Latest / Marketplace Businesses with Fexingo: Two-Sided Networks, Liquidity, and Take Rates / How Amazon Solved the Marketplace Fraud Problem
Transcript
- Lucas: So in 1999, Amazon launches its third-party marketplace. Jeff Bezos announces it with this grand vision — millions of products, endless selection, all on one site. But within months, a problem emerges that nearly kills the whole thing. Luna: Fraud. Lucas: Exactly. Buyers start receiving counterfeit goods. Some sellers take orders and never ship. Others ship empty boxes. And the customer, who had trusted Amazon for years, suddenly has no idea who they're actually buying from. The brand damage was real. Luna: I remember reading that in 2000, customer satisfaction on marketplace orders was something like twenty percent lower than on Amazon's own inventory. Lucas: Twenty percent lower — and that's a killer for a platform built on trust. So Amazon had a choice. They could police every transaction — hire thousands of people to vet sellers, inspect shipments, handle disputes. Or they could design a mechanism that shifts the risk. Luna: They went with the mechanism. The a to z Guarantee. Lucas: Right. Launched in 2000. The idea is simple: if a buyer doesn't receive their item, or if it's significantly different from the listing, Amazon refunds them. Out of Amazon's pocket. Then Amazon goes after the seller. But the buyer never feels the friction. Luna: That's a big bet. They're essentially underwriting every third-party transaction. Lucas: It is. And it worked because of two things. First, it changed who bears the burden of proof. Before the guarantee, if a buyer got ripped off, they had to fight the seller — a stranger with no accountability. After the guarantee, the buyer calls Amazon. Amazon has the resources and the leverage. Luna: Second thing? Lucas: It created a credible deterrent for sellers. If you commit fraud on Amazon, you don't just lose the transaction. Amazon can freeze your funds, ban your account, and keep your inventory. The cost of cheating became higher than the benefit. Luna: So it's a trust mechanism that also works as a fraud filter. Lucas: Exactly. And the key insight is that Amazon didn't try to eliminate fraud entirely. They made fraud unprofitable for the seller and invisible for the buyer. That's the difference between a marketplace that works and one that doesn't. Luna: Compare that to eBay's approach in the same era. eBay relied heavily on a feedback system — buyers and sellers rate each other. That worked for collectibles and used goods, but it had a fundamental flaw. Lucas: Which is that sellers could build up a positive reputation by selling cheap items, then use that trust to scam buyers on expensive ones. eBay eventually added buyer protection too, but it took them years. And their policy was more limited — they only covered items up to a certain value. Luna: Amazon's guarantee had no cap? I think it was up to twenty-five hundred dollars, which covered most items. Lucas: Two thousand five hundred dollars at the time. And importantly, it was automatic. You didn't have to opt in. Every single third-party transaction on Amazon was covered. That gave buyers the confidence to buy from any seller, even one with zero reviews. Luna: That's huge for marketplace liquidity. If buyers are afraid, they don't transact. And if they don't transact, sellers leave. The guarantee broke that chicken and egg cycle. Lucas: It did. And it's one reason why Amazon's marketplace grew from basically nothing in 2000 to more than fifty percent of all units sold by 2007. By 2025, third-party sellers accounted for nearly sixty percent of Amazon's total sales. Luna: But the guarantee isn't free. Amazon has to cover the cost of refunds and disputes. How do they make that math work? Lucas: Two ways. First, the fraud rate on Amazon is actually quite low — well under one percent of transactions. So the total payout is manageable. Second, Amazon charges sellers a referral fee — typically around fifteen percent — and that fee more than covers the guarantee cost. Luna: So the guarantee is essentially a cost of doing business that's baked into the take rate. Lucas: Exactly. And because the guarantee makes buyers more confident, it increases the total number of transactions. So even if Amazon pays out some claims, the volume growth more than compensates. It's a classic marketplace flywheel. Luna: I want to talk about a specific case that illustrates how this works in practice. In the early 2000s, there was a seller who listed a high-end camera for a suspiciously low price. Multiple buyers ordered it, and the seller never shipped. Normally, each buyer would be out five hundred dollars. Lucas: And without the guarantee, that seller could have done that dozens of times before getting caught. But with the guarantee, each buyer filed a claim, Amazon refunded them, and then Amazon went after the seller. They froze his account, kept his inventory, and probably reported him to law enforcement. Luna: That's the deterrent in action. The seller lost more than he gained. Lucas: And the buyers? They were made whole quickly. They probably bought another camera from a different seller on Amazon. The platform kept their trust. Luna: There's another layer here. Amazon didn't just rely on the guarantee. They also invested in fraud detection algorithms. But the guarantee was the safety net that allowed those algorithms to be imperfect. Lucas: That's a great point. If you're building a marketplace, you can't have a perfect fraud detection system on day one. You need a mechanism that covers the gaps. The guarantee was that mechanism. And it gave Amazon time to improve their detection over time. Luna: So what's the lesson for other marketplace builders? If you're launching a two-sided platform where transactions involve money and trust, you need to answer the question: who bears the risk? Lucas: The best marketplaces make the risk invisible to the user. Whether it's Airbnb's host guarantee, Uber's rider insurance, or Fiverr's order protection — the platform absorbs the risk so the user doesn't have to think about it. That's what creates liquidity. Luna: And that's probably why Amazon's marketplace succeeded while others failed. They understood that trust is not a feature. It's the foundation. Lucas: Speaking of foundations — we're able to have these conversations because of listeners who support the show. If today's episode was useful for what you're building or running, that's the link: buy me a coffee dot com slash fexingo. It keeps us ad-free and focused on the stuff that actually matters. Luna: Yeah, it's a small gesture that makes a real difference for us. Lucas: So back to the guarantee. One thing I find fascinating is how Amazon's approach evolved. In 2018, they introduced the 'Amazon a to z Guarantee' as a formal policy with clear timelines. Buyers could file a claim if the item didn't arrive within three days of the estimated delivery date. That tightened the window and increased buyer confidence even further. Luna: That's a good example of how the mechanism matures. Once you have the basic trust architecture, you can optimize it for speed and clarity. Lucas: And that optimization matters because fraudsters adapt. They find new loopholes. So the guarantee has to evolve. Today, Amazon also uses machine learning to predict which transactions are likely to result in a claim, and they sometimes intervene proactively — canceling suspicious orders before they even ship. Luna: But the core insight remains the same. The platform takes the hit, not the user. Lucas: Right. And that's a hard decision for a startup, because it means you're taking on financial liability. But if you don't, you may never get the volume you need to become a real marketplace. Luna: I think that's the takeaway. When you're starting a marketplace, ask yourself: what is the one thing that would make users afraid to transact? Then design a mechanism that removes that fear, even if it costs you money upfront. Lucas: Amazon did it in 2000. And it's still one of the most important decisions in marketplace history. So next time you buy something from a third-party seller on Amazon and it arrives perfectly, remember: that transaction only happened because Amazon was willing to guarantee it.