By DEEKSHA HEGDE
I wanted to draw parallels between the two. Structural facts kept lining up in ways I couldn’t rule out, and eventually I gave up trying.
Function’s product is an app: You pay $365 a year, go to a Quest Diagnostics location, get tested for more than 160 biomarkers twice a year, and receive notes written by a doctor interpreting your results. Laboratory tests are completely outsourced. Function is the top layer: panel design, member experience, medical note generation, longitudinal tracking.
FunctionHealth raised $298 million at a valuation of $2.5 billion in November 2025. At 25x revenue, the market is clearly not buying a lab reseller. You’re buying into the data flywheel: longitudinal biomarker histories that increase in clinical value over time, aggregated across hundreds of thousands of members into a data set that health plans, pharmaceutical companies, and AI developers can’t otherwise build. A member with four years of data cannot switch to a cheaper competitor without losing the trend. Unit economics work if the interpretation layer scales without proportionally scaling the workforce, which is what Medical Intelligence Lab, its generative AI model launched in November 2025, is designed to do. The feature is also building toward a B2B enterprise channel, positioning the product as a way to keep employees “healthy, focused and ready to perform.”
It meets a pressing need for specific people: the health-conscious, the health optimizers, the people who saw their loved ones diagnosed too late, the people who don’t wait for a diagnosis before they start providing care. These are people that the rest of the industry has mostly left alone. I wrote earlier this year, in an article on Hinge Healthon the prevention paradox: the employer return on investment model is structurally blind to the member who benefits most from early intervention. The feature completely skips the employer ROI story, charges the member directly, and doesn’t try to prove a CFO case it can’t make. Still.
Take apoB, short for apolipoprotein B, one of Function’s most publicized biomarkers, a better predictor of cardiovascular risk than the LDL test on which most annual physicals still rely. He Swedish AMORIS Cohort Study They followed 137,000 people for an average of 17.8 years and found that elevated apoB separates cases from controls about 20 years before a major cardiovascular event, with the gap becoming more acute during the decade closest to the event. Detect it, act accordingly (statins, dietary changes, closer monitoring) and you can prevent a heart attack. That is the true clinical value.
Now run the employer’s calculations. The average tenure of the private sector in the United States is 3.5 years. The employer funding the Feature membership will almost certainly not insure this person when that cardiovascular event materializes. The avoided cost is borne by whoever covers them in fifteen years. The CFO writing the check today doesn’t get any of that. What Function is really selling to employers now is a wellness benefit. Higginbotham, an insurance broker, announced a association in January 2026 offering Function to employees at $334 instead of $365. A discount of $31. You don’t build a $2.5 billion company on a gym stipend.
What health plans and pharmacy partners would actually pay for doesn’t exist yet: actuarial risk models and research data sets based on years of longitudinal biomarker data at scale. The D2C subscription is the data acquisition engine for that product. It all depends on the steering wheel working long enough to close the gap.
That’s where the 23andMe comparison clicked for me.
23andMe was a D2C healthcare data company whose valuation was based on a data monetization thesis: pharmaceutical research partnerships, data licensing. Kit sales alone couldn’t justify it. It was not a HIPAA-covered entity and collected sensitive health data under a privacy policy that reserved the right to transfer it in the event of a sale or bankruptcy. Consumer marketing, however, never said any of this. 23andMe led with ascendancy, making it viral, emotionally catchy, and FDA-adjacent rather than FDA-regulated. The health reports were the upsell. “Contributing to science” was how they framed the data acquisition: altruism, not a transaction. He GSK Association ($300 million for research access to the genetic database) was disclosed in the terms of service, for people who read the fine print. When the business failed, data was a salable asset and Regeneron got it at a garage sale.
The function is structurally similar, with greater risks in each dimension. D2C Health Data Company, Data Monetization Thesis: Business Risk Models, Research Partnerships, AI. It is also not a HIPAA covered entity. It does not bill insurance and is not a covered healthcare provider, which means the legal protections that most members assume apply to your health data do not. He Same loophole in bankruptcy transfer that caught the members of 23andMe would also catch those of Function.
23andMe’s revenue model was broken from day one: buy a kit once and you’re done. The feature has recurring subscriptions and real switching costs. 23andMe also imploded while trying to become a therapeutics company, a pivot that required capabilities completely foreign to what it had. The feature’s monetization path is at least adjacent to what it already does.
But the regulatory exposure is identical and that window is closing. He The FTC updated its health breach notification rule in April 2024, explicitly extending it to health apps and D2C platforms, and has already used it against GoodRx, Premom, and BetterHelp. HIPRAThe Health Information Privacy Reform Act, introduced in November 2025, would extend HIPAA-equivalent obligations to exactly the business function category.
The data contained in the function makes this exposure worse. 23andMe contained genetic predisposition data: probabilistic, future-oriented, a risk modifier. Function biomarker data is up to date: thyroid function, metabolic status, hormonal profile, cardiovascular markers, updated twice a year. More actionable, more temporally accurate, more directly useful to anyone making decisions based on their health.
Which leads to a problem that the feature hasn’t fully named. Your D2C brand is built on a specific promise: Your data is yours, and we help you understand it. That’s what got 200,000 members. 23andMe acquired its first millions the same way, using aspirational identity marketing to capture data assets, while creating real value required a B2B transaction that consumer marketing never prepared the user for. Function’s member base is more sophisticated than 23andMe’s: they are people who already believe their data has clinical value. Function marketing hasn’t yet built the narrative architecture to invite them to participate in a transaction they’re already in.
The playbook exists. Oura chose the NBA and UCSF as research partners. These are institutions that made their members feel like they were participants in something elitist and scientific. Whoop post findings using aggregated member data. In both cases, the same underlying data transaction reads as a contribution to improving identity because the framework was built into the brand from the beginning, not buried in the terms of service. Function’s biomarker data is more clinically sensitive than wearable data like HRV and sleep stages, and its B2B thesis is more loaded, meaning the manual itself needs more rigor in execution. Partner selection is branding work: the first disclosed research partnerships will lay the groundwork for everything. Publishing the findings normalizes the data relationship while building the case for product validity. Function needs before health plans issue a check.
The feature is already being performed in parts of this manual: the NBPA Association in February 2026 it will bring biomarker testing to professional athletes and the brand signal is real. But a distribution agreement is not a research relationship. What is missing are structured research partnerships to produce and publish findings, the kind that turns distribution into evidence and evidence into retention.
There is another layer that neither Oura nor Whoop have built. Proactive data governance is what makes the contribution framework credible and not just a brand copy. The feature currently describes itself as “HIPAA-aligned,” following key requirements of the Security Rule on a voluntary basis. That’s the gap in a sentence. Security practices without legal commitment do not close the bankruptcy transfer loophole, do not satisfy IRB consent requirements for future pharmaceutical partners, and do not provide insurance partners with the security they need to license a data set without regard to regulatory exposure. Adopting HIPAA-equivalent data handling before HIPRA is passed and framing it as a product decision rather than a compliance response protects the flywheel from being dismantled before it matures.
The function has a window to act on its own terms. 23andMe waited for regulators to write the rules and became the warning that gave those rules their name. The function could write the playbook instead.
Deeksha Hegde is a bioengineer who writes about health technology and digital health positioning in Substack.
