Moderator: Intaver Support
But what should one do if the activity is partially completed and certain events are assigned to the activity? If the event has already occurred, will it occur again? Or vice versa, if nothing has occurred yet, will it happen?
There are four distinct approaches to this problem:
1. Probabilities of a random event in partially completed activity stay the same regardless of the outcome of previous events. This is mostly related to external events, which cannot be affected by project stakeholders. It was originally determined that “bad weather” event during a course of one-year construction project can occur 10 times. After a half year, bad weather has occurred 8 times. For the remaining half year, the event could still occur 5 times. This approach is related to psychological effect called “gambler’s fallacy” or belief that a successful outcome is due after a run of bad luck.
2. Probabilities of events in a partially completed activity depend on the moment of the event. If the moment of risk is earlier than the moment when actual measurement is performed, this event will not affect the activity. For example, activity “software user interface development” takes 10 days. Event “change of requirements” can occur any time during a course of activity and can cause a delay (a uniform distribution of the moment of event). 50% of work is completed within 5 days. If the probabilistic moment of event happens to be between the start of the activity and 5 days, this event will be ignored (not cause any delay). In this case, the probability that the event will occur will be reduced and eventually become zero, when the activity approaches the completion.
3. Probabilities of event can be calculated based on original probability and historical data related to accuracy of previous assessment of the probability. In this case probability of event can be calculated using Bayesian Theorem:
P(E|H) = P(H|E)*P(E) / ( P(H|E)*P(E) + P(H|E’)*P(E’) )
P(E|H) – probability of event
P(E) – original probability of event (e.g. 30%).
P(E’) – probability of normal flow of the activity (event did not occur) (e.g. 70%)
P(H|E) – accuracy of event assessment based on historical data (i.e. the probability the event was properly identified ) (e.g. 90%)
P(H|E’) – accuracy of normal flow of activity assessment (e.g. 80%)
In this example, the probability of the event calculated taking into account the accuracy of the assessment of historical data equals 32.5%. Probability of the event has slightly increased because the previous assessment of probability was not 100% accurate. This approach to probability calculations is effective if there is an established process or tools to record the actual occurrence of events.
4. Probabilities of events need to be defined by the subjective judgment of project managers or other experts at any stage of an activity. For example, the event “change of requirements” has occurred. It may occur again depending on many factors, such as how well these requirements are defined and interpreted and the particular business situation. To implement this approach excited state activities should be explicitly subscribed or not subscribed to certain events. For example, a new excited state after the event “change of requirements” may not be subscribed to this event again, and as a result this event will not affect the activity a second time.