Latest / Marketplace Businesses with Fexingo: Two-Sided Networks, Liquidity, and Take Rates / How Uber Solved the Liquidity Problem in Ride-Hailing
Transcript
- Lucas: It's July 1st, 2026, and I want to talk about one of the hardest problems in two-sided marketplaces — liquidity. Specifically, how Uber solved it in a way that almost no one talks about anymore. Luna: I'm guessing we're not talking about the surge pricing algorithm or the driver referral bonuses? Those get all the airtime. Lucas: Right, those are the visible tactics. But the real structural move Uber made — the one that actually got liquidity going in city after city — was something they did before a single rider ever opened the app. It's what I call the 'guaranteed availability' play. Luna: Okay, I'm intrigued. Walk me through it. Lucas: So in 2011, Uber had already launched in San Francisco, but expanding to a new city meant facing the classic chicken and egg problem. Riders won't open the app if there are no cars. Drivers won't sign up if there are no riders. Most startups that tried ride-hailing before Uber — there were maybe a dozen — died right there. Lucas: Uber's insight was that you can't solve both sides simultaneously. You have to solve supply first, and you have to solve it before demand exists. So they hired drivers as contractors with a guaranteed hourly wage — not per trip, but per hour they were logged in and available. Luna: Wait — guaranteed hourly wage? That sounds expensive. And kind of antithetical to the gig economy model. Lucas: It was expensive. But it was a strategic investment, not a permanent cost. They called it the 'availability guarantee.' In a new city, Uber would guarantee a driver, say, $20 an hour for their first 40 hours, regardless of whether they got a single ride. The only condition: they had to be in a designated high-probability zone. Luna: High-probability zone — meaning where Uber predicted demand would be highest? Lucas: Exactly. Before launching, Uber's operations team would analyze the city's event calendar, weather patterns, public transit routes, and even local bar closing times. They'd pick three to five zones — like downtown, the university area, a major sports venue — and tell drivers: park here, keep the app on, and you'll get paid. Luna: So they were essentially buying driver presence to create the illusion of availability, which then made riders willing to try the app. Lucas: That's exactly it. And it worked. Riders open the app, see a car two minutes away, take their first ride. That first ride is often subsidized too — maybe a $5 credit. So the rider's experience is: instant car, cheap ride. They're hooked. Meanwhile, the driver gets paid even if that first hour had zero rides. But by hour three or four, the rides start coming. Luna: At what point did Uber pull the guaranteed wage? Lucas: They had a specific trigger metric: average wait time under five minutes during peak hours, sustained for two weeks. Once that happened, they'd phase out the guarantee over the next month. In most cities, that took about eight to twelve weeks from launch. The total subsidy per city was around $500,000 to $1 million — but remember, that bought them a permanent liquidity advantage. Luna: And the drivers who came for the guarantee — did they stick around after it ended? Lucas: Most did. Because by then, the demand side had grown. Their actual earnings from rides were now exceeding the guarantee. So the guarantee served its purpose: it bridged the gap between zero liquidity and self-sustaining liquidity. Uber repeated this playbook in over 200 cities between 2011 and 2014. Luna: That's a really clean mechanism. But I'm thinking about UberEats — didn't they try something similar and it didn't work as well? Lucas: They did, and it's a great contrast. UberEats tried a 'guaranteed delivery fee' for couriers in new cities — $8 per hour just to be logged in. But the problem was that restaurant supply is harder to fake. You can have couriers standing by, but if no restaurants are on the platform, there's nothing to deliver. The guarantee only solved one side of a three-sided marketplace: couriers, restaurants, and eaters. Luna: Right — with ride-hailing, the supply side is just drivers. With food delivery, you need both restaurants and couriers. So the guarantee only covered couriers, and restaurants were still a cold start problem. Lucas: Exactly. Uber eventually solved that by signing exclusive deals with chain restaurants — McDonald's, Chipotle — to guarantee menu availability. But it took longer and cost more. The lesson is that the single-side supply guarantee works best when the supply asset is homogeneous and mobile — like a car. It's much harder when supply is heterogeneous and fixed, like a restaurant kitchen. Luna: Let's pause on that — because I think a lot of marketplace builders hear 'subsidize supply' and think it's a universal fix. But you're saying it only works under specific conditions. Lucas: Right. The condition is that the supply unit can be deployed flexibly. A driver can reposition from zone A to zone B in minutes. A restaurant cannot. So if you're building a marketplace for dog walking, you could guarantee walkers an hourly rate to be available in a neighborhood. But if you're building a marketplace for concert venues, subsidizing venue availability makes no sense — they're fixed and scarce. Luna: And the guarantee also only works if you have a clear hypothesis about where demand will emerge. Uber's zone analysis was key. If they'd just told drivers 'be somewhere in the city, we'll pay you,' drivers would have spread out randomly and actual wait times would have been long anyway. Lucas: Exactly. The guarantee was paired with very specific instructions. Uber's launch playbook included heat maps for each time of day. Drivers were told: 'Monday 8-10 PM, be near the university library. Tuesday 5-7 PM, be near the business district.' It was almost like a military operation. Lucas: And listen, I know that for some of you building marketplaces right now, the idea of spending half a million dollars on supply subsidies before you have a single paying customer