Portfolio risk reporting for RIAs

Two readers, one document

The client is reading for reassurance and for surprise. They want to know what a bad year looks like in money, whether anything has changed since last time, and whether you noticed before they did.

A process review examines the stated objective, the method chosen for the report and the inputs behind each figure. Retain those details so a reviewer can trace the calculation.

A report can serve both readers through a short client-facing section and a methodology appendix. State losses in currency with their horizon and assumptions. Keep the conventions and coverage available in the same record so clients, advisers and reviewers can inspect the basis.

What clients actually understand

Explain modeled losses in account currency, with a stated horizon, confidence level and assumptions.

• Report loss in currency first, with the percentage in brackets.
• Give a frequency for every threshold. “One year in twenty” is a unit people own.
• Show what changed since last review, and lead with it. A stable report with an unexplained change is worse than a moving one that explains itself.
• Name the two or three exposures driving the risk. Connect each reported contributor to the client's holdings and explain the assumptions.

The methodology appendix, in full

Everything in this list is a choice that changes the number. Stating them is what turns a figure into a defensible figure.

1. Confidence level and horizon. A daily loss threshold and an annual loss estimate answer different questions. State the confidence level and horizon beside each figure, and explain any change between reports.
2. Estimation window and weighting. The history included and the weight placed on recent observations affect the estimate. Keep those settings in the report and explain changes before comparing periods.
3. Ex-ante or ex-post. Whether the number applies today's weights to a covariance model or describes the realised path of a book that has since changed.
4. Exposure convention. Whether weights are over NAV or over invested capital. The two differ by the cash weight, and a report that does not say which will be asked.
5. Coverage. The share of the portfolio by weight that could actually be priced, and what was excluded.
6. Data as-of and dataset version. The date the inputs were current, distinct from the date the report was produced.
7. Model changes since the last report. If the methodology moved, the comparison to last quarter is not like-for-like and must say so.

Cadence, consistency and the review that goes wrong

Changes in reporting methods can make quarterly risk figures difficult to compare. State whether a change comes from the portfolio, inputs or methodology, and restate comparisons where appropriate.

• Fix the method for a full year at a time. If it has to change mid-year, restate the prior period on the new method and show both.
• Regenerate rather than edit. A pack assembled by hand from last quarter's file inherits last quarter's mistakes and hides this quarter's changes.
• Keep the report and its inputs. A PDF alone cannot reproduce a figure. Retain the dataset version, as-of date, holdings and model settings used to calculate it.
• Separate the alert from the review. A quarterly cadence is fine for the review and far too slow for a breach. Limits should be monitored continuously and reported on a schedule.

Keep the evidence needed to explain a report: holdings, input dates, model settings, coverage and the released document. The SEC marketing and books-and-records references below provide context for US advisers. Compliance staff should decide which requirements apply to a firm and its communication.

How this runs in TradePolaris

Aegis computes risk from the entered holdings, supported instrument mappings and available modeled history. Its CSV and PDF exports include method and confidence information alongside concentration, factor, correlation, drawdown, liquidity, options and caveats. Review those sections and the coverage before using the report in a client discussion.

The risk-limit arithmetic is published as a page, so a client or reviewer can inspect its definitions. Review coverage beside the result. Decision Review withholds checks that require complete modeled coverage when history is missing. A standing monitor can evaluate its modeled subset and states what it measured.

For reproducibility, retain the dataset version alongside the holdings, method, parameters and report date. A data version alone does not freeze a model or recover missing inputs. Read point-in-time data and survivorship bias for the data boundary.