Case file
- What happened: Theranos claimed its Edison blood-testing devices could run hundreds of tests from a single finger-prick drop of blood. The technology never worked reliably and produced inaccurate results.
- Scale: Roughly $700 million in investor capital raised; peak valuation around $9 billion. Partnerships with Walgreens and Safeway were built on the premise.
- Root cause: No valid design verification or clinical validation. Failed Edison results were secretly replaced by commercial analysers while patients, investors, and regulators were told otherwise.
- The bill: Felony fraud charges against CEO Elizabeth Holmes and COO Ramesh Balwani. Company dissolved. Patients received false or unreliable medical results from a device that, for most tests, was not doing the work.
Every quality professional has seen a process bypass. An operator skips a drying step to hit takt. A supplier quietly runs parts on an unqualified machine because the approved line is down. We catch these with layered process audits, gemba walks, the occasional whistleblower on the shop floor. What I had not encountered before reading the public record on Theranos was the bypass becoming the entire operating model—scaled to nine billion dollars, with real patients on the receiving end.
The situation
Elizabeth Holmes founded Theranos in 2003. The pitch: replace venous blood draws with a finger-prick and a compact desktop analyser running hundreds of tests from a few microlitres. The Edison device—small, automated, patient-friendly. Investors committed roughly $700 million. Walgreens and Safeway signed on for retail deployments. The board featured former cabinet secretaries and generals. By 2014, Theranos was valued at about $9 billion.
The core technology did not work.
How it unfolded
The Edison had fundamental engineering problems. Pipetting nanolitre volumes of blood is mechanically and chemically brutal. Dilution errors, evaporation, inconsistent cartridge manufacturing—all produced unreliable results. Internal complaints were raised and suppressed. Employees who flagged issues were managed out or threatened with litigation. Rather than halt and fix, Theranos began running most patient tests on commercial analysers from established manufacturers while presenting results as Edison output.
John Carreyrou's investigation in the Wall Street Journal broke the story in October 2015. CMS inspections followed. Walgreens pulled out. In 2018, Holmes and Balwani were indicted on federal wire-fraud charges. Holmes was convicted and sentenced to over eleven years in prison.
Root-cause anatomy
The technical failure is well documented in public reporting. Blood testing from capillary samples is inherently harder than from venous draws. Small volumes amplify error. Hemolysis during finger-prick collection degrades samples. The Edison's robotics were not precise enough for the chemistry involved.
This is the kind of failure surface a disciplined PFMEA would have mapped. Insufficient sample volume: what is the severity, what is the detection rating? Pipetting variation above threshold: same question. Either these were never asked formally or their answers were buried. The organisational layer compounded the technical one. Holmes built a culture of secrecy, surveillance, and retribution against dissenters. The board lacked diagnostic-device expertise. There was no independent quality voice with real escalation authority.
Where the quality system failed
APQP was never meaningfully applied. Design verification—the gate where you prove your device meets design inputs under controlled, repeatable conditions—was faked or skipped. Clinical validation, the medical-device equivalent of a PPAP, was circumvented. Theranos exploited a regulatory classification that allowed certain tests to run as Laboratory Developed Tests, dodging the FDA device-approval pathway that would have demanded transparent evidence.
Change control was meaningless when the actual change was swapping an unproven Edison for a Siemens analyser in another room. Proficiency testing—the external benchmarking every credible clinical laboratory submits to—was gamed or avoided. CAP and CLIA audits that should have caught discrepancies were managed through restricted access and stage-managed demonstrations.
When the quality function cannot stop the line, it is not a function—it is a decoration.
What would have caught it
The simplest control is the one Theranos fought hardest to avoid: independent proficiency testing. Send identical samples to Theranos and to peer laboratories using standard methods. Compare. The Edison would have failed on day one, and it would have been visible to regulators.
Pre-registered validation protocols with acceptance criteria—reviewed and signed off by independent clinical reviewers, not internal stakeholders whose compensation depended on a passing result. A PFMEA examined by competent external scientists who understood microfluidics and clinical chemistry. The failure modes were not exotic; they were classical. Unrestricted facility access for inspectors: no demo rooms, no scripted tours, real CAPA closure on every complaint traceable to root cause. And a culture where the quality manager can escalate to the board without being fired. The employees who tried to raise concerns internally did everything right. The system punished them for it.
My take
I have spent twenty years building and auditing quality systems across automotive and aerospace. The Theranos pattern is familiar in miniature. An operator on a Friday afternoon re-runs a failed test on a different machine and reports the better number. A shift supervisor knows the drying oven runs cold but has not submitted a deviation because production targets are non-negotiable. A supplier changes a sub-process without a new PPAP because the launch window is tight and nobody asked.
Small bypasses. We catch them with layered process audits, cross-shift gemba walks, QRQC stand-ups, and a culture where quality can stop shipment without career consequences. The cost of poor quality in my world is a field return, a warranty claim, an audit finding. At Theranos, the CoPQ was patients receiving false cancer markers, incorrect cholesterol readings, unreliable medication dosing from a machine they were told existed but, for most tests, did not. The scale is different. The mechanism is identical: a gap between the validated process and the actual process, maintained through power asymmetry.
What this means on your floor
- If your design validation was conducted in-house only, with no independent verification, you do not have validation—you have a claim.
- A PFMEA that never triggers a design or process change is paperwork, not engineering. It must be a decision tool.
- Your audit function needs an escalation path that bypasses the CEO. If it does not, your first real problem will be the one nobody told you about.
- The first time someone is disciplined for raising a quality concern, you have built a concealment system. This is the most dangerous organisational failure mode in any industry.
Theranos did not collapse because blood diagnostics are hard. Blood diagnostics are hard and thousands of laboratories do the work honestly every day. Theranos collapsed because leadership treated quality as an obstacle to vision rather than the mechanism that makes vision real. Nine billion dollars of valuation, a board stacked with political figures, a magazine cover—none of it substituted for a validated design and a transparent clinical trial. No amount of capital ever can. The quality system is not the obstacle to your business model. It is the only thing that proves your business model exists at all.
