Build an evidence ledger for a capacity scenario

Assign a status to each input

Use explicit categories such as hypothetical fixture, sourced observation, derived statistic and unsupported estimate. A hypothetical number can be useful when its purpose is clear. The problem arises when a displayed decimal or source link makes an invented value look like an observation.

For each field, retain the source owner, URL or licensed dataset identifier, relevant observation period, release or availability time, retrieval date, unit, currency, session and transformation. Record the person or process responsible for checking it. These fields let a second reviewer trace the input without reconstructing the analysis from conversation history.

An example with mixed evidence

Imagine a scenario containing $25 million AUM, a 6% weight, 50% trade fraction, $18 million ADV and a 20 bp impact coefficient. Suppose all five are invented for an educational example. Label all five hypothetical. The resulting $750,000 order and 4.167% one-day participation are derived from those assumptions, not observed facts.

Now imagine the ADV is replaced with a licensed 20-session calculation. The order remains hypothetical, the ADV becomes a sourced derived statistic and the coefficient remains an unsupported scenario parameter. That change improves one input’s provenance; it does not convert the total cost estimate into empirically validated execution research.

Preserve transformations and corrections

A volume average should identify which observations were included and how they were converted. If a currency conversion is applied, preserve the rate source and timestamp. If corporate-action adjustments are used, preserve the adjustment basis. A final CSV containing only ticker and ADV is not enough to reproduce these choices.

Keep an immutable version of each decision’s input set. When a vendor corrects a print or a calculation bug is fixed, create a new version with a change note. Do not overwrite the original and then imply the corrected data were available at the historical decision time. A current corrected analysis and a point-in-time evaluation can both be useful if distinguished.

Rights are separate from accessibility

NYSE’s market-data policy and connectivity pages describe documentation governing receipt, use and redistribution. A publicly viewable webpage is therefore not blanket permission to redistribute a market-data history. Check the terms that apply to the precise dataset, use and publication format.

For a public guide, a source citation and original hypothetical fixture may be sufficient to explain a method without republishing restricted observations. If a downloadable empirical dataset is necessary, record its licence or permission. A hash can help identify a retained file, but hashing does not grant rights or demonstrate that the file is accurate.

Use the ledger to constrain claims

An evidence table can map a claim such as 'this scenario reaches 8 bps at $31.25 million' to the formula and hypothetical inputs. A stronger claim such as 'this strategy can manage $31.25 million' would additionally need representative execution evidence, coverage and operational context. The first record cannot silently support the second.

Include failed retrievals and missing evidence. A source blocked by a website should remain unconfirmed rather than being counted as reviewed. A complete ledger is not one with no empty fields; it is one that makes material gaps visible and prevents the released wording from exceeding what the retained evidence actually supports.

Keep in the decision record

• Label assumptions, observations and derived values distinctly.
• Preserve timestamps, transformations and versions.
• Source access, redistribution permission and empirical validation are separate checks.

Source notes

1. NYSE — Market Data Policies
2. NYSE — Connectivity Documents
3. Federal Reserve — Supervisory Guidance on Model Risk Management, April 2026

Sources inform the definitions and limitations. Scenario inputs and arithmetic are publication illustrations, not extracted market observations. Access checks are dated in the source inventory.

Continue the analysis