliers or customers and give them a deal to join early. Luna: Like what Uber did in San Francisco — they recruited a handful of black cars and gave free rides to tech workers. Lucas: Right. And for a vertical marketplace, say one connecting plumbers with homeowners, you might offer the first 50 plumbers a six-month free listing, and give the first 100 homeowners a $50 credit. The size of the subsidy should be just enough to get them to try, not enough to become a profit center. Luna: Is there a formula for calculating the optimal subsidy amount? Lucas: There's a framework called the 'liquidity threshold.' Calculate the minimum number of transactions per week needed to make the marketplace useful. Then work backwards: how many suppliers and customers do you need to generate that volume? Then figure out the acquisition cost for each side. The subsidy per participant should be less than the lifetime value they'll generate. If your LTV is $100, you can afford to spend $30 to acquire them. If it's $10, you can't. Luna: So the subsidy is really just a marketing expense, but one that's directly tied to the network effect. Lucas: Exactly. And the most sophisticated marketplaces use dynamic subsidies — they adjust the subsidy in real time based on supply-demand balance. If there are too many buyers and not enough sellers, they increase the seller subsidy. If sellers are abundant, they cut it. Luna: That's what Uber does with surge pricing, but in reverse. Surge pricing charges riders more, which effectively subsidizes drivers by giving them higher earnings. Lucas: Right. It's a real-time subsidy to the supply side. And the data shows that dynamic subsidies increase transaction volume by 20 to 30 percent compared to static pricing. Luna: Let's talk about a failure. A marketplace that subsidized the wrong side. Lucas: One classic example is the original version of Homejoy, the home cleaning marketplace. They subsidized customers heavily — first cleaning for $19. That attracted a ton of demand, but the cleaners were paid poorly and had high churn. The supply side never stabilized because the subsidy didn't go to them. Homejoy eventually shut down. Luna: So they subsidized demand, but supply was the real bottleneck. Cleaners needed better pay to stick around. Lucas: Exactly. A cleaner who only gets one booking a week from your platform won't stay. You need to subsidize supply to ensure enough inventory. Homejoy's mistake was treating cleaning as a commodity when it's actually a high-churn labor market. Luna: That's a great lesson. Now, if I'm building a marketplace tomorrow, what's the one number I should calculate first? Lucas: The acquisition cost ratio — the cost to acquire one supplier divided by the cost to acquire one customer. If that ratio is low, meaning suppliers are cheap to get, subsidize customers. If it's high, subsidize suppliers. Then test it with a small cohort and measure retention after the subsidy ends. Luna: And if both sides are expensive? Lucas: Then you probably don't have a viable marketplace yet. You need to find a cheaper channel — maybe referrals, SEO, or partnerships. Marketplaces are hard because you have to solve two acquisition problems at once. The subsidy just buys you time. Luna: What about marketplaces that don't use direct cash subsidies? Like using content or tools as subsidies? Lucas: That's actually the smartest approach. Instead of paying people, give them something valuable that costs you little to reproduce. For example, a marketplace for photographers might provide free editing software to photographers. That's a supply-side subsidy that builds loyalty and creates switching costs. Luna: So the best subsidies are non-monetary when possible. Lucas: Yes. Because cash subsidies attract mercenaries — people who leave as soon as the cash stops. Tool subsidies attract loyalists. And loyalists build the network effects that make a marketplace defensible. Luna: It's like the difference between paying someone to attend a party versus giving them a reason to stay. One is a transaction; the other is an investment. Lucas: Perfect analogy. And the best marketplace operators I've studied think of subsidies as investments in liquidity, not expenses. They measure the return on that investment by tracking how many transactions happen after the subsidy ends. Luna: So the final takeaway: know your bottleneck side, measure elasticity, subsidize with retention in mind, and prefer non-cash subsidies when you can. Lucas: That's the playbook. And if you're building something, start with the smallest subsidy you can, measure everything, and let the data tell you which side to bet on.