The Verified Badge Isn't Free. It Bills the Poorest Patients.

Based on the research ofZhan, Zhang, and Fu, "Information Disclosure via Platform Endorsement in Online Health Care," Journal of Marketing Research, 2026

What Haodf's Endorsement Actually Buys

Zhan, Zhang, and Fu study Haodf (好大夫在线, "Good Doctor"), one of China's leading third-party online doctor-consultation platforms, founded in 2006 and home to roughly 280,000 registered licensed doctors by 2024, more than 10 percent of all hospital-affiliated licensed doctors in the country, about 73 percent of them from top-tier public hospitals. The platform has served 89 million cumulative patients and facilitated around 50 million consultations in 2023 alone, split across two very different lanes. In paid consultations, a patient picks the doctor directly and 95 percent of requests are accepted within 12 hours. In free consultations, a patient submits a request into a public pool with no doctor selection at all, demand so far outstrips supply that over a third of requests go unanswered for more than a week, and the resulting uncertainty means free care is used overwhelmingly by underprivileged patients, a pattern the authors confirmed directly with the platform's director and its doctors.

Using a unique platform dataset and a generalized synthetic control design, the authors track what happens when Haodf endorses a doctor with a featured badge. The badge is not neutral. Over the full post-treatment period, endorsement drives a 19.7 percent increase in paid-service quantity (p = .007) and an 8.7 percent increase in paid-service price (p = .047), average price rising from 102 to 111 yuan per consultation, and average weekly paid consultations climbing from 27.2 to 32.6 for an endorsed doctor. Patients respond to the badge exactly as designers hope: it reads as a quality signal and shifts demand outward. Doctors respond too, gradually raising prices over the first six months before the new price level stabilizes.

The same endorsement, over the same period, produces a 35.2 percent decline in free-service quantity (p = .001), weekly free consultations per endorsed doctor falling from 9.4 to 6.1. Because doctors voluntarily choose which free requests to pick up from the public pool, this decline is not patients turning away; it is doctors reallocating their own finite hours. Combining both service types, endorsed doctors do work more overall, total weekly output rises 5.7 percent, from a pretreatment base of 36.6 consultations, but the composition of that work tilts hard toward what pays.

The Ledger, Itemized

The paper does not stop at percentages; it prices the trade. Across the 156 doctors newly endorsed in its sample, the paid-side gains translate into roughly 44,000 extra paid consultations a year, a 1.5 percent lift in total platform paid volume, worth about $1.1 million in additional annual platform revenue. Each newly endorsed doctor personally gains about 45,313 yuan, roughly $6,971, in additional annual income, equivalent to nearly half the average annual salary of a full-time doctor in China. Endorsement, in other words, is a real and substantial payday, for the platform and for the doctor.

Set against that is the free-service side of the same ledger: about 27,000 fewer free consultations a year from these same doctors, a 3.3 weekly decline per doctor that adds up to a 3.8 percent platform-level drop in free care in 2019. Valuing each forgone consultation at the 111-yuan price endorsed doctors now charge, the authors estimate roughly 3 million yuan, about $500,000, in forgone care value, transferred away from the patients who relied on it and toward the doctors and platform now competing for paying customers. The authors add a sharper reference point: households receiving China's subsistence allowance earn less than 4,200 yuan a year, meaning a single paid consultation with one of these leading doctors would cost such a household over 2.6 percent of its annual income, precisely the population the free-service queue exists to serve, and precisely the population now waiting longer for it.

The paper's heterogeneity analysis adds a design-relevant twist: doctors it classifies as "high-prosocial," based on above-median free-service provision before endorsement, largely refrain from raising their paid prices after being endorsed, while "low-prosocial" doctors capture more of the price premium. Endorsed doctors also maintain or improve measured service quality despite the higher paid volume, this is not a story of endorsed doctors getting sloppier. It is a story of the same fixed set of hours being repriced and reallocated, with who gets endorsed determining how much of the bill lands on the patients who cannot pay it.

The badge did not ask a single doctor to work less for the poor. It simply made working for the paying patient valuable enough that the free queue lost the doctor's time by default.

Every Badge Economy Runs a Version of This Trade

This is not a China-specific or health-specific artifact of one platform. Joost Rietveld, Robert Seamans, and Katia Meggiorin, writing in Strategy Science, studied what happened when Kiva, the microfinance lending platform, introduced social-performance badges for a subset of its microfinance-institution partners. Certified institutions did gain more borrowers, lenders, and funding, exactly as a trust-signal story predicts, but they also reoriented their loan portfolios to better match what the badge rewarded, shifting the composition of who they lent to rather than simply doing more of the same lending. A certification badge, their findings suggest, functions less like a costless label and more like a governance lever that reshapes complementor behavior toward whatever the badge measures and rewards.

Western health-care marketplaces show that this outcome is a design choice, not a law of nature. Zocdoc's own explanation of its ranking, "How Search Works," states plainly that the platform verifies provider credentials and specialties before listing them, but does not let doctors pay for better placement in organic search results, search rank instead follows patient-specific factors like appointment availability and insurance fit, and Zocdoc's business model charges providers per booking rather than per placement. That is a structural decision to keep the endorsement-like signal (verification) separate from the demand-shifting lever (ranking), which is exactly the coupling that generates Haodf's equity cost. It does not prove Zocdoc has no analogous dynamic buried elsewhere in its marketplace, only that the specific mechanism the Chinese study isolates is not baked into its stated search design.

Labor marketplaces show the badge machinery in a different professional context. Upwork awards its "Top Rated" badge to the top 10 percent of freelancers, requiring a Job Success Score of at least 90 percent sustained over 13 of the last 16 weeks, and steers client attention toward badge holders. Whether Top Rated freelancers subsequently shift their limited hours away from lower-paying or pro bono work in a way that mirrors Haodf's endorsed doctors is a reasonable hypothesis given the same underlying logic, finite attention, a badge that concentrates demand, a provider who must decide whose request to accept, but it is a hypothesis, not a finding either Upwork or an independent field study has quantified. Treat it as illustrative of the mechanism's generality, not as an established result.

The Rule for Anyone Shipping a Badge

A verified, featured, or endorsed tag is not a free quality signal layered harmlessly on top of a marketplace. It is an allocation mechanism, and allocation mechanisms have losers as well as winners. Zhan, Zhang, and Fu show the winners clearly, patients get an honest quality signal, endorsed doctors and the platform get more revenue, and measured quality does not fall. They also show the loser clearly: patients who could never pay in the first place lose access to a queue that was already failing to meet demand before the badge arrived, and lose it precisely because the badge worked as intended.

Before a platform ships an endorsement, verification, or featured-status feature, the question worth asking is not "does this help patients (or buyers, or clients) choose better", the Haodf data says badges do that. The question is: whose existing time or capacity does this feature reallocate to produce that improvement, and who was depending on the capacity it takes away. Haodf's own numbers say the answer is rarely nobody.