Tip of the Week #101 Tip Index

Go to the Prior Tip Extraordinary Popular Delusions and the Madness of Crowds
Go to the Next Tip Neural Network Control System for Steam Injection
Return to MaxValue Home Page

"Why Monte Carlo Simulations of Project Networks Can Mislead"

by Terry Williams, Project Management Journal, Sep 2004, pp. 53-61.

This is the best article that I've seen in years about project modeling. Terry Williams highlights two flaws commonly seen in Monte Carlo simulation models of project networks:

If a project is running late, the manager will take action to recover the schedule; and
Catastrophic overspends occur because of positive feedback loops.

Managers Affect the Distribution

In reading Critical Chain by Eliyahu Goldratt, I was most struck by the human behaviors that affect judgments about activity times. The 'student syndrome' characterizes the tendency of people to waste the slack. People procrastinate in starting activities with lots of slack (e.g., students put off working on assignments until timely completion is at risk). Williams works the other side: If a project (or activity) is running late, then the manager will take corrective action to bring it back toward schedule. Both behaviors are somewhat analogous to mean reversion behavior in statistics.

Remedies:

Anticipate the behavior in judging time to complete distributions.
Important decision points (where resources can be redirected) should be embedded in the project model.

The project model is our attempt at understanding the project as a system. Terry shows many the complexities introduced. For example, adding manpower to the project (or activity) results in:

Overcrowding, multi-tasking, and using untrained staff ---> lower productivity per person
Higher communication overhead, and trained workers diverted to train new workers ---> Lower proportion of time spent on productive work

In addition to expediting an activity, Terry discusses increased parallelism and temporary workarounds to bring a project back on schedule.

Positive Feedback Loops

Most project managers have experienced this situation: When some things start to go wrong, most everything starts to go wrong. There is often a cascading effect of delays and cost overruns. Terry offers a detailed example, modeled in Excel with Palisade Corporation's @RISK add-in. (Palisade also offers @RISK for Project). He discusses system dynamics, which is the science of modeling the behavior of complex systems. There are no easy solutions here, but the implications are important considerations in project modeling. I often include a 'System Complexity' variable to represent inadequate scope definition; higher complexity tends to impact most activities together.

There is an excellent bibliography. I highly recommend this article for those persons responsible for project modeling.

—John Schuyler, Sep 2004.