One Badge, Many Winners: The Spillover Effects of Platform Endorsement
Platform endorsement increases sales for endorsed items—but the bigger story is what it does for everything else.
Platform Papers is a blog about platform competition and Big Tech. Prominent academics discuss their latest research. The blog is linked to platformpapers.com, a repository that collects and organizes academic research on platform competition.
By Mimansa Bairathi, Xu Zhang, and Anja Lambrecht.
If you’ve ever searched for a product on Amazon and seen an “Amazon’s Choice” badge, you’ve encountered platform endorsement. Intuitively, the badge helps the endorsed item—it signals quality and draws attention. But what happens to everything else? Do merchants whose items don’t receive a badge lose out? And does the platform itself come out ahead?
These questions aren’t just academic. Merchants worry that endorsement diverts demand away from their non-endorsed listings. Regulators have raised concerns about whether such practices are anticompetitive. And platforms themselves need to know whether endorsement grows the overall pie or merely redistributes it.
Our research, published in Marketing Science, investigates exactly this, using data from a large-scale field experiment conducted by one of the world’s largest online freelance platforms.
The Experiment
The platform ran a 27-day field experiment involving nearly 600,000 registered users. A proprietary algorithm, accounting for price, average rating, and number of completed orders, identified high-quality services as eligible for endorsement. Users were randomly assigned to either a treatment group, who saw an endorsement badge on qualifying listings in their search results, or a control group, who saw no badge at all. Everything else, including the ranking algorithm, service listings, prices, remained identical across groups.
Critically, only about 1.2% of all services on the platform were eligible for endorsement. This means that for most searches, no badge appeared even for treatment group users, and the vast majority of services users encountered were unendorsed. This is the setting in which we can cleanly examine spillover effects on unendorsed items.
The Headline Finding: A Rising Tide
The most striking result is not what the badge does for endorsed services—that part is intuitive. Exposure to platform endorsement increased clicks on endorsed services by 25% and orders by 40.6%. What is surprising is what the badge does for everything else.
Users who were exposed to an endorsement badge also increased their search and purchases of unendorsed services. Specifically, the endorsement led to a 4.1% increase in impressions of unendorsed services, a 2.0% increase in clicks on unendorsed services, and a 2.2% increase in orders of unendorsed services. Putting this in perspective: because there are so many more unendorsed services on the platform than endorsed ones, 66.7% of the total increase in orders stems from unendorsed services. The badge lifts the sales of the entire marketplace.
At the platform level, the total number of orders increased by 3.1%, with average order prices remaining unchanged—meaning revenue grew by approximately the same amount.
Why Does This Happen?
The positive spillover to unendorsed services raises an obvious question: why? We investigated three potential explanations.
Perceived quality change. Platform endorsement may signal to users that the platform actively curates its offerings. By making high-quality services more visible, endorsement may shift users’ beliefs about the general quality of services available—not just the endorsed one. If users expect to find better services, they search more, and the extra search translates into more purchases across the board. We find strong support for this mechanism. Users who were exposed to an endorsement badge in one product category went on to browse and purchase more in other categories where they were not exposed to any badge at all—a result hard to explain by anything other than a platform-wide shift in perceived quality.
Attention spillovers. Perhaps the badge simply attracts attention, which spills over to nearby listings? We find little support for this. If attention were the driver, we would expect the biggest gains to accrue to services placed spatially close to the endorsed service in search results. In fact, the reverse is true: services located near an endorsed service receive fewer clicks and orders, while services located farther away benefit more. This is consistent with limited attention being diverted toward the endorsed service in its immediate vicinity, not with attention spreading outward.
Novelty. Maybe the badge is simply novel, and users explore more out of curiosity? We find no evidence of this either. If novelty drove the effect, we would expect it to wear off on subsequent exposures to endorsement. Instead, the effect on clicks becomes more pronounced over repeated exposures—consistent with a gradual updating of quality perceptions, not a novelty response that fades.
Not All Unendorsed Services Benefit Equally
The spillover effects are positive on average, but there is meaningful heterogeneity depending on an unendorsed service’s position relative to the endorsed one.
Services located in the same row as the endorsed service, or in the row immediately above or below, suffer from the badge. Users’ limited attention is drawn to the endorsed service, crowding out interest in what sits directly beside it. Services farther away, however, benefit from the overall quality perception upgrade without suffering the attention trade-off.
Price similarity tells a related story. Services priced similarly to the endorsed service see no significant increase in orders—arguably because, at a similar price point, users who have been drawn to the endorsed service simply buy that one instead. Services priced differently, and therefore less directly competitive with the endorsed item, capture more of the spillover benefit.
Who Responds Most?
Platform endorsement is not equally effective for all users. We find that users with a higher propensity to purchase—measured by their past purchase history and whether they self-identified as business users—respond more strongly to endorsement. Business users, who purchase roughly 1.6 times more than personal users in the control group, see their total orders increase by about 8.5% more than personal users in response to endorsement.
