Build prototype

Once we have enough data we can build a prototype. The causal model approach
models the cause/effect data and analyzes the relationship between them. If the
relationship has been determined in a delphic workshop it is input directly. The
incorporation of balanced scorecarding may also be necessary if the firm are keen
on this approach. This is where the competitor or industry benchmark data will be
found useful.
An example of balanced scorecarding could be in a dealing room where the firm
could set standards for each unit:
Ω ‘No more than 5 settlement fails per 1000 trades’
Ω If exceed score, allocate extra capital and charge them for it
A charge for allocated capital, that affects the profit and loss account, certainly
focuses the mind.
The use of risk visualization can also be incorporated into the prototype using
exception reporting of a ‘red, amber, green’ nature where the exceptions are against
quality control standards. A frequency distribution of high-frequency low-impact
events should be modeled and a distribution (a non-normal one) fitted. (This is a
non-trivial exercise as causes may be a Poisson distribution with losses a Weibull
one. After setting an appropriate confidence interval (say, 95%) the firm can measure
the unexpected loss and then be able to allocate risk capital for operational risk.