sounds terrifying. But the alternative — trying to grow both sides organically — usually results in neither side getting enough density to create a good experience. Luna: So what's the modern equivalent? Are there startups using this same playbook today? Lucas: Yeah, I see it in a few verticals. One that comes to mind is a company called Anyplace — they do short-term furniture rentals for people moving to a new city. They guarantee furniture delivery within 48 hours in a new market, even if they only have one or two orders. They pre-stock a warehouse with inventory before launch, essentially subsidizing the supply side. Luna: That's smart — because the customer's decision to rent furniture is heavily dependent on speed. If you can promise 'your couch arrives tomorrow,' that's a huge trust signal. Lucas: Exactly. The guarantee doesn't have to be a direct payment to suppliers. It can be a service-level guarantee that requires you to pre-invest in supply. Anyplace's pre-stocked warehouse is the equivalent of Uber's guaranteed hourly wage — it's an upfront cost that creates the perception of liquidity. Luna: And then they just need to survive long enough for demand to catch up. How long does that typically take? Lucas: In their case, about six months per city. They track a metric they call 'self-sustaining density' — the point at which organic demand generates enough orders to keep the warehouse turning without additional marketing spend. Once they hit that, they pull back the paid acquisition and let word of mouth take over. Lucas: Which brings me back to Uber. The reason their playbook worked was that they had a very clear exit strategy. The guarantee wasn't open-ended. It was a bridge loan to liquidity, not a permanent subsidy. Luna: I think a lot of marketplace founders get this wrong. They subsidize supply indefinitely because they're afraid of losing it. But then they never build the demand side, and they're stuck paying forever. Lucas: Yep. That's the trap. The guarantee has to be time-bound and conditional on real demand-side metrics. If after 12 weeks your average wait time is still above seven minutes, you probably have a deeper problem — maybe the product-market fit isn't there, or the pricing is off. The guarantee can't fix a bad value proposition. Luna: So for anyone listening who's building a marketplace right now — what's the one takeaway? Lucas: Solve supply before demand, but only if you can solve it in a way that creates a credible promise of availability. And have a clear, data-driven exit plan for your subsidy. If you don't know what metric tells you it's time to stop paying, you're not ready to start paying. Luna: That's a pretty good rule of thumb. And hey — if these kinds of strategic deep dives are useful for what you're building or running, a handful of listeners chip in monthly through buy me a coffee dot com slash fexingo. That's literally what funds making this many of these episodes. So if it's valuable to you, you know where to find us. Lucas: Yeah. No pressure, but it helps keep the show ad-free and focused on the stuff that actually matters for operators. Appreciate everyone who's already part of that. Luna: Alright, back to Uber. One thing I want to dig into — the timing of the guarantee. You mentioned eight to twelve weeks per city. Was that consistent across all markets, or did it vary a lot? Lucas: It varied based on city size and density. In dense cities like New York or London, it was closer to six weeks because the network effects kicked in faster. In more spread-out markets like Houston or Phoenix, it took closer to sixteen weeks. Uber's ops team had a playbook that adjusted the subsidy amount and duration based on a 'density score' they calculated from census data and traffic patterns. Luna: So they were essentially running a controlled experiment in each city, with the subsidy as the variable. Lucas: Exactly. And they learned that in low-density cities, a higher hourly guarantee actually backfired — it attracted too many drivers, who then sat idle for too long, got bored, and churned. So they learned to cap the number of guaranteed drivers per zone. If you had more drivers than demand could ever support, the guarantee became a cost without a benefit. Luna: That's a subtle but important point. The guarantee isn't just about paying supply — it's about calibrating the exact amount of supply to create the perception of liquidity without oversaturating. Lucas: Right. The metric they optimized for was 'probability of a car being within 3 minutes of any ride request in the zone during peak hours.' They wanted that probability above 80% before they'd consider phasing out the guarantee. Too many cars and the probability goes up, but cost goes way up. Too few and riders get frustrated. Luna: So the optimal number of drivers is the smallest number that gets you above that 80% threshold. Lucas: Exactly. And that number varies by time of day and day of week. So they'd have a higher guarantee pool for Friday and Saturday nights, and a smaller one for Tuesday afternoons. It was dynamic. Luna: I love that. It's so much more sophisticated than just 'pay drivers to show up.' It's a real operational discipline. Lucas: Absolutely. And I think that's the real lesson for marketplace builders today. The competitive advantage isn't in the idea of subsidizing supply — everyone knows that. It's in the precision of how you do it, the data you use to calibrate, and the discipline to stop when the job is done. Luna: Alright, so to wrap: if you're building a marketplace, think about whether your supply side is homogeneous and mobile. If it is, consider a time-bound guarantee paired with zone-based deployment. If it's not, you need a different approach — maybe exclusive deals or inventory pre-stocking. Lucas: And whatever you do, define your exit metric before you start. Otherwise you'll end up like the startups that burned through venture capital subsidizing both sides indefinitely, with nothing to show for it. Luna: Good advice. That's it for this episode of Marketplace Businesses with Fexingo. We'll be back next week with another deep dive. Lucas: See you then.