Analyze
How likely, how severe, how sure, and whether the method can be trusted. 9 components, chapter 16 of the book.
The scenario triplet
Every analyzed risk is recorded as a scenario, a likelihood and a consequence, each with its uncertainty; likelihood is a probability, a frequency or a calibrated word, never a bare adjective; consequence is stated per objective; a margin is reserved for unknown risks.
Source: Risk management handbook [11]
Honest scales
Every label maps to a versioned numeric band; consequence is assessed per objective and the highest severity governs; every estimate carries a precision field; probability words carry their bands; a matrix is used for communication only, and only after its cells pass a consistency test.
Source: Probability and impact literature; Cox's critique [12]
Action Priority
Severity, occurrence and detection are rated one to ten; priority is read from a table in which severity dominates; a severity of nine or ten requires action or a signed acceptance by someone with authority; the product survives only as a legacy figure; ratings are revisited after action with history kept.
Source: Failure mode and effects analysis, harmonized handbook [8]
Simulation that proves itself
Every simulated result reports convergence with a standard error, the sampling scheme, a validation against a closed-form case, a heavy-tail check, a global sensitivity ranking by share of variance, and the seed and model version; a result missing any of these is a draft.
Source: Open engineering course [26]
Composite indices
Indicators are normalized to zero to ten; dimensions are aggregated geometrically so that a zero zeroes the score; a reliability index from missing and aged data is published beside every score; the index runs on a cadence matched to what it measures, with three clocks kept apart.
Source: Humanitarian risk index methodology [7]
Regulatory model discipline
A model is legitimate only while named tests pass: backtesting with zones and a fallback, attribution per unit of accountability, outlier tests against a stated base; whatever can be summed is summed once across all books and entities against one limit set in advance, with every excess recorded; parameters are the organization's own.
Source: Market-risk and banking-book standards [5]; credit principles 2025 [19]
Integrated stress and reverse stress
One scenario moves losses, income, capital or cash and liquidity together, year by year, under rules fixed in advance; an inventory records scenarios, owners and last run; sensitivity, scenario and reverse stress test are distinguished; a management action counts only with a prior decision, a trigger and a feasibility test under the stress itself.
Source: European stress-test methodology and guidelines [25]
Climate and physical scenarios
Reference scenarios on a long and a short horizon; transition and physical variables kept apart, acute and chronic damages not summed; physical scenarios parameterized as hazard by code, exposure, vulnerability and capacity; sector mapping joins scenario to exposure; every input carries version and retraction status.
Source: Climate scenarios [13]; hazard profiles and disaster framework [24]
Structural models
Where failure logic can be drawn, fault and event trees with common-cause failure and human reliability; minimal cut sets listed; importance measures rank controls by contribution to the top event; where the logic cannot be drawn, the record names the method used instead and why.
Source: Probabilistic risk assessment procedures guide [11]