Overview
CrowdMed was a crowdsourced diagnosis marketplace where medical detectives reviewed hard cases and a prediction market ranked likely diagnoses, from YC Winter 2013. CrowdMed was founded in 2012 to use crowdsourcing and prediction markets for difficult medical cases. [1]
CrowdMed was right that rare cases need more minds, but the product's trust boundary was wrong. The crowd could suggest possibilities; the health system still controlled diagnosis, reimbursement, and liability. The outcome shows the difference between product need and standalone company structure.
Founding Story
CrowdMed's origin was specific rather than generic. CrowdMed was founded in 2012 to use crowdsourcing and prediction markets for difficult medical cases. [1] The early product insight was this: CrowdMed attacked a real failure in healthcare, but the model asked patients to trust strangers and pay out of pocket for advice that still had to be validated by clinicians.
Jared Heyman started CrowdMed after his sister spent more than three years and over $100,000 seeking a diagnosis. [2] That setup mattered because the company was not selling a thin interface. It asked users to trust a new workflow for a decision that already had entrenched habits.
CrowdMed told WIRED: "Instead of relying on individual physicians, CrowdMed harnesses the collective intelligence of hundreds of 'medical detectives'" [4] Jared Heyman told TEDMED profile: "medical detectives could help solve difficult medical mysteries online" [2] Those quotes define the company better than a feature list: CrowdMed tried to compress an emotionally noisy decision into a structured product.
The founding gap is also worth stating. Public sources do not fully explain every early team decision, board conversation, or financing constraint. The available record is strongest on product shape, funding or acquisition events, and the strategic reason the idea ended up inside a larger system.
Timeline
- 2012: Founded by Jared Heyman, Axel Setyanto, and Jessica Greenwalt.
- 2013-04: Launched public beta at TEDMED.
- 2014-05: Raised $2.4 million seed round with strategic celebrity and venture investors.
- 2016: Charged patients $149 to $749 per month for difficult-case review.
- 2023-10: Website appeared down.
What They Built
CrowdMed built a crowdsourced diagnosis marketplace where medical detectives reviewed hard cases and a prediction market ranked likely diagnoses. The first user experience was designed to replace an inefficient default: patients who had already spent years, money, and emotional energy on unresolved diagnoses; later enterprise prospects included insurers and self-insured employers. The product's promise was not novelty for its own sake. It was a cleaner decision loop.
The key workflow had three parts. First, the user supplied context. Second, the system turned that context into a ranked recommendation, assessment, or plan. Third, the user or buyer acted on the output with less search cost. That pattern is visible across the public facts: CrowdMed launched public beta at TEDMED in 2013. [4]
The product differed from alternatives because it packaged judgment, not just information. Directories, search results, and generic software leave the hard ranking work to the user. CrowdMed tried to own the ranking layer. In complex medical case navigation, that is valuable only when the ranking is trusted and tied to a transaction or operating workflow.
The company raised $2.4 million from Patrick Dempsey, NEA, Andreessen Horowitz, Greylock, SV Angel, Khosla Ventures, and Y Combinator. [3] That evidence suggests the product had real substance. The harder question was whether that substance created a standalone distribution advantage.
Market Position
Target Customers
Patients who had already spent years, money, and emotional energy on unresolved diagnoses; later enterprise prospects included insurers and self-insured employers.
Market Size
The public record does not provide a clean market-size model for CrowdMed. That absence matters. The company operated in a large category, but broad category size was not the binding constraint. The binding constraint was whether enough users would change behavior through this specific workflow and whether the company could capture revenue at the point where value was created.
Competition
The relevant competitors were not only startups with similar copy. They were incumbents that controlled demand, data, reimbursement, purchase intent, or workflow. Included Health, Grand Rounds legacy services, Mayo Clinic second opinions, Fabric, Navina, and AI medical-record summarization tools.
CrowdMed's position was therefore structurally awkward: it had a sharper product surface than many incumbents, but the incumbents had more of the transaction context. That is the recurring pattern in the company's outcome. The product became more valuable when attached to a larger data or distribution base.
Business Model
CrowdMed charged patients a monthly fee for case attention and paid detectives for useful contributions. The model asked the sickest users to pay while they were least able to evaluate quality. Public sources do not disclose enough revenue detail to calculate reliable unit economics. The absence of that data is itself a signal: when a startup's strongest public evidence is product quality, funding, or acquisition language rather than durable revenue, the analyst should be careful about assuming a repeatable go-to-market engine.
An inference is still fair. CrowdMed's economics depended on reducing decision cost enough that a buyer would pay repeatedly. In complex medical case navigation, that means the product needed either recurring workflow usage or a direct share of downstream transaction value.
Traction
Public sources cite hard-case activity and investor backing, but the company did not publish durable revenue or enterprise penetration data. The DO reported that patients paid $149 to $749 per month after a two-week free trial. [5] The available traction evidence supports product demand, not necessarily venture-scale independence.
Post-Mortem
trust was the product
Rare-disease patients did not only need ideas. They needed their physician, insurer, and family to believe the next step was safe and worth pursuing. The DO reported that each case was overseen by a licensed physician and that about 60% of medical detectives worked in or studied healthcare. [5] This is the primary mechanism: the product's judgment layer was valuable, but the strongest owner of that layer was the party with the underlying context.
the payer was misaligned
Out-of-pocket pricing from $149 to $749 per month selected for desperate patients, while the savings from faster diagnosis often accrued to insurers or employers. The team addressed the problem by building a structured workflow rather than a passive directory or content product. That helped users understand the output, but it did not erase the cost of trust, distribution, or buyer education.
the crowd needed clinical rails
The prediction market produced ranked suggestions, but healthcare requires accountable documentation, credentials, privacy controls, and follow-up. Wikipedia reports that the CrowdMed website appeared down as of October 2023. [6] The counterargument is that acquisition can be a success. That is true. But for Startups.RIP, the useful question is narrower: why did the product not keep compounding as an independent company? The answer is not simply execution. The market rewarded the capability when it moved closer to distribution, data, or institutional trust.
The non-obvious lesson is that judgment products do not fail only when the judgment is wrong. They fail when the company cannot own the moment where the judgment becomes action.
Key Lessons
- CrowdMed's core product should be evaluated as a workflow, not a feature. The company won attention by reducing a messy decision into a ranked next step.
- The strongest 2026 rebuild should attach to the buyer's existing data and action path. A standalone destination repeats the hardest part of the original go-to-market problem.
- The company-specific risk is trust. complex medical case navigation users need to believe the recommendation, simulation, plan, or diagnosis before they act.
- The acquisition or inactive outcome does not invalidate the idea. It shows where the idea needed to live.