Latest / Marketplace Businesses with Fexingo: Two-Sided Networks, Liquidity, and Take Rates / How Marketplaces Use Referral Programs to Build Liquidity Cheaply
Transcript
- Lucas: You know what's one of the most under-discussed tools in the marketplace playbook? Referral programs. Not the splashy ad campaigns, not the subsidised pricing — just asking your existing users to bring in their friends. Luna: It sounds almost too simple. But I've seen the data — referrals consistently have the lowest customer acquisition cost of any channel for marketplaces. Lucas: Exactly. Let's ground this in a concrete number. In 2013, Uber was running a driver referral program that paid five hundred dollars per new driver who completed a certain number of trips. And the driver who referred them also got a bonus. That's a lot of money, but consider the alternative: they were spending even more on Facebook ads, on Google search, on billboards. Luna: Right. And the referral channel had a built-in quality filter. A driver who joins because their friend referred them is more likely to stay, more likely to understand the platform, less likely to churn. Lucas: That's the key insight — referral programs aren't just cheaper, they're stickier. The referred user has a trusted guide already on the platform. For Uber, that meant lower churn and higher lifetime value. They calculated that the effective customer acquisition cost via referral was about sixty to seventy percent lower than paid channels in those early years. Luna: What about Airbnb? I remember they had a host referral program too, but the economics were different. Lucas: Airbnb's early host referral paid a hundred dollars per new host, but the catch was the new host had to complete a booking within a certain timeframe. That booking requirement ensured the referral was real — that the new host actually listed a property and got a guest. Without that step, you'd get a lot of sign-ups that never transact, which doesn't build liquidity. Luna: So the design of the referral incentive matters. It's not just any referral — it has to be a referral that results in a transaction. Lucas: Precisely. And that's where the unit economics get interesting. If you're a marketplace, you care about liquidity — the number of completed transactions. A referral that brings in a user who never transacts is worthless. So the best programs tie the reward to a real action, like a completed trip, a booked night, a delivered order. Luna: DoorDash did that with their Dasher referral. They paid the referring Dasher a bonus only after the new Dasher completed a certain number of deliveries. It aligns incentives — the existing Dasher has a reason to help the new one succeed. Lucas: Exactly. And that's a beautiful feature of well-designed referral programs — they turn your existing users into a part-time growth team, but a growth team that's motivated by real alignment. They're not just broadcasting a link; they're actually coaching the new user through the onboarding process because they want that bonus. Luna: Let's talk numbers. What's a typical referral program cost per acquired user, and how does it compare to paid channels? Lucas: For a mature marketplace like Uber or DoorDash, the fully loaded cost of a referred acquisition — including the reward, the fraud detection, the operational overhead — is often between twenty and forty dollars per user. Compare that to paid search, which can easily be sixty to a hundred dollars per user in competitive markets, or display ads, which are even higher with lower conversion rates. Luna: So it's roughly half the cost. But there's a ceiling, right? You can't just keep paying for referrals forever — eventually you saturate your existing users' networks. Lucas: That's the limitation. Referral programs have a natural cap because each user has a finite number of friends interested in driving or hosting. Uber saw the referral conversion rate drop dramatically after the first few waves. Early on, every driver could refer two or three friends easily. After that, the well runs dry. So referral programs are fantastic for that initial liquidity boost, but they can't be the sole growth engine forever. Luna: And there's a risk of fraud. If you're paying a hundred dollars per new host, some people will try to game the system — creating fake accounts, referring themselves. Lucas: Right. TaskRabbit had to shut down their referral program temporarily in 2015 because of rampant fraud. People were referring themselves using prepaid credit cards and fake email addresses. The company lost a significant amount of money before they realised what was happening. They later relaunched with stricter identity verification and a waiting period before the reward was paid out. Luna: So the program design has to include fraud prevention from day one. Things like requiring a real bank account, or a verified phone number, or a minimum time on platform before the reward vests. Lucas: Exactly. And the best marketplaces treat their referral program as a product — they A/B test the incentive amount, the payment structure, the eligibility criteria. Uber famously tested whether