POOLED RISK SCHEDULING

• Task Variance Overview

A project task’s duration often has a degree of uncertainty. A single task is shown in the figure below. The task duration has a range from a minimum value to a maximum value as shown in the gray uncertainty bar. The distribution of the duration is often skewed from the mode to longer values as illustrated by the blue distribution curve. In project management software, the duration of a task is entered as a single value. The task duration is then saved as a baseline. Increases to the duration are considered overruns as noted by the red area of the task bar even though the “overruns” are simply common cause variation that should be expected on some tasks in a project. Pooled Risk Scheduling provides a method for integrating this task variance into a project schedule. Importantly, Pooled Risk Scheduling supports the CCPM concept of aggregating schedule margin (safety) to account for variance in a buffer instead of in each task.

• Methodology

A PRS project is planned using standard critical path methods by constructing a project network of tasks with resource assignments. A special task called a buffer task is placed at the end of the project network. Additional buffer tasks are placed in front of key milestones that reflect external commitments for a project. A buffer task has special characteristics in that it provides schedule margin between a commitment date and the predecessor tasks driving the commitment date. Extensions to predecessor tasks absorb the buffer margin without pushing the commitment date as long as there is margin left in the buffer. The CCPM feeding buffers, which add to complexity in a CCPM project, are avoided by basing PRS on standard critical path methods and capitalizing on traditional float as a safety factor.

Monte Carlo simulation is used to size the buffers. Task variance is simulated through the use of statistical distributions to define the potential task duration distribution. Resource leveling can be performed in each simulation iteration to incorporate resource constraints into the determination of the critical path and the buffer sizes. The PRS simulation results in a planned project with buffers sized as a function of the task variance, resource constraints, and specified confidence level. The following figure shows a simple example of a planned five task project with a PRS buffer (yellow) protecting a commitment date milestone.

• Tracking & Monitoring

In a PRS tracking update cycle, tasks are updated to reflect completion status and remaining work. If increases in task durations are made that affect the critical path, the buffer is absorbed but the committed project date is maintained as illustrated below where an increase is made in task 3’s duration as shown in red.


Project monitoring is accomplished by comparing the original buffer size with a re-estimated buffer size after each update cycle. This re-estimation is done by running a Monte Carlo simulation that leaves the original buffer’s schedule position intact but calculates an Expected Buffer Finish date for each buffer. This Expected Buffer Finish date is logged with a date stamp. Then a re-estimated buffer size is calculated and plotted as a ratio of the baseline buffer size to provide time-phased insight into project performance as illustrated below.


The above illustration shows a 10 task project with a baseline buffer and a commitment date and the Expected Buffer Finish date as a red inverted triangle. Four update snapshots are provided to reflect task & buffer status at periods: 4, 6, 10, and 14. The buffer is consumed faster than expected starting in period 4 but is brought close to expectations by period 10. The project is finished early at period 14.

The PRS process of running a simulation on the current state of an updated project while leaving the original buffers in place means that all of the changes made to task estimates, task durations, resource assignments, task additions, and task dependencies are taken into account in evaluating buffer consumption.

References

Putting quality in project risk management, part 1: understanding variation / Lawrence P. Leach
PM network. Vol. 15, no. 2 (Feb. 2001), p. 53-56., PMI

A CRITICAL LOOK AT CRITICAL CHAIN PROJECT MANAGEMENT Article from: Project Management Journal Article date: December 1, 2003 pg 11 for buffer complexity support, Author:    Raz, Tzvi ; Barnes, Robert ; Dvir, Dov

On the merits and pitfalls of critical chain scheduling, Willy Herroelen*, Roel Leus, Journal of Operations Management, Journal of Operations Management 19 (2001)