risk metrics.

Where risk metrics fits on a long engagement.

Value at risk answers a narrow question — a loss threshold at a confidence level — and says nothing about what happens beyond it. Two portfolios with identical VaR can have entirely different behaviour in the tail, which is why expected shortfall, averaging losses beyond the threshold, has become the preferred measure where the tail is what matters.

Every metric rests on a distributional assumption. Normal-distribution VaR understates tail risk substantially because real returns have fatter tails, and historical simulation only contains the crises present in its window. A number quoted without its method is not comparable to anything.

What an assigned team does with risk metrics.

Backtesting is what separates a risk system from a risk report. Counting how often actual losses exceeded the stated threshold, and comparing that to what the confidence level predicted, is the only evidence the model works.

Running that continuously rather than at model approval is ongoing work, held within an agreed committed monthly capacity.

What we use risk metrics for.

• Tail behaviour measured Expected shortfall alongside VaR, because the threshold alone hides the tail.
• Method quoted with the number Assumptions stated, so figures are comparable across time and desks.
• Backtesting as routine Exceedance counts tracked continuously, which is the only evidence the model holds.

How risk metrics capacity is assigned.

Risk analytics capacity is assigned under managed data services, with backtesting treated as continuous rather than an approval step.