Glossary

Stressline uses 13 specialized terms across its methodology. Every term is defined below in plain English — no jargon, no circular definitions. The corpus these definitions reference contains 241 confirmed failures. For context on how these terms connect, see the methodology hub.

Causal chain

The Gioia three-level sequence that maps raw observations to a structural failure node. Level 1 captures facts from primary sources — post-mortems, investor write-ups, press coverage, SEC filings. Level 2 groups those facts into structural patterns through constant comparison. Level 3 assigns the pattern to a single structural failure node. Every corpus entry has a full causal chain recorded for auditability.

Concentration ratio

The proportion of a failure node's entries that come from a single sub-vertical. A high concentration ratio means the failure pattern is strongly associated with one sector. A low ratio means the pattern appears broadly across healthcare B2B. Concentration is monitored to prevent any single sector from dominating a node's pattern definition.

Consensus node

The structural failure node that appears most frequently among corpus entries that structurally resemble a thesis, as determined by the structural matcher. A consensus node requires at least 3 matching entries in the relevant sub-vertical before it can serve as the primary prosecution charge. Nodes with fewer than 3 matches are flagged as having insufficient precedent.

Corpus match

A corpus entry that shares structural dimensions with the thesis under analysis. Matching is computed via embedding proximity across buyer type, mechanism, and status quo disruption — not keyword matching. The number of corpus matches determines the evidence depth tier: 10 or more matches is STRONG, 5 to 9 is MODERATE, and fewer than 5 is THIN.

Evidence depth

A classification indicating how much structural precedent the corpus contains for a specific thesis. STRONG means 10 or more structural matches — enough for high confidence. MODERATE means 5 to 9 matches — directional but not statistically saturated. THIN means fewer than 5 matches — early signal only; findings may shift as the corpus grows.

Failure node

A structural failure mechanism that recurs across multiple healthcare startups in the corpus. Each node represents a distinct category of systemic cause — not a symptom, not a proximate trigger, but the underlying structural mechanism. Examples include reimbursement dependency, buyer fragmentation, and CAC payback crisis. Every corpus entry maps to exactly one primary failure node.

Gioia coding

A qualitative research methodology adapted from Gioia, Corley, and Hamilton (2013) for systematic causal analysis. Applied here as a three-level process: first-order codes capture observable facts from primary sources, second-order themes identify structural patterns through constant comparison, and aggregate dimensions map themes to the failure taxonomy. The methodology ensures consistent coding across all corpus entries.

Kill threshold

The Bayesian posterior probability at which the structural evidence is strong enough to classify a thesis as KILLED. The threshold is calibrated against the sub-vertical base rate and corpus match strength. A thesis crosses the kill threshold when the evidence from corpus precedents outweighs the structural differentiators identified in the defense case.

Prosecution / Defense / Judge

The three roles in the adversarial analysis engine. The prosecution presents the strongest evidence-based case for failure, constrained to charges backed by corpus precedents. The defense argues structural differences from known failures using surviving company data. The judge reconciles both cases and renders a verdict. Three independent LLM families — OpenAI, Google Gemini, and Anthropic — are assigned one role each to ensure genuine adversarial independence.

Shannon entropy

A statistical measure of distribution diversity applied to the failure node distribution across the corpus. High entropy means failures spread across many nodes in roughly equal proportions. Low entropy means a small number of nodes dominate. Stressline monitors entropy at each corpus update — the corpus must maintain entropy at or above 70% of its theoretical maximum to prevent any narrow range of failure patterns from dominating the analysis.

Stressline score

The structural risk score assigned to a thesis after adversarial analysis. The score measures how closely a thesis's structural profile resembles companies that failed in similar ways. It is not a probability of failure — it is a measure of pattern proximity to the failure corpus. A high score means strong structural overlap with documented failure modes, not a guaranteed outcome.

Structural risk

Risk that arises from the inherent design of a business model — how it interacts with buyers, payers, regulators, and clinical workflows — rather than from operational execution or market timing. Structural risks are the kind that would persist even if the team executed flawlessly. Stressline specifically measures structural risk, not execution risk, team risk, or market timing risk.

Sub-vertical

A segment of healthcare B2B with distinct buyer dynamics, regulatory structures, and failure patterns. Stressline tracks 15 sub-verticals including clinical workflow, revenue cycle, prior authorization, telehealth infrastructure, and digital therapeutics. Sub-verticals with 20 or more confirmed failures are classified as saturated, meaning the corpus has enough density for reliable structural signal. Sub-verticals below that threshold produce directional findings only.