to pay the reward as a lump sum or as a series of smaller payments over time. They found that smaller, more frequent payments kept drivers engaged longer and reduced fraud because the bad actors didn't want to wait. Luna: Interesting. So it's not just about the amount, it's about the psychology of when and how you pay. Lucas: Right. And this is where marketplace operators can learn a lot from each other. The playbook isn't secret — it's just that most people don't think of referrals as a core liquidity strategy. They think of it as a nice to have feature. But in the early days of a marketplace, referrals can be the difference between a cold start and a thriving network. Luna: I want to bring in a more recent example. The Indian food delivery platform Zomato had a referral program that was hugely successful in 2020 and 2021. They offered a discount to the referrer and the referee on their next order. But they also capped the number of referrals per user to prevent abuse. Lucas: That cap is smart. It forces the program to be about genuine recommendations, not spamming. And it keeps the cost predictable. Zomato's referral program reportedly accounted for over thirty percent of new customer acquisitions at its peak, with a cost per acquisition below twenty percent of what they were paying for digital ads. Luna: Thirty percent from referrals — that's huge. And it's not just for consumer marketplaces. B2B marketplaces like Upwork also use referrals. In fact, Upwork's referral program for freelancers is a key part of their supply-side growth. Lucas: Great point. Upwork pays a referral bonus to freelancers who bring in other freelancers who then complete projects. Again, the reward is tied to actual work completed, not just sign-ups. That ensures the new supply is real and contributes to liquidity. Luna: It makes me think about the viral coefficient. A good referral program can give you a viral coefficient above one, meaning each new user brings in more than one additional user. That's how you get exponential growth without spending on ads. Lucas: Exactly. But the viral coefficient is hard to sustain. Most marketplaces see it drop after the initial few months as the low-hanging fruit is exhausted. That's why you see them layer multiple referral programs — one for customers, one for suppliers, maybe even one for both sides simultaneously. Luna: And this is where I think the conversation connects to something broader. If these conversations are useful for what you're building or running, a couple of dollars a month is genuinely what keeps these going. Lucas: Yeah, it's true. This is entirely listener-supported, no ads. Buy me a coffee dot com slash fexingo — if you've gotten something out of them, it makes a real difference. Luna: We'll keep doing the deep dives on marketplace mechanics either way. Lucas: Alright, back to referrals. One more dimension I want to touch on is how marketplaces use referral programs to target specific segments. For example, a marketplace might offer higher referral bonuses for suppliers in undersupplied categories. Or for customers in geographic areas where they need more demand. Luna: That's a smart way to use the program not just for volume but for balance. If you need more plumbers in a certain city, you can target your referral bonus to plumbers in that city. Lucas: Right. And you can do it without advertising the different rates publicly, because it's a direct incentive to the user. Thumbtack has done this — offering higher referral bonuses to pros in categories where they have more customer demand than supply. Luna: It's a dynamic pricing technique applied to referrals. Very marketplace-native thinking. Lucas: So to pull back a bit — what's the one thing a new marketplace founder should keep in mind when designing their first referral program? I'd say: tie the reward to a completed transaction, not just a sign-up. And start with a small incentive, then test higher. You can always increase, but it's hard to decrease without upsetting users. Luna: And have a fraud detection plan from the start. Even if you think you're too small to attract fraudsters, you will be surprised. Lucas: Absolutely. And if you do it right, referrals can be the most efficient liquidity-building tool you have. The question is whether AI will make referrals even more powerful — by predicting which users are most likely to refer high-value new users, and then giving them a tailored incentive. Luna: That's a good forward-looking note. I imagine within a few years, every marketplace will have a machine learning model scoring referral propensity and customising the offer in real time. Lucas: It's already happening at the bigger players. Uber uses ML to decide which drivers to target with referral offers, and at what amount. They've reportedly increased their referral conversion rate by over twenty percent just through better targeting. Luna: So the playbook keeps evolving. But the core insight — ask your users to bring their friends — has been around since the earliest days of eBay. Lucas: Right. And it works because it's built on trust. That's something no amount of ad spend can buy.