Projects

Critical Chain: more projects, finished sooner

Mohit Saini · 17 Jan 2018 · 5 min read

Critical Chain: more projects, finished sooner

Project management has developed enormously over the last fifty years: standards, certifications, software, templates. Yet across software, engineering, construction and product development, project performance hasn’t improved much. A large project finished on time, on budget and in full scope is still rare enough to make the news.

Ask project managers what it takes to deliver reliably, and the answers sound familiar:

  • “A detailed, accurate plan and estimates.”
  • “The best resources, available when needed.”
  • “Vendors who deliver perfectly.”
  • “Requirements frozen up front.”

These would work in a world without uncertainty. Projects, by definition, are full of it. Every project is somewhat new. By the time a company has perfected its processes for one kind of project, the next one is different.

The solution that keeps failing

The usual cycle goes like this:

  1. A detailed plan is made, often extremely detailed, with estimates for every task and resources assigned to meet every deadline.
  2. Reality intervenes: requirements change, vendors slip, approvals are delayed, work takes longer, people are pulled to other projects.
  3. Deadlines are missed. People are moved around to fight fires. Project and resource managers argue about what is really urgent.
  4. The project is late. The post-mortem concludes that the plan wasn’t detailed enough.
  5. Next time, people protect themselves with more generous estimates. Sales then complains the estimates are unreasonable.

The surprising root cause

The intuitive explanation for late projects is not enough safety time. Dr. Eliyahu M. Goldratt pointed to the opposite: projects contain more safety than they need, and it is wasted.

When estimates become commitments, people sensibly add safety to every task. Then three well-known behaviours consume it:

  • Student syndrome: knowing there is slack, people start late, attending to other urgent work first, and the safety is used up before the work even begins.
  • Parkinson’s law: work expands to fill the time available. A task finished early is rarely reported early (it might lead to tighter estimates next time), so the extra time goes into polishing.
  • Bad multitasking: people juggle tasks across several projects, so every task takes longer.

The result: gains are never passed on, but delays always are. The teams at the end of the project (installation, testing, commissioning) inherit all the accumulated delay and work under the most pressure.

Other practices add to the problem:

  • Progress measures like percentage complete or earned value can reward people for finishing many easy tasks instead of moving the longest chain of dependent work.
  • Contract payment milestones that don’t match the flow of work push contractors to chase billing points instead of completion.
  • Starting projects as early as possible floods resources with work and makes multitasking inevitable.

Critical Chain Project Management

Goldratt’s answer, Critical Chain Project Management (CCPM), manages uncertainty rather than trying to eliminate it:

Planning

  • Task estimates are cut to focused, realistic durations, with roughly even odds of finishing on time, so student syndrome and Parkinson’s law have less room.
  • The safety removed from tasks is pooled: a project buffer at the end of the longest chain (the critical chain, which accounts for both task dependencies and resource availability), and feeding buffers where other paths join it. Pooled safety needs to be much smaller than the sum of individual safeties, so the planned lead time shrinks.

Releasing projects

  • In a multi-project environment, new projects are released according to the capacity of the most loaded resource, not as soon as they are approved. This limits work in progress and multitasking.

Execution

  • Since task estimates are deliberately tight, task deadlines are not used to judge people.
  • Projects are tracked by the rate of buffer consumption compared with progress on the critical chain.
  • Daily priorities, within and across projects, are set by buffer status.

What pooling does: a simple example

Picture a small project with five tasks in a chain. Each task owner, asked for a safe estimate, says 10 days. Of those 10 days, perhaps 5 are the work itself and 5 are protection against things going wrong. The plan says 50 days.

Now watch what happens to that protection. If a task goes well, it rarely finishes on day 5: the start slips or the work expands to fill the time. If a task goes badly, it overruns its 10 days, and the delay passes straight to the next task. Each task’s safety protects only that task, and only against its own bad luck.

Critical Chain plans the same project differently:

  • Each task is planned at its focused duration: 5 days. The chain is now 25 days.
  • The safety is pooled into one project buffer at the end. It doesn’t need to be the full 25 days that were taken out, because not every task will hit trouble, and an early finish on one task can now absorb a late finish on another. Half of what was removed, about 12 days, is a common rule of thumb.
  • The plan: 25 days of work plus 12 days of buffer, about 37 days instead of 50.

The promise to the customer is made at day 37. And because nobody is judged against the 5-day durations, a task that takes 8 days is not a failure. It simply uses 3 days of the shared buffer.

Reading the buffer

During execution, the question is no longer “Is each task on time?” It is: “Is the project eating its buffer faster than it is completing the chain?”

Suppose 40% of the critical chain is complete:

  • 20% of the buffer used: the project is healthy (green). Leave it alone.
  • 50% used: it is consuming protection faster than it progresses (yellow). Prepare what you would do if it gets worse.
  • 80% used: it is in danger (red). