This instructor-only report is bound to the outcome-blind October 28, 2015 packet. It tests the decision process against later evidence without rewriting what was knowable at the cutoff.
Outcome in brief
The frozen recommendation—to pause patient testing and retail expansion, preserve records and specimens, and require independent assay-by-assay and end-to-end workflow validation—was process-sound. It did not require a finding that every Theranos test was inaccurate or that anyone had committed fraud. It responded to asymmetric patient harm, a wide evidence gap, partner hesitation, and fresh quality-system observations while preserving the option to relaunch a validated subset. @judgment.theranos.outcome.process-quality
The public record after the cutoff developed along four dimensions:
- An independent black-box study of 60 healthy adults found significant interservice variability in 15 of 22 measurements and said Theranos flagged out-of-range results 1.6 times as often as the other services. The authors could not isolate collection, processing, instrumentation, timing, or their interaction as the source. @claim.theranos.outcome.jci-variability @src.theranos.outcome.jci-2016
- CMS determined that the Newark laboratory had condition-level deficiencies and that specified hematology practices posed immediate jeopardy. It later imposed sanctions after concluding the deficiencies had not been corrected. @claim.theranos.outcome.cms-immediate-jeopardy @claim.theranos.outcome.cms-sanctions
- Walgreens terminated the relationship and closed all 40 Theranos Wellness Centers in its Arizona stores. @claim.theranos.outcome.walgreens-terminated @src.theranos.outcome.walgreens-termination-2016
- By September 2018, Theranos said it intended an assignment for creditors and dissolution, reported at least USD 60 million of unsecured claims and about USD 5 million of estimated net remaining cash, and expected no shareholder distribution. @claim.theranos.outcome.wind-down
What the technical evidence establishes
The later independent comparison is strong evidence of service-level disparities, but it is not a controlled decomposition of the proprietary device. The blood-collection method, shipment to a central facility, sample preparation, timing, conventional versus new instrumentation, and absent Theranos technical replicates remained confounded. The correct conclusion is narrower than “all Theranos technology failed”: the commercial service lacked enough independently reproducible, workflow-specific evidence to justify broad patient use. @claim.theranos.outcome.jci-variability @conflict.theranos.outcome.technical-source-attribution
The FDA record at the cutoff also matters. One HSV-1 assay had a defined intended use and reported performance package. That evidence remains positive for that narrow assay; it does not validate hundreds of other assay-workflow combinations. Later appellate language about the number of small-sample assays and the subset run on Edison is relevant current record, but it must not be smuggled into the 2015 decision packet. @claim.theranos.cutoff.one-assay-clearance @claim.theranos.outcome.device-scope
CMS's immediate-jeopardy determination and sanctions are regulator actions about a laboratory and specified condition-level failures. They carry more adjudicative weight than a Form 483, but they still do not prove that every reported result was wrong or isolate one device as the sole cause. The durable analytical unit is the assay plus specimen, device, site, operator, process, and quality system—not a company-wide slogan.
Legal status is evidence, not decoration
The SEC complaint alleged that Theranos raised more than USD 700 million while making materially misleading claims and that its proprietary analyzer performed only about 12 tests from a menu exceeding 200. Those are complaint allegations in this corpus. Holmes's civil judgment was entered by consent without a general admission or denial and included specified remedies. Neither record is silently upgraded into litigated findings. @claim.theranos.outcome.sec-allegations @claim.theranos.outcome.sec-consent-remedies
The current appellate record likewise requires defendant-specific precision: Holmes was convicted on four investor-related counts and acquitted on patient-related counts, while Balwani was convicted on all submitted investor- and patient-related counts. The verdicts cannot be collapsed into one generalized finding about every claim, test, executive, or victim. @claim.theranos.outcome.verdict-scope @conflict.theranos.outcome.legal-scope
Arizona's consent terms, SEC remedies, criminal restitution, regulator sanctions, and creditor outcomes measure different legal and economic objects. The report preserves those distinctions instead of using the magnitude of later remedies as retroactive proof of a technical proposition.
