Latest / Marketplace Businesses with Fexingo: Two-Sided Networks, Liquidity, and Take Rates / How Airbnb Uses Dynamic Pricing to Balance Supply and Demand
Transcript
- Lucas: So Airbnb in June 2026 — peak booking season in most of the Northern Hemisphere. Hosts can charge three hundred dollars a night for a one-bedroom in Lisbon, and guests will pay it. But what about Lisbon in February? That same apartment sits empty unless the price drops. And Airbnb has a structural answer: dynamic pricing. Luna: We talk about pricing algorithms a lot on this show. But for a two-sided marketplace, dynamic pricing isn't just about maximizing revenue per transaction — it's about keeping both sides alive during the slow months. Lucas: Exactly. If the market is illiquid on the demand side — no guests searching in February — then supply side hosts get frustrated, they delist, and the platform shrinks. So Airbnb's dynamic pricing is really a liquidity tool dressed as a revenue tool. Luna: Before we go deep — this is the kind of practical operator thinking that keeps this show ad-free. If these conversations are useful for what you're building or running, you can support the show at buy me a coffee dot com slash fexingo. Seriously, it makes a difference. Lucas: It really does. And it means we keep the episodes focused on what works, not on selling ads. So back to pricing — let's talk about Airbnb's Smart Pricing feature, which launched back in 2015 and has been iterated on ever since. Luna: Smart Pricing is basically an algorithmic price suggestion that hosts can opt into. Airbnb's model looks at demand signals — search volume, booking windows, similar listings in the area — and recommends a nightly rate. If the host accepts, the rate adjusts automatically. Lucas: Right. And the key insight is that Airbnb doesn't just set a price and leave it. The algorithm predicts demand curves for specific dates. A Thursday in March might be cheap, but the same apartment on New Year's Eve — that's a different elasticity entirely. The system can recommend a two-hundred-dollar premium for that one night. Luna: But here's the tension: if the algorithm pushes prices too low, hosts revolt. If it pushes too high, guests go to hotels. So the take rate — Airbnb's commission — is directly tied to how well the pricing engine balances those two forces. Lucas: And that's the multiplier effect. A well-calibrated dynamic pricing system increases total transaction volume. More bookings happen because prices match willingness to pay. Airbnb's take rate is roughly 14 to 16 percent on average, so a ten percent increase in bookings means a ten percent increase in revenue — without raising the take rate. Luna: Let's compare with ride-hailing. Uber and Lyft use surge pricing, which is a different animal. It's real-time, often algorithmically triggered by a spike in demand — a concert ending, a rainstorm. Prices go up to attract more drivers to the area. Lucas: Surge pricing is dynamic, but it's reactive, not predictive. Airbnb's system is more like a forward curve: it sets prices for dates weeks or months out, based on historical patterns and current booking velocity. Lyft's system adjusts minute by minute based on live supply and demand imbalances. Luna: Both are solving the same core problem — thin liquidity on one side. For Airbnb, thin demand in February. For Lyft, thin supply during a storm. The pricing signal rebalances the market. Lucas: And both platforms face a trust problem. When a Lyft ride costs triple the normal fare, riders feel gouged. When an Airbnb host sees their price drop forty percent because the algorithm says so, they feel devalued. So the communication around dynamic pricing is almost as important as the algorithm itself. Luna: Airbnb has tried to address this by giving hosts control. Smart Pricing is optional, and hosts can set floor and ceiling prices. If the algorithm suggests a price below the floor, it won't go lower. That preserves the host's sense of agency. Lucas: But here's the interesting thing — data from Airbnb suggests that hosts who opt into Smart Pricing earn about ten to fifteen percent more on average than those who set static prices. The algorithm is better at predicting demand than individual host intuition. That's a data moat. Luna: A data moat that gets deeper with every booking. Every search, every click, every booking adds a signal. A new marketplace trying to compete with Airbnb would need years of transaction data to build a pricing model that works as well. Lucas: Exactly. And it's not just the demand data. Airbnb also sees supply-side data — how many hosts are willing to list at different price points. That's the two-sided network effect. The pricing algorithm benefits from scale on both sides, which makes it harder for a competitor to replicate. Luna: But dynamic pricing isn't a silver bullet. OpenTable, for example, uses a different approach. They don't set prices for restaurants — restaurants set their own. OpenTable's dynamic element is time-based: popular slots cost more points in their loyalty program, but the actual cover price is fixed. Lucas: Right, because OpenTable's marketplace is different. Supply is limited to the number of tables, and demand is highly time-sensitive. A 7 PM Saturday slot is worth more than a 5 PM Tuesday slot. But OpenTable's primary revenue is from a subscription fee plus a per-cover charge, not a percentage of the meal price. So dynamic pricing of the cover charge doesn't directly increase their take rate. Luna: Interesting. So the applicability of dynamic pricing depends on the marketplace's revenue model. If you earn a percentage of transaction value, you want to maximize transaction volume and value. If you earn a flat fee per booking, you just want volume. Lucas: Right. And that's why Airbnb's incentive aligns with dynamic pricing. More bookings at optimal prices = more commission. For OpenTable, dynamic pricing of the reservation itself doesn't move the needle as much, because their take rate is per cover regardless of the meal's cost. Luna: Let's talk about the risks. When Uber first introduced surge pricing, it faced massive backlash. People felt it was price gouging during emergencies. Uber eventually added features like upfront pricing and surge notifications, but the trust damage was real. Lucas: Airbnb has avoided that kind of public blowback partly because the price changes are less dramatic and partly because hosts, not guests, see the price adjustments. But there's a subtler risk: algorithmic collusion. If all hosts in a city use Smart Pricing, they're effectively following the same algorithm. That could lead to coordinated pricing without explicit collusion. Luna: Regulators have started looking at that. If the algorithm effectively sets a floor price for all listings in a neighborhood, it could be seen as price-fixing. The host might not have intended it, but the effect is the same. Lucas: Yeah, and that's a real antitrust concern. But so far, Airbnb's algorithm doesn't force hosts to stay at a minimum. It suggests a range, and hosts can override. So the risk is theoretical, but it's something marketplace operators need to monitor. Luna: Let's bring it back to the operator perspective. If you're building a marketplace today, should you implement dynamic pricing from day one? Lucas: Probably not. Dynamic pricing requires a lot of data to be effective. If you're a new marketplace with a few hundred transactions, an algorithm won't have enough signals to set good prices. You're better off with manual pricing or simple rules — like a fixed percentage discount for off-peak times. Luna: So the threshold is something like ten thousand transactions per month? Or is it more about the diversity of demand patterns? Lucas: I'd say it's more about demand variability. If your marketplace has clear peak and off-peak periods — like a vacation rental platform — then even with moderate volume, a simple rule-based system can help. But if demand is relatively constant, dynamic pricing adds complexity without payoff. Luna: And the take rate matters too. If your take rate is low, the extra revenue from dynamic pricing might not cover the engineering cost. Airbnb's take rate is high enough that the investment pays off. Lucas: Right. So the decision to invest in dynamic pricing depends on three things: demand variability, transaction volume, and take rate. If all three are high, it's a no-brainer. If any is low, you might want to wait. Luna: What about marketplaces where the supply side is professional, not peer to peer? Like, say, a B2B services marketplace. Lucas: In B2B, pricing is often negotiated, not set by an algorithm. Professional suppliers have established rate cards and don't want a platform undercutting them. So dynamic pricing is much harder to implement. You might use it for lead routing — charging different prices for different quality leads — but not for the service itself. Luna: That's a good distinction. In peer to peer marketplaces, supply is more elastic — individual hosts or drivers can adjust prices more freely. In B2B, supply is fixed in the short term, and pricing is relationship-based. Lucas: Exactly. And that's why Airbnb's dynamic pricing is such a great case study. It's a pure peer to peer marketplace with high demand variability, high transaction volume, and a healthy take rate. All the conditions are met. Luna: One more thing — Airbnb has also experimented with dynamic pricing on the guest side, like offering discounts for longer stays or last-minute bookings. That's essentially price discrimination, which is another layer of complexity. Lucas: Yeah, and it works. A guest who books a week-long stay in February is price-sensitive. A guest who books a weekend in August is not. The algorithm can segment by booking window and length of stay, offering different prices to different user segments. That's where the real optimization happens. Luna: So dynamic pricing isn't just about supply and demand in aggregate. It's about micro-segments — specific dates, user types, listing types. The more granular, the better the liquidity. Lucas: Right. And that granularity is a data moat. A competitor can copy the concept of dynamic pricing, but they can't copy the years of booking data that allow Airbnb to predict demand for a specific apartment on a specific Tuesday in March. That's the advantage that compounds over time. Luna: Alright, let's leave it there. Next time, maybe we can talk about how marketplaces handle the opposite problem — too much demand and not enough supply. Lucas: The waitlist problem. That's a good one. For now, if you're running a marketplace, think about whether dynamic pricing fits your liquidity problem. If it does, start simple, gather data, then iterate. If not, spend your engineering dollars elsewhere.