A board does not act on a 97.5th percentile. It acts on: what can we lose, how likely, how fast, and what is our move. The model knows all four - the discipline is translation.
The gap is organisational, not mathematical
Most firms do not lack quantitative capability; they lack a translation layer. The quant answers in percentiles, liquidity horizons and confidence levels; the board asks in decisions - commit the cargo, extend the facility, hedge the book. When the two never meet, the meeting falls back on narrative and the model becomes an expensive compliance artefact. The failure shows up as a standing item: a "risk report" delivered, filed, unacted upon.
Four translation disciplines that work
- Answer the decision, not the model. Every board risk artefact should be structured around a choice: exposure against limit, capital against appetite, liquidity against the forward calendar. The number exists to move a decision; if no decision is pending, it belongs in an appendix.
- One page, one feed. Margin, cash, exposure and liquidity on a single screen sourced from the governed layer - the Control Tower pattern. Boards trust what they can trace; reconciliation disputes in the boardroom are data-layer failures surfacing at the worst possible moment.
- Quantify the tail in money. "97.5% ES of $42m over 10 days" and "a repeat of last year's Q4 dislocation costs us $60m and breaches two limits for a week" are the same statement. Only one of them gets decided on.
- Pre-agree the tripwires. Escalation thresholds defined while calm - which metrics trigger what action at which level - turn model output into governance. Without them, every breach becomes a negotiation.
Why the partners deliver this themselves
Translation is not a reporting task; it is a design constraint on the whole platform. It is why we stay engaged at C-suite level through implementation: the models, the feed and the boardroom artefacts are one system, and the system works only when all three were designed to answer the same questions.