Merge bias in Schedule Risks Analysis, or Portfolio Effect

This forum includes discussion about project risk analysis and risk management theory: Monte Carlo simulations, Event Chain Methodology, schedule and cost risk analysis. Please submit questions and case studies about your experience with our project risk analysis software.

Moderator: Intaver Support

Post Reply
Intaver Support
Posts: 989
Joined: Wed Nov 09, 2005 9:55 am

Merge bias in Schedule Risks Analysis, or Portfolio Effect

Post by Intaver Support » Sat Aug 29, 2015 9:44 am

This is a question that we often get on our support email.

After importing or creating a schedue into RiskyProject, they input the information that they have gathered from a risk workshop. The steps are to this are usually straight forward:

1. Enter in the Risk name in the Risk Register.
2. Add additional metadata about the risk to the Risk Properties form.
3. Open the Drag N' Drop Risks view.
4. Select the tasks to which you want to assign a risk
5. Drag the risk onto the tasks and the Assign risks to tasks dialog box opens. Here you can quickly add the chance (probability) and outcomes (impacts).

When you calculate the results, it appears that you have almost no chance of completing on the original budget or schedule. This is a very common. First, deterministic schedule rarely account for 'merge bias' or " portfolio effect'. These are two factors that that cannot be measured by critical path calculations and tend to expand the expected values for both schedule and budget (due to time dependent costs). Merge bias occurs when parallel (activities occurring at the same time) all share the same successor and the successor cannot start until all of the predecessors are complete. This is very common and if any of the activities is delayed, even if they are not on the original critical path, they can cause delays. In large projects, there tend to be many parallel activities and without Monte Carlo simulations we can not measure the possible implications.

Second, the portfolio effect occurs when you have many activities with distributions linked in a predecessor network. In this case, project tasks tend to have right-skewed distributions because activities have only a limited amount in which you can decrease length, but theoretically at least, the activities of durations have no limit. In these circumstances, the causes the average values (p50) of the task's duration distribution to be to the right of the Most Likely (the peak). The combination of the precedent network and risk skewed distributions means that mathematically (and in reality) that the Most Likely value for the Project shifts towards the average value and in practical terms can significantly extend what you can realistically expect your schedule to be.
Attachments
ParallelActivities.PNG
ParallelActivities.PNG (178.41 KiB) Viewed 26469 times

Post Reply