Latest / Marketplace Businesses with Fexingo: Two-Sided Networks, Liquidity, and Take Rates / How Marketplaces Use Search Ranking to Control Take Rate
Transcript
- Lucas: If you run a marketplace, the most powerful lever you have for influencing take rate isn't your fee schedule. It's the search algorithm. Luna: That's a provocative claim. You're saying how you rank listings matters more than what you charge? Lucas: Exactly. Because the algorithm determines which products buyers see first, and that directly shapes conversion rates. A seller who ranks higher gets more sales, and the marketplace collects a percentage of each sale. So if you can nudge buyers toward higher-margin listings without raising fees, you effectively increase your take rate without the backlash of a price hike. Luna: But isn't that manipulative? If the algorithm favors certain sellers, doesn't that erode trust? Lucas: It can, if it's done poorly. But the smartest marketplaces align ranking with what genuinely makes buyers happy. For example, Etsy made a change a few years ago where they started boosting listings that were in stock and ready to ship quickly. That sounds innocent, but it had a huge effect on which sellers got traffic. Luna: Right, because previously a handmade seller might list an item but take weeks to produce it. Etsy was essentially deprioritizing made-to-order in favor of ready to ship. Lucas: Exactly. And that shift increased conversion rates because buyers got items faster. But it also meant sellers had to change their behavior—hold inventory, which costs money—or lose visibility. And Etsy's take rate went up because more transactions completed, and because the algorithm tended to favor sellers who could afford to hold inventory, who also tended to be higher-volume sellers paying a higher effective take rate from listing fees. Luna: So the algorithm becomes a way to push the marketplace toward a certain kind of supply without explicitly mandating it. That's pretty subtle. Lucas: And it's everywhere. Airbnb prioritizes listings that enable instant booking. That means guests can book without waiting for host approval, which improves the user experience and increases bookings. But it also means hosts who want to manually approve guests lose visibility. Airbnb's take rate benefits because more bookings happen, and because instant-book listings tend to be professionally managed, which often means higher prices and thus higher absolute fees per booking. Luna: But what about the host who wants to approve guests? They might have a good reason—security concerns, or wanting to know who's staying. If they get pushed down in search, they might leave the platform. Lucas: That's the risk. Over-optimizing for take rate can drive away valuable supply. The key is to balance ranking signals that improve the buyer experience with signals that maintain seller diversity. Airbnb still allows manual approval, but it's deprioritized. The question is: at what point does that hurt the overall marketplace? Luna: Let's talk about the other side: when the algorithm explicitly favors higher-priced listings because they generate more fee revenue per transaction. Is that common? Lucas: It's more common than most marketplaces admit. If you're a platform that takes a percentage commission, you have a natural incentive to show buyers more expensive items. But if you do that too aggressively, you hurt the perception of value. Buyers come to a marketplace expecting a range, and if they only see premium listings, they might go elsewhere. Luna: So what's the right approach? How do you design a ranking system that balances take rate with marketplace health? Lucas: The most durable approach is to rank primarily by predicted conversion probability and buyer satisfaction, not by price. When you rank by conversion, you naturally surface listings that buyers want—which often includes a mix of price points. Then you can layer in business objectives as secondary signals. For example, you might give a small boost to listings that are likely to lead to repeat purchases, or to sellers who provide good customer service. That way, you're aligning take rate with long-term value, not just short-term fee extraction. Luna: That sounds good in theory, but how do you measure 'predicted conversion probability' without biasing toward certain seller types? Lucas: You have to be really careful about your training data. If your model is trained on past transactions, it will naturally favor sellers who have already succeeded, which can entrench incumbents. You need to incorporate exploration—show new sellers to a fraction of users to gather data. And you need to explicitly model fairness constraints: for example, ensure that a certain percentage of search impressions go to new or small sellers. Luna: And that's where it gets complex. You're essentially running an experiment every time someone searches. Lucas: Exactly. And marketplaces that do this well treat ranking as a continuous optimization problem, not a set-it-and-forget-it. They A/B test different ranking schemes and measure not just conversion and take rate, but also long-term seller retention and buyer repeat rate. That's the real meta-lesson: the algorithm is a product, and it needs just as much care as any feature. Luna: Before we go deeper, I want to touch on something you mentioned earlier—the idea that marketplaces don't always tell