Financial record: deliberately nonadditive
The deterministic table records six sourced amounts: the 2017 exchange-offer original issue price, Arizona consumer restitution, the Holmes SEC civil penalty, issuer-estimated remaining cash, issuer-reported minimum unsecured claims, and criminal restitution. @table.theranos.outcome.nonadditive-recorded-amounts
They must not be summed. The USD 582.2 million Form D figure was the aggregate original issue price of securities tendered by existing preferred holders in an exchange offer—not verified new cash. The USD 452 million restitution order is a criminal remedy involving identified victims, not enterprise value. Remaining cash and minimum claims were management estimates in a wind-down letter, not audited liquidation endpoints. @claim.theranos.outcome.exchange-not-cash @claim.theranos.outcome.restitution @claim.theranos.outcome.nonadditive-financial-record
No audited bridge in the corpus allocates value destruction among technical validation, laboratory operations, misrepresentation, financing, partner decisions, regulation, litigation, and later management. Warren therefore abstains from a target price, total enterprise-loss estimate, or causal percentage. @judgment.theranos.outcome.no-single-loss-bridge
Best-supported causal model
The primary hypothesis is a validation-governance gap:
flowchart LR
A["Broad patient-facing claims"] --> B["Evidence and quality-governance gap"]
B --> C["Unresolved technical and laboratory risk"]
C --> D["Regulator and partner response"]
D --> E["Trust, financing, and operating contraction"]
E --> F["Creditor wind-down"]
This is a directional mechanism, not a quantified loss model. At the cutoff, one narrow clearance coexisted with scientific criticism, reported partner hesitation, and newly public FDA observations. Later independent, regulator, partner, and wind-down records are consistent with the risk that unsupported scope allowed technical and quality uncertainty to propagate through the enterprise. @hypothesis.theranos.validation-governance-gap
The rival hypothesis deserves real weight: collection, processing, conventional laboratory operations, financing, partner behavior, and later legal events may explain material portions of the outcome independently of the proprietary device. The black-box study's explicit limitations and the narrow HSV-1 evidence prevent a universal device-failure conclusion. @hypothesis.theranos.measurement-operations-and-other-causes
The evidence supports the primary hypothesis at moderate confidence, not a monocausal verdict. Board minutes, raw full-menu validation, device-to-result mapping, quality logs, partner contracts, audited financials, and capitalization records remain missing.
Counterfactual and signals
The feasible cutoff action was a controlled pause: preserve evidence, stop unsupported patient workflows, independently validate every offered combination, close quality-system issues, notify affected clinicians and patients where required, and relaunch only a passing subset. @counterfactual.theranos.independent-assay-gates
Four signals were available before the decision:
| Signal | Lead time | False-positive boundary |
|---|---|---|
| One narrow clearance versus a broad menu | About 3.5 months | Other assays may have had nonpublic evidence or conventional workflows |
| Missing independent scientific transparency | More than 4 months | Confidential or unpublished data can still be valid |
| FDA quality and validation observations | Hours | A Form 483 is not a final determination and may be remediated |
| Reported Walgreens expansion pause | 4 days | Partner diligence can be temporary and reversible |
See @signal.theranos.clearance-menu-gap, @signal.theranos.scientific-transparency, @signal.theranos.quality-system-observations, and @signal.theranos.partner-pause.
Transferable rules
@rule.patient-facing-assay-validation-gate is a candidate safety-and-governance rule: when patient-facing claims outrun independent assay and workflow evidence, pause unsupported use and require predeclared validation and quality gates. It is not a fraud classifier, and passing the gate does not establish attractive unit economics or valuation.
@rule.preserve-legal-and-measurement-status is a candidate analytical rule: preserve whether a proposition is an observation, allegation, consent term, acquittal, conviction, appellate holding, company estimate, or audited fact. Aggregate monetary values only after deterministic reconciliation proves compatible scope and no overlap.
Most important lessons
- Match validation to the marketed claim. One assay, one specimen, or one device configuration cannot validate a heterogeneous service menu.
- Underwrite the full workflow. Collection, transport, preparation, instrumentation, software, operators, quality controls, and result correction all affect patient outcomes.
- Use gates before patient scale. Predeclared independent tests preserve option value while limiting asymmetric harm.
- Treat missing transparency as a diligence trigger. It supports a pause and evidence demand, not an automatic accusation.
- Keep regulator records in status. Inspectional observations are not final findings; later CMS determinations and sanctions have their own scope.
- Keep legal propositions defendant- and count-specific. Pleadings, consent judgments, acquittals, convictions, and appellate holdings are not interchangeable.
- Do not manufacture a loss bridge. Exchange values, remedies, creditor claims, cash estimates, and restitution are not additive.
- Separate process quality from hindsight. The pause was justified by cutoff uncertainty and patient-risk asymmetry, not by foreknowledge of later outcomes.
- Abstain when valuation inputs are missing. Technical conviction cannot substitute for verified cash, burn, liabilities, capitalization, and contracts.
Final boundary
This case does not establish that every Theranos assay was inaccurate, that every service difference came from Edison, that every public claim was adjudicated false, or that technical validation alone caused the enterprise failure. It establishes a narrower and more transferable principle: patient-facing scale should stop when marketed scope outruns independent workflow-specific evidence, and both legal status and financial measurement must survive intact from source to conclusion.