What to read first: Worst case analysis: When, why, and how
The reasonable worst case necessarily has consequence more severe than some other cases. That does not mean that the reasonable worst case represents the least acceptable risk.
The likelihood of the reasonable worst case will probably be low. It may be so low that the risk from that case alone is well within an acceptable range.
Less severe but more likely cases may not be so acceptable.
In the product launch example, there is a milder risk case than the reasonable worst case. The milder case is Actual average sales 110% to 130% of break-even volume. The likelihood of this milder case is assessed indicatively at 35%. Its consequence is assessed as ‘qualified success on profitability’, defined on the profitability consequence scale as Enterprise is just breaking even, stakeholders looking for better investment opportunities, or continuing for reasons other than profitability.
The risk from this case (alone) can be summarised as 35% likelihood of a qualified success on profitability.
The enterprise might well set a maximum acceptable likelihood of 30% for qualified success on profitability, based on risk capacity and appetite. Such a limit might result from enterprise thinking about the attractiveness of alternative investments of its resources, coupled with a fairly high appetite for uncertainty: these are hypothetical reasons.
It would follow that the ‘fair’ case, milder than the reasonable worst case, would represent unacceptable risk, on its own. This could be true even if the risk from the reasonable worst case is acceptable, as we said it was in the main post.
For some types of risk, you might want to first separate ‘cases’ by the size of the event or deviation leading to an unplanned outcome. You can then identify varying potential consequences for events of each size.
Going back to the data centre power failure example, you might first consider facility outages of different duration. Each outage duration has a likelihood. Then you could identify different potential consequences from an outage of each duration. You split the likelihood of each outage duration across the consequence cases that might follow from it.
The product launch example showed varying sales volumes but assumed that each sales volume range had a known implication for profitability. The relation to profitability might also be uncertain.