Hal Macomber provides an introduction to the problem of the Critical Path method not meeting the needs of project management.
This is the same old song. Project management teams that are too lazy to keep up their schedules. Ill prepared to actually "manage" the project in a proactive way. While this is all too common, it is not the problem of the Critical Path Method per se. There are issues with the standard A+2B+C approach to calculating the most likely durations for the task. (Which itself is not representative of the actual underlying distributions) But these issues are well known and solutions well described in the literature. Some references are provided below.
In the government contracting domain as well as large construction it is simply not allowed to not know the probabilistic critical path on a week by week basis.
Hal makes a previous statement
There is no critical path. Of course I'm not saying that one can't calculate a critical path. Of course you can calculate it. I'm saying that it is not a thing, just a characterization.
Yes it is a characterization - a critically important one, used to provide insight into the performance of the project, it's programmatic risks and their mitigations. What's missing here is the recognition that a probabilistic critical path Always exists, is Always changing and Always assessable with he right tools and a well formed task network.
By the way - if you don't have a well formed network of tasks, with no widows and no orphans, a single start and a single end, logical relationships between the tasks, then you're pretty much screwed to begin with. If I can't see all the work that is required to complete the project - or have some way to assess this work in a planning package - and see how this work is related, its probabilistic duration and the required resources - then there is really no plan for the project. This plan can be represented in many ways - but there needs to be one.
Remember there is a difference between a plan and a schedule
The aerospace approach to this problem is to use a Monte Carlo simulator to reveal where the schedule has problems. Then require that actual progress is captured in the schedule from some form of measurement - Earned Value is common in many domains, but other approaches are useful too. The use of the probabilistic critical path calculation on a weekly basis provides insight into the performance of the plan, so that the project management team can:
- Update the plan to reflect the current reality
- Assess the risk for the out date activities to further reduce the risk
- Provide a view of the risk buy down activities
- Support the assessment of the increasing maturity and increasing technical performance measures of the plan
These are all commonly used approaches in any aerospace or defense contractor. To not use them, or to claim they are not effective is to fail to understand how the probabilistic CPM processes and their support risk mitigation activities are applied in practice.
Revisiting a Dead Horse
David Schmaltz refers to a paper about the issues with the CPM method. The referred paper is really about the PERT estimating parameters. These problems have been well documented in the literature for decades - which is likely why the paper did not get accepted at Cross Talk. See "PERT Completion Times Revisited," Ted Williams, University of Michigan-Flint as a starting point for a survey of the source as well as solutions to this problem.
Monte Carlo simulations with risk adjusted task completion estimates derived in a variety of ways is the standard approach on NASA and Defense systems Integrated Master Plan / Integrated Master Schedule programs prescribed by DID 81650.
My personal experience includes Crew Exploration Vehicle ($3.8B), Hubble Robotic Servicing Mission ($670M) as well as working directly with those having managed 100's of Billions of dollars of other spacecraft and launch vehicle projects - all on time, budget, on specification. The I-25 TREX project just completed under budget and 8 months ahead of schedule here in Denver using probabilistic critical path management techniques.
It's a Poor Workman That Blames His Tools
It is not the tool that is failing the project. It is the failure of the people managing the project to adequately understand and mitigation the inherent uncertainties that reside in the project. Start with "Quantitative Risk Analysis for Project Management: A Critical Review," Lionel Galway, RAND Working Paper, WR-112-RC, Feb 2004.
For the treatise on probabilistic risk management for projects, see Effective Risk Management: Some Key to Success, 2nd Edition, Edmund H. Conrow, AIAA Press, 2003. The materials in this book are mandatory for any aerospace or defense system program manager.
A final note on the "wrongness" of the CPM calculation. This is well known and well understood in many domains. The solution is provided in several tools - Risk+, @Risk for Project, SAS, and Crystal Ball. Restating the problem without a known solution adds little to the profession of Project Management.
Anyone wishing to learn more about the probabilistic management of complex projects, drop me a note for the introduction briefing we provide to all entry level planners in the spacecraft development business.
