A tracking implementation rarely slips because one task was hard. It slips because the same analyst is booked for three tasks at once, the client approval waits in an inbox, and every estimate already hides extra days that nobody returns. Critical Chain Project Management (CCPM) is a planning method built for exactly these problems. This article applies it to one hypothetical analytics project, with every number checked.
Where Critical Chain comes from and what it changes
Eliyahu Goldratt introduced the approach in his 1997 book Critical Chain, as an application of his Theory of Constraints to projects. That summary comes from Geekie and Steyn's paper on buffer sizing (South African Journal of Industrial Engineering, 2008). This article relies on such secondary sources, not on the book itself.
Larry Leach's paper on CCPM (Advanced Projects Institute, 2000) names two differences from the classic critical path. The critical chain includes resource dependencies, and it does not change during the project. The method also changes how time is estimated:
- Each task gets a median estimate, about 50% likely to be met, instead of a safe one.
- The removed safety goes into shared buffers. A project buffer sits at the end of the chain. Feeding buffers sit where other paths join it.
- People work on one task at a time and hand it over as soon as it is done.
Leach explains why per-task safety gets lost. Work expands to fill the time available (Parkinson's law). People start late when they know there is slack (Goldratt calls this the student syndrome). Early finishes are rarely reported, so delays pass on and gains do not.
