Slow-Rolling a Quality Tier Is the Riskiest Way to Launch One

Based on the research ofFeng, Li, and Shen, "The Ripple Effect of Platform Quality Controls in a Submarket: Evidence from a Secondhand Marketplace," Marketing Science, 2026

Feng, Li, and Shen study a natural experiment on a Chinese secondhand marketplace that opened a quality-controlled submarket inside its mobile phone category. Adopters got a physical quality check, a detailed condition report, standardized listing fields, a recommended price, a post-purchase guarantee, and free delivery. The immediate result is unsurprising: adopters sold more. The two results that are not obvious are what make the paper worth a platform operator's attention. First, sellers who joined the submarket got better at selling everywhere, carrying improved listing habits into their ordinary, non-certified inventory. Second, sellers who never joined did not experience a uniform penalty. They lost ground while adoption was low, then gained it back, and then some, once adoption crossed a threshold. The rollout speed itself, not just the program design, determined how much damage non-adopters absorbed.

A submarket most marketplaces already run

The structure in the paper is not exotic. It is the same structure eBay built when it layered Authenticity Guarantee on top of its existing peer-to-peer marketplace. Authenticity Guarantee routes eligible purchases through a third-party inspection before the item reaches the buyer, and eBay covers the cost. It started with sneakers and watches and has since been extended to handbags, fine jewelry (in partnership with the Gemological Institute of America), trading cards, and streetwear. Crucially, it is gated by price and category: Jordan and Yeezy sneakers at $150 and up, watches at $2,000 and up, handbags over $500 from a defined list of luxury brands. Below the threshold, or outside the covered brand list, an otherwise-identical item sits in the ordinary marketplace, competing against inventory that now carries a trust badge it cannot get.

Vinted runs a lighter version of the same idea with its Item Verification service: buyers can pay a small fee to have an eligible designer item checked by an in-house authentication team before it ships, currently limited to a handful of European markets and a defined brand list. In both cases, a platform did not convert its whole marketplace to a higher standard. It carved out a submarket, gated by price, brand, or geography, and left the rest of the inventory to compete under the old rules. That is exactly the setting Feng, Li, and Shen study, and it is common enough that most large resale marketplaces are running some version of it right now.

The carryover effect: certification as an on-the-job course

The paper's carryover finding is the more surprising of the two, because the intuitive model of a certification program treats it as walled off: it should help adopters and, at most, compete with everyone else. Instead, sellers who went through the submarket's standardized process, structured photos, condition disclosure, guided pricing, appear to have absorbed habits that improved their performance back in the regular marketplace too. The certification program functioned as an unplanned training course in how to list well.

You can see the shape of the same incentive in Amazon's Renewed program. Sellers must maintain an Order Defect Rate under 0.8 percent, offer at minimum a 90-day warranty, and demonstrate substantial refurbishment volume to qualify. Carvana's model compresses the same idea further: every vehicle it sells passes a 150-point inspection and carries the same reconditioning standard, because Carvana is the seller of record on all of its inventory rather than a marketplace hosting independent sellers. Neither company has published data on whether meeting a certification bar changes seller behavior in uncertified listings elsewhere on the same platform, so treat the parallel as illustrative rather than as a second data point. What Feng, Li, and Shen add is the part nobody had measured directly: the discipline a quality submarket imposes does not stay contained to the submarket. It travels with the seller.

The U-shaped trap for everyone else

The nonadopter result is where the paper earns its counterintuitive angle. Early in a rollout, when few sellers have joined the certified tier, nonadopters are simply outcompeted: buyers now have a visibly better, guaranteed option, and demand shifts toward it. That is the losing half of the U. But as adoption keeps rising, a second force takes over. Buyers' overall confidence in the category improves, because a large and growing share of what they see carries the platform's quality stamp, and that improved perception spills into listings the platform never touched. Once adoption crosses a threshold, nonadopters are better off than they were before the submarket existed at all.

The practical consequence is that the "stuck in the middle" zone of a rollout, not the launch and not full maturity, is where nonadopters take the worst hit. eBay's own rollout illustrates how long that zone can last in practice: sneakers launched in 2020, watches followed in 2021, handbags over $500 arrived later that year, fine jewelry with GIA came in 2022, streetwear in 2023. Within any one category, an item just under the dollar threshold, or from a brand not yet on the eligible list, sat next to authenticated competitors for years before the program matured or expanded to cover it. A staged, cautious expansion is the intuitive, risk-averse way to launch a quality tier. The paper's evidence says caution is exactly what prolongs the period in which the sellers you have not yet reached are worst off.

What a fast tip looks like, and what a capped one looks like

Airbnb's Guest Favorite badge, introduced in November 2023, is worth holding up as a contrast rather than a confirmation. It is awarded to listings in roughly the top 1 to 2 percent by guest ratings, review volume, and reliability, evaluated over the trailing two years. That ceiling is structural: because the badge is defined by percentile rank rather than by meeting a fixed bar, its "adoption rate" cannot rise toward the tipping point the way submarket enrollment can. A badge capped near the top 1 percent stays scarce by design, so the category-wide trust spillover that eventually rescues nonadopters in the Feng, Li, and Shen setting has no mechanism to arrive here; the badge is built to confer exclusivity, not to lift the median listing. That is a legitimate design choice, but it is a different one from a submarket that is meant to expand, and platform teams should be clear about which model they are actually running before they promise either outcome to the sellers not yet inside it.

A quality tier that grows slowly does not protect the sellers left outside it. It only lengthens the time they spend losing.

eBay's authenticator network, and Vinted's in-house verification team, are real capacity constraints, not design choices, and they cap how fast either company could plausibly expand its submarket even if it wanted to sprint past the tipping point. That tension is the honest takeaway for anyone building a similar program: the paper's evidence argues for speed, but inspection capacity, fraud review, and brand-eligibility vetting are genuine bottlenecks that do not scale as fast as enrollment could. The fix is not to skip the constraint. It is to budget for it explicitly, size the inspection or verification pipeline to the adoption curve you actually want to hit, and treat "how fast can we get to the tipping point" as a resourcing question decided before launch, not a pace that gets set by default caution.