How Stressline works
Stressline tests healthcare startup theses against 241 documented failures across 15 sub-verticals. Each failure is Gioia-coded with a 3-level causal chain. Three adversarial LLMs — prosecution, defense, and judge — evaluate structural risk. The result is a corpus-grounded diagnostic, not a prediction.
How scoring works →
Bayesian inference, prior calibration, kill thresholds, and why the score is not a prediction.
15 sub-verticals covered
Failure patterns →
The 13 structural failure nodes derived from Gioia coding. Every company maps to exactly one primary node.
13 structural failure patterns
The adversarial engine →
Prosecution, defense, and judge — three independent LLM families arguing every thesis from opposing positions.
Corpus methodology →
Gioia three-level coding, provenance tagging, inter-rater reliability, and data sources used to build the corpus.
Sub-verticals →
Coverage across 15 healthcare B2B sub-verticals. Failure counts, capital destroyed, and top risk patterns per sector.
Limitations →
What the engine cannot do, where corpus coverage is thin, and what a stress-test score does not tell you.
Glossary →
Plain-English definitions for every term used in Stressline reports: failure nodes, kill thresholds, structural risk.
Data Sources: Rock Health, CB Insights, SBIR Phase I awards, company post-mortems, SEC filings, founder interviews, and press archives. All entries Gioia-coded and provenance-tagged. Full methodology in Corpus methodology.