Page 46 - Pipeline Risk Management Manual Ideas, Techniques, and Resources

Risk assessment models 2/25
miles. This finding now tells us that condition G has increased purchased, time of day, and other factors to rate the proba-
the risk by a small amount. Suppose we change roadway D to bility of a fraudulent card use. Scoring systems are also
roadway H and find that our accident frequency is now once used for psychological profiles, job applicant screening,
every 300,000 miles driven. This tells us that by using road H career counseling, medical diagnostics, and a host of other
we have reduced the risk quite substantially compared with applications.
using road D. Chances are, however, we could have made these
general statements without the complicated exercise of calcu- Choosing a risk assessment approach
lating statistics for each variable and combining them for an
overall accident frequency. Any or all ofthe above-described techniques might have a place
So why use numbers at all? Suppose we now make both vari- in risk assessment/management. Understanding the strengths
able changes simultaneously. The risk reduction obtained by and weaknesses of the different risk assessment methodologies
road H is somewhat offset by the increasedrisk associated with gives the decision-maker the basis for choosing one. A case can
road and weather condition F, but what is the result when we be made for using each in certain situations. For example, a
combine a small risk increase with a substantial risk reduction? simple matrix approach helps to organize thinking and is a first
Because all of the variables are subject to change, we need step towards formal risk assessment. If the need is to evaluate
some method to see the overall picture. This requires numbers, specific events at any point in time, a narrowly focused proba-
but the numbers can be relative-showing only that variable H bilistic risk analysis might be the tool of choice. Ifthe need is to
has a greater effect on the risk picture than does variable G. weigh immediate risk trade-offs or perform inexpensive overall
Absolute numbers, such as the accident frequency numbers assessments, indexing models might be the best choice. These
used earlier, are not only difficult to obtain, they also give a options are summarized in Table 2.1.
false sense of precision to the analysis. If we can only be sure of
the fact that change X reduces the risk and it reduces it more Uncertainty
than change Y does, it may be of little Wher value to say that a
once in 200,000 frequency has been reduced to a once in It is important that a risk assessment identify the role of uncer-
2 10,000 frequency by change X and only a once in 205,000 fie- tainty in its use of assumptions and also identify how the state
quency by change Y. We are ultimately most interested in the of “no information” is assessed. The philosophy behind uncer-
relative risk picture of changeXversus change Y. tainty and risk is discussed in Chapter 1. The recommendation
This reasoning forms the basis of the scoring risk assess- from Chapter 1 is that a risk model generally assumes that
ment. The experts come to a consensus as to how a change in a things are “bad until data show otherwise. So, an underlying
variable impacts the risk picture, relative to other variables in theme in the assessment is that “uncertainty increases risk.”
the risk picture. If frequency data are available, they are cer- This is a conservative approach requiring that, in the absence of
tainly used, but they are used outside the risk analysis system. meaningful data or the opportunity to assimilate all available
The data are used to help the experts reach a consensus on the data, risk should be overestimated rather than underestimated.
importance of the variable and its effects (or weighting) on the So, lower ratings are assigned, reflecting the assumption of rea-
risk picture. The consensus is then used in the risk analysis. sonably poor conditions, in order to accommodate the uncer-
As previously noted, scoring systems are common in many tainty. This results in a more conservative overall risk
applications. In fact, whenever information is incomplete assessment. As a general philosophy, this approach to uncer-
and many aspects or variables must be simultaneously consid- tainty has the added long-term benefit of encouraging data col-
ered, a scoring system tends to emerge. Examples include lection via inspections and testing. Uncertainty also plays a role
sporting events that have some difficult-to-measure aspects in scoring aspects of operations and maintenance.
like artistic expression or complexity, form, or aggressive- Information should be considered to have a life span because
ness. These include gymnastics, figure skating, boxing, and users must realize that conditions are always changing and
karate and other martial arts. Beauty contests are another appli- recent information is more useful than older information.
cation. More examples are found in the financial world. Many Eventually, certain information has little value at all in the risk
economic models use scoring systems to assess current condi- analysis. This applies to inspections, surveys, and so on.
tions and forecast future conditions and market movements. The scenarios shown inTable 2.2 illustrate the relative value
Credit card fraud assessment is another example where some of several knowledge states for purposes of evaluating risk
purchases trigger a model that combines variables such as where uncertainty is involved. Some assumptions and “reason-
purchase location, the card owner’s purchase history, items ableness” are employed in setting risk scores in the absence of

Table 2.1 Choosing a risk assessment technique

When the need is to. . . A technique to use might be
Study specific events. perform post-incident investigations, compare risks of specific failures. Event trees. fault trees, FMEA. PRA, HAZOP
calculate specific event probabilities
Obtain an inexpensive overall risk model, create a resource allocation model, model Indexing model
the interaction of many potential failure mechanisms, study or create an operating discipline
Better quantify a belief, create a simple decision support tool, combine several Matrix
beliefs into a single solution, document choices in resource allocation