The five dimensions
The same headline traction can come from different expansion configurations. PRISM examines the structure behind that traction, rather than treating the existence of a US pipeline as sufficient evidence of readiness.
- Timing — when capital meets the market.
- Trigger — what initiated the expansion.
- GTM motion — how demand converts to revenue.
- Execution — whether infrastructure can carry the motion.
- Signal misread — where traction is mistaken for conversion.
From expansion profile to a Risk Verdict
The public deliverable comprises a structural risk score and expansion archetype, calibrated outcome probabilities, and an assessment of capital consequences and the corrective window.
The website describes nine documented expansion archetypes. The public methodology is an explanation of scope, not a complete specification of model features, coefficients, source records or proprietary implementation.
Keep the evidence populations separate
- Case log: the current website reports 3,232 documented cases. This is a record count, not a client count or a count of prospectively verified companies.
- Classification: the published 84.4% accuracy applies to negative-outcome classification in the top quartile, with 1.967× lift above a naive baseline.
- Statistical model: the homepage reports XGBoost cross-validated AUC of 0.9594. AUC is a discrimination metric, not a percentage probability of an individual company's success.
- Survival analysis: the homepage reports Cox model concordance of 0.7904 for estimating time to structural friction. Concordance is not an outcome probability.
- Prospective validation: locked cohorts and their verification windows are described separately in the original homepage record.
A record that changes as outcomes resolve
Every resolved outcome feeds back into the case log. PRISM retrains. The record compounds.
That feedback describes how the record is maintained; it does not mean every model update improves every metric or removes uncertainty. A dated deliverable remains a snapshot of its own assessment.
What this methodology does not establish
A risk classification is not a certainty. Historical cases may not represent a new company or a changed market, and model metrics do not imply a guaranteed individual outcome.
This public explanation is not an independent audit or a reproducibility package. Read the exact cohort, metric and date attached to a claim; do not merge figures from different records.
The Risk Verdict is not a substitute for legal, tax, immigration, accounting or regulated investment advice.