sellers how the algorithm works. Etsy has faced criticism for being opaque about ranking factors. Is transparency always better? Lucas: I think transparency is a spectrum. If you tell sellers every single signal you use, they'll game it. But if you hide everything, they feel powerless and distrustful. The best middle ground is to explain the principles: we rank based on relevance and customer satisfaction, and here are the top three things you can do to improve your visibility—like ship on time, have good photos, and maintain high ratings. That gives sellers a clear path without exposing the algorithm to manipulation. Luna: And that's actually a form of take rate control too, right? Because if sellers know what to optimize, they'll invest in those areas, which improves the marketplace overall, and the marketplace captures more value. Lucas: Exactly. It's a virtuous cycle. But there's a darker side: some marketplaces have been accused of using ranking to push sellers into paying for advertising. For example, if you reduce organic visibility and then offer promoted listings, that's effectively a way to raise take rate through a side door. Luna: And that can backfire. Sellers might feel extorted, and eventually leave. We saw that with some Amazon marketplace sellers who felt they had to buy ads to get any visibility. Lucas: Right. That's a classic example of over-optimizing take rate at the expense of supply. The best marketplaces use ranking to create a win-win: buyers find what they want quickly, sellers get rewarded for good behavior, and the platform earns its fee fairly. It's a delicate balance. Luna: Speaking of balance, I want to bring up something I've noticed in our own work here at Fexingo. We deliberately keep these episodes ad-free because we believe that helps us focus on what's useful for you, the listener, without any distortion. Lucas: That's exactly right. We'd rather have the conversation be clean and independent. If you value that approach and want to support it, the simplest way is to buy me a coffee dot com slash fexingo. It's a tiny gesture that helps us keep doing this without any outside pressure. Luna: Yeah, no sponsors, no ads, just the content. We appreciate everyone who's chipped in. Now, back to ranking—let's talk about the actual mechanics of how you test a change to the algorithm. Lucas: Sure. So imagine you're a marketplace like Etsy, and you want to test whether boosting listings with faster shipping improves take rate. You'd run an A/B test where a portion of users see the new ranking, and you measure not only immediate conversion and fee revenue, but also whether those buyers come back, and whether the sellers who got deprioritized reduce their listing activity over the next month. Luna: So it's not just about the first click. You have to look at the entire ecosystem response. Lucas: Right. And that's hard because the effects can take weeks to show up. A seller who loses visibility might not delist immediately, but over a month they might list fewer items or leave. That's a lagging indicator, but it's critical. The best marketplaces have dashboards that track seller engagement metrics alongside buyer metrics. Luna: And there's another layer: the algorithm itself can create feedback loops. If you boost instant-book listings on Airbnb, those listings get more bookings and reviews, which improves their ranking further. That's good for those hosts, but it can create a winner-takes-all dynamic. Lucas: Exactly. That's why you need to build in decay or randomization. For example, you might give every listing a small random boost occasionally to ensure that new or less popular listings get a chance to prove themselves. Or you might have a 'freshness' signal that boosts newly listed items for a short period. Luna: So the algorithm is not just about ranking; it's about shaping the entire marketplace culture. One interesting case is how Uber Eats ranks restaurants. They have to balance delivery time, rating, and price. But they also have an incentive to push higher-priced restaurants because they take a percentage. Do they? Lucas: They likely do, but they're careful about it. If you search for 'pizza' and the top results are all expensive gourmet places, you might get frustrated and switch to another app. So they have to keep the ranking relevant. I've seen research suggesting that platform-owned delivery services tend to favor restaurants that use their own delivery fleet, because those are more reliable and the platform captures more value. That's another subtle form of take rate optimization. Luna: So the common thread is: ranking is the invisible hand that shapes both sides of the market. It's the single most powerful tool for influencing behavior, and therefore take rate. Lucas: Absolutely. And the operators who understand that—who treat ranking as a strategic lever, not just a technical detail—are the ones who build lasting marketplaces. It's not about tricking users; it's about aligning incentives so that what's good for the platform is also good for buyers and sellers. Luna: And that alignment is the holy grail. Because when it works, you don't need to raise fees. The growth itself lifts take rate. Lucas: Exactly. And that's the most sustainable path. So if you're building a marketplace, spend less time on your pricing page and more time on your ranking algorithm. It's where the real leverage is.