As you may have noticed in my previous posting on this subject, @RISK for Project‘s Probabilistic Calendars are a great way of modelling multiple occurrences of the same risk event over a defined period and as a result avoid having to set up lots of risk tasks. Furthermore, you can use them for modelling any type of calendar-related risk and affect just the tasks taking place in the period concerned.
The main disadvantage is that they are less visible in the plan, compared to a separate RISK/IF statements or discrete task, and the impact is not so variable. In addition, they have no effect on cost (if you have costed your plan), as the assumption is that non-working resources will be deployed elsewhere, which is not always the case.
Perhaps my favourite Probabilistic Calendar example is modelling hurricane seasons. Here we have a long project (4 years) where there is a 20% chance that 1 month will be lost each year due to the annual hurricane season.
(click to view larger image)
As no uncertainty has been put in the tasks (in order to see only the effect of the hurricanes each October) the pre-risk finish date is 19th February 2010. However, after taking the hurricane season into account, the finish date has the following distribution:
Probabilistic Calendars are therefore very appropriate for this type of long timescale forecast.
Out of interest, the model is also very similar to tossing a biased coin 4 times, as in the binomial distribution below, where the chance of more than 2 years with hurricanes is only 5% but there is a 40% chance of at least 1 hurricane in the 4 year period.
Ian Wallace, ACMA
Palisade Training Team