Critical Path Method developed by James E. Kelley Jr. and Morgan R. Walker (Critical-Path Planning and Scheduling, 1959); Critical Chain by Eliyahu M. Goldratt (Critical Chain, 1997).
Finds the longest run of dependent tasks, which sets a project's earliest possible finish, and in the Critical Chain version swaps padded task estimates for shared buffers.
Seven tasks in an office fit-out, and the one chain that sets the move-in date
Three chains of dependent tasks run from the start to move-in, and the longest one, the critical path, sets the earliest finish. Stretch a task on the wiring chain and the finish moves at once; stretch the furniture delivery and nothing moves until its two weeks of slack run out. Illustrative.
An example
A law firm had twelve weeks before the lease on its old floor ran out, and the fit-out of the new one had seven tasks. The office manager's first worry was the furniture: eight weeks from order to delivery, the longest single item on the list. She was ready to pay for rush delivery.
The project lead laid the tasks out by dependency. Strip-out came first, then wiring and plumbing side by side, then walls and ceilings once both were done, then flooring, then the move. Furniture ran on its own from day one. The wiring chain added up to eleven weeks, the plumbing chain to ten, the furniture chain to nine. The finish date was set by the wiring, not the furniture, and the furniture had two weeks of slack.
The rush fee would have bought nothing. The firm spent the money on a second electrical crew instead, which took a week off the wiring chain and gave it two weeks of margin before the lease ended.
Why it works
On a project with dependent tasks, the finish date is set by one chain, the longest run of tasks that each wait on the one before. Everything else has slack: it can slip by some amount without moving the end. People manage by the biggest or most worrying task, and that task is often not on the critical path, so effort and money go where they change nothing.
Tracing the chains shows which tasks deserve attention and how much any other task can slip. The critical chain version adds two observations. Padded estimates tend to get used up, since work expands to fill the time, so it strips the padding from each task and pools it in shared buffers at the end of the chain. And people shared across tasks create dependencies the task list does not show, so it treats a person's availability as a link in the chain.
The thinking behind it
James E. Kelley Jr. James E. Kelley Jr., a researcher at Remington Rand, developed the critical path method in the late 1950s with Morgan R. Walker, an engineer at DuPont, for planning and scheduling large projects. Their paper Critical-Path Planning and Scheduling (1959), in the Proceedings of the Eastern Joint Computer Conference, set out the core move: model a project as a network of activities and find the longest chain of dependent tasks, which sets the shortest possible duration.
Eliyahu M. Goldratt Eliyahu M. Goldratt, an Israeli physicist turned management thinker, created the theory of constraints, which holds that a system's output is limited by its constraint. He applied it to project scheduling in Critical Chain (1997), a business novel, treating the chain of dependent tasks and shared resources as the project's constraint and managing it with buffers in place of padded estimates.
How to run it
- List the tasks and what each waits on. For every task, name the tasks that must finish before it can start.
- Estimate each duration. Use a realistic figure; for the critical chain version, use a median estimate with no padding.
- Trace every chain. Add up each run of dependent tasks from start to finish.
- Mark the longest. That chain is the critical path, and its length is the earliest finish.
- Note the slack. For every other chain, the gap to the longest is how far it can slip unseen.
- Add the shared people. Where one person or crew works two tasks, treat that as a dependency, and retrace.
- Pool the padding in buffers. Put the safety margin at the end of the chain, and watch how fast it is used against progress.
Where it goes wrong
The path is traced once and forgotten. Durations change as work starts, and a chain with one week of slack becomes critical the day a task on it slips a week. Recompute every time an estimate moves.
Shared people are left off. Two tasks that look parallel on paper run in sequence if the same electrician does both. When a plan assumes a crew can be in two places, the real critical path is longer than the one drawn.
Buffers become new padding. A team that learns the project buffer exists will spend it early, and the protection is gone when it is needed. Report buffer use against progress, so a buffer consumed faster than work is completed is seen at once.
When to use something else
- Theory of Constraints. When the work is a repeating flow of orders or units, not a one-time project, and the bottleneck is a station.
- Kanban Method. When the work arrives continuously in small items with no fixed end date to plan back from.
- Monte Carlo Simulation. When the durations are uncertain enough that the question is the chance of finishing by a date.
Where it sits in SOLVE eX
Go to the source
- Critical-Path Planning and Scheduling, Eastern Joint Computer Conference, 1959
- Critical Chain, North River Press, 1997