Flatiron Health

Flatiron Health is a New York oncology technology company, founded in 2012, that structures the messy real-world data trapped in cancer clinics into research-grade evidence and sells the electronic health record that captures it. Its two founders had zero healthcare background, and both stories below turn on the same trait: outsiders who treated their ignorance of oncology's conventions as an asset, then used disciplined feedback to find the one product the market actually needed.

Founder-Market Fit: outsiders who turned no medical background into an advantage

The problem. Nat Turner and Zach Weinberg had just sold their ad-tech company Invite Media to Google for roughly $81 million in 2010, at age 24, and had no experience in medicine or oncology. Turner's motivation was personal, not professional: his young cousin Brennan Simkins had been diagnosed with a rare, recurring leukemia in 2008, and Turner saw firsthand how siloed and un-analyzed cancer data was. The obvious objection was that two ad-tech founders had no business rebuilding cancer informatics.

The approach. Rather than hire their way to credibility or defer to industry veterans, they leaned on an outsider's willingness to question constraints that insiders had stopped seeing, a stance Turner argues is a feature: too much domain expertise makes you "so used to the constraints of the industry ... that you can't see through the mess." To close the knowledge gap fast, they worked under Google Ventures partner Krishna Yeshwant, a physician, who introduced them to dozens of oncologists and researchers, and took $8 million in GV seed funding to start helping cancer centers with analytics.

How it solved it. The constraint the insiders accepted was that oncology's records were too messy to become research-grade, so incumbents built workflow software and left the notes as notes. Epic and Cerner sat on far larger volumes of cancer records and never turned them into regulatory evidence; specialist oncology EHR vendors sold charting, not curation. Flatiron treated the mess as the product. The outcome settled the question: in February 2018 Roche agreed to buy the 87.4% of Flatiron it did not already own for about $1.9 billion, on top of roughly $1 billion invested earlier, and Roche said plainly it was buying the real-world data asset to speed its own cancer-drug development. By then Flatiron partnered with over 260 community cancer clinics and 14 of the top 15 oncology-focused pharma companies. A price that size, paid by a pharma major for the data itself, is the counterfactual: the people who knew oncology best had the same records and did not build this.

PMF: distilling the good 10% by owning the data source

The problem. Flatiron launched in 2012 as a big-data analytics offering for cancer centers, but the founders hit a wall that analytics alone could not clear: the real-world oncology data they wanted to analyze barely existed in usable form. The vast majority of cancer patients, roughly 96%, are treated outside clinical trials, and their records sat in unstructured physician notes rather than clean structured fields. An analytics layer on top of nonexistent structured data was the wrong 90% of the idea.

The approach. Consistent with Turner's discipline of "distilling all the information you have, 90% of which will be crap, and finally figuring out what is the good 10%," they concluded they had to own the data pipe itself, not just analyze its output. In May 2014 they acquired Altos Solutions, maker of OncoEMR, the first web-based oncology-specific electronic medical record, alongside a $130 million Series B led by Google Ventures, GV's largest health-IT investment at the time.

How it solved it. Owning the EHR gave Flatiron the source: Altos already served over 1,300 oncology clinicians seeing more than 550,000 unique cancer patients a year. The pipe alone was not the moat, though. What made the data uncopyable was the layer on top: trained oncology nurses doing human chart abstraction against standardized definitions, quality-controlled NLP, and endpoints built to satisfy a regulator rather than a dashboard. Flatiron ran a collaboration with the FDA from 2016 and later published a validated real-world mortality endpoint, work that converts unstructured physician notes into evidence an approval can rest on. That is what Roche paid about $1.9 billion for in 2018, and it is why a well-funded EHR incumbent with more records could not simply have produced the same asset: the differentiator was curation and abstraction at scale tied to FDA-relevant endpoints, not the software. The database grew to millions of active cancer patients across hundreds of sites, the durable product-market fit the analytics-only version never had.