This matters for platform strategy. When piloting or scaling endorsement programs, platforms can maximize their return by prioritizing endorsement in categories that attract high-propensity users. The badge is most effective where it meets users who are already inclined to buy.
About the research: We study a 27-day field experiment run by a major online freelance platform involving 598,772 users. The platform randomly assigned users to see or not see an endorsement badge on qualifying service listings in their search results. Using Poisson regression with inverse probability weighting to estimate average treatment effects on the treated, we find that exposure to endorsement increases clicks on endorsed services by 25%, orders of endorsed services by 40.6%, and—critically—orders of unendorsed services by 2.2%, with 66.7% of the total order increase coming from unendorsed listings. We trace the mechanism to an improvement in users’ perceived quality of services on the platform.
What It Means
For platforms, the results demonstrate that even a small design intervention—a single badge on a search results page—can meaningfully shift platform-wide revenue. Endorsement does not merely redistribute demand from unendorsed to endorsed items; it expands total demand. And the insight that high-propensity users are most responsive helps platforms allocate endorsement efforts more efficiently.
For merchants whose items are not endorsed, the results offer genuine reassurance. The fear that a competitor’s badge will cannibalize your sales is largely unfounded—unless your listing sits directly adjacent to the endorsed one. On average, the badge raises the tide.
For regulators, who have grown increasingly concerned about whether platform endorsement practices are anticompetitive, our findings complicate the standard narrative. In contexts where platforms endorse items on the basis of quality, the practice does not appear to harm unendorsed sellers. The European Union’s platform-to-business regulations, designed to address fairness concerns about prominence, may need to account for the possibility that selective endorsement can be ecosystem-positive rather than zero-sum.
There are, of course, limits to what one experiment in one setting can tell us. Endorsing items on the basis of something other than quality—say, payment—could produce very different dynamics. And the long-run effects on seller entry, pricing, and quality investment remain open questions. But as a first systematic account of what platform endorsement does to the whole ecosystem, not just the lucky few who receive the badge, the evidence is more optimistic than the conventional wisdom suggests.
This post is based on research published in Marketing Science and is included in the Platform Papers references dashboard:
Bairathi, M., Zhang, X., & Lambrecht, A. (2025). The value of platform endorsement. Marketing Science, 44(1), 84-101.
Platform Upates
EU-DPRN program live (📑): The premier European academic conference on digital platforms is set to take place this year from June 11-12 at Aalto University, Finland. The annual summit of the Europan Digital Platform Research Network features presentations by academics and practitioners from Europe and beyond. The program features a mix of research presentations, practitioner panels, and social events!
EU’s DMA may fail to ignite multi-homing (📑): Two years in, the Digital Markets Act faces backlash for possibly degrading core services rather than dismantling tech monopolies. Gatekeeper firms argue that forced interoperability and sideloading may fracture user experience and compromise security, without meaningfully empowering rival ecosystems.
Apple taps Intel to diversify its silicon supply (📑): Ending its exclusive reliance on TSMC, Apple agreed for Intel to manufacture its advanced chips. This strategic multi-sourcing bypasses vendor lock-in amid an AI capacity crunch, validating Intel’s pivot toward a manufacturing platform through its foundry model, open to external customers.
Alphabet commits $190bn on AI infrastructure (📑): Following a 63% surge in cloud revenue, Alphabet raised its 2026 capex to $190bn. While scaling its data centers to alleviate computing constraints, Google is doubling down on its generative AI enterprise solutions, embedding them as an integral part of its vertically integrated software stack.
Google targets orbital data centers to bypass terrestrial friction (📑): To bypass severe grid constraints, Google is negotiating with SpaceX to launch space-based AI data centers. By advancing its native Project Suncatcher and engaging rival launch providers simultaneously, Google aims to avoid vendor lock-in in extraterrestrial logistics.
Meta harvests employee data for an AI agent moat (🔈): While investing in its $135bn AI-related infrastructure, Meta is tracking workers’ keystrokes to train autonomous digital agents. Deploying these assistants across its 3.5 billion-user network aims to supercharge targeted ads and remain indispensable in the generative AI era.
eBay declines GameStop’s $55.5bn omnichannel takeover bid (📑): The online marketplace rejected an unsolicited $55.5 billion acquisition offer from GameStop. Instead of pursuing this offline integration, eBay opted to protect its pure-play digital marketplace, focusing its turnaround on defending against Amazon and Temu.
Nintendo hikes global prices to defend hardware margins (📑): Nintendo is raising prices across its Switch consoles and online subscriptions. By elevating hardware and digital entry costs, the gaming giant may test the inelasticity of its user base, positing that exclusive IP would sustain cross-side network effects despite steeper financial barriers.
Deliveroo bundles multi-store orders and cuts delivery fees (📑): Deliveroo now lets users combine items from three different shops into a single order for no extra delivery fee. By eliminating cross-vendor transaction friction, the platform aims to drive larger basket sizes and discourage multi-homing among price-sensitive customers.
Platform Papers is published, curated and maintained by Joost Rietveld.