Version the holdings, inputs, scenarios, mappings, and methods; retain prior reports; and explain material revisions through controlled reruns. Separate proposed changes, reviewed changes, and authorized use rather than treating every saved file as an approved result.
Assign responsibility to specific objects
A stress report depends on several different objects: holdings, market data, mappings, scenario assumptions, valuation methods, and interpretation. Assign an owner to each rather than naming one person responsible for an opaque total. The owner should be able to explain the relevant evidence and limitations. Review responsibility should include the authority to question an assumption, not merely confirm that a document exists.
The Basel Committee's stress-testing principles discuss objectives, governance, resources, methodology, and documentation as connected parts of a framework. This guide adapts that organizational idea to report interpretation without claiming bank-level compliance. The practical question is whether someone can identify an error, assess its consequence, and prevent the unsupported result from being reused while the issue is still unresolved or awaiting review.
Preserve the report as it was understood
Retain the original report with the exact inputs and calculation version that produced it. Record the holdings and market-data timestamps separately from the report-generation time. Preserve scenario shocks and mapping decisions rather than only their human-readable names. A later download from the same source may contain revisions, so a source address alone is not enough to reconstruct what an earlier report actually used.
Keep proposed and reviewed changes distinguishable. Saving a revised mapping or scenario does not establish that its implications were assessed. Where the process uses manual adjustments, retain the amount, rationale, author, and review state. A reproducible record should show both the computational result and any judgment layered onto it, including limitations that were known when the report was originally interpreted.
Hypothetical calculation: bridge a revision
Suppose an invented prior report shows USD 400,000 of loss magnitude. Holding market starting values and the valuation method fixed, updated holdings increase that loss by USD 60,000. A scenario revision adds USD 30,000. A documented data correction then reduces the loss by USD 20,000. The revised loss magnitude is USD 470,000, a net increase of USD 70,000 from the preserved original.
The table uses positive numbers for loss magnitude, unlike a signed profit-and-loss ledger. Each step is a controlled rerun in the stated order, not an independently additive causal estimate. With nonlinear models, changing the order may change intermediate contributions. These hypothetical amounts demonstrate a revision bridge only; they are neither actual portfolio losses nor evidence that any particular governance process has been completed.
| Hypothetical revision step | Change in loss magnitude | Resulting loss magnitude |
|---|---|---|
| Original report | Starting point | USD 400,000 |
| Updated holdings | +USD 60,000 | USD 460,000 |
| Revised scenario | +USD 30,000 | USD 490,000 |
| Data correction | -USD 20,000 | USD 470,000 |
Distinguish corrections from new assumptions
A correction fixes something that was wrong under the original specification, such as an inverted exchange-rate quote or duplicate holding. A new assumption changes the specification, such as a different liquidation horizon or a broader scenario. These events deserve different labels. Calling a favorable assumption change a correction can conceal judgment, while calling a genuine error a routine update can hide its consequences.
For a material correction, identify affected reports and interpretations. Preserve the original with a clear superseded or corrected status, and provide the corrected result without erasing the audit trail. Explain whether the issue changes rankings, threshold assessments, or only display precision. The purpose is to make downstream readers aware of the correction and its scope, not to manufacture a history in which the mistake never occurred.
Control changes before looking for favorable outcomes
State the reason for a proposed methodological change and its evaluation criteria before selecting among results. Compare old and new methods on frozen inputs and include cases where the proposed change performs poorly or becomes inapplicable. A lower reported loss is not, by itself, evidence of a better model. Evaluate coverage, consistency, conceptual suitability, and the sensitivity of conclusions to the revised assumption.
Where an urgent workaround is needed, document its scope, owner, limitations, and review condition. Do not let a temporary manual override become an undocumented permanent rule through repeated reuse. Keep the unadjusted calculation available alongside the workaround. A reader should be able to tell whether a result comes from the ordinary method, a reviewed exception, or an unresolved analytical limitation.
Protect the meaning of aggregation and thresholds
The Basel Committee's risk-data aggregation principles provide an official reference for dependable risk information and reporting. In a portfolio review, a basic application is consistent definitions across versions: the same meaning of covered value, the same sign convention, and an explicit denominator. Changing those definitions can move a headline percentage even when the underlying currency loss remains exactly the same.
Thresholds also need versioned definitions and owners. A loss budget expressed against total starting value is different from one expressed against modeled value or available cash. Record when a threshold changes and why. Do not revise a threshold after seeing an unfavorable result and then describe the original assessment as passed. Any new assessment should retain its own assumptions and be distinguishable from the earlier decision.
Close revisions with evidence and a readable explanation
A revision is ready for interpretation when the change is described, its effect is reconciled, relevant checks are complete, and remaining limitations are visible. Name the reviewed version and preserve the comparison with its predecessor. An issue log should distinguish open questions from completed corrections. Avoid declaring the entire process validated merely because one numerical discrepancy was fixed or one reference test passed.
The final note should answer three practical questions: what changed, why it changed, and what conclusions can now be supported. Keep the explanation understandable without requiring the reader to inspect source code. This governance discipline supports an educational stress-report process; it does not authorize investment actions, confer regulatory approval, or turn hypothetical illustrations into verified records of historical portfolio performance.
Review checkpoints
- Preserve earlier reports and the exact inputs that produced them.
- Label corrections separately from changes in assumptions or scope.
- Bridge material revisions and retain the meaning of denominators and thresholds.
Source notes
Continue the analysis
Read the stress report before accepting its headline
Audit scope, coverage, sign conventions, denominators, loss bridges, and limitations before interpreting a portfolio stress total.
Explain a scenario loss without inventing a unique cause
Reconcile position and factor contributions, expose interactions and ordering choices, and distinguish attribution from causal explanation.
Challenge the model, not just its arithmetic
Separate implementation tests from economic validation, construct informative challengers, and document unresolved limitations and permitted uses.
First edition, 16 October 2026. Hypothetical values are teaching inputs, not observed portfolio or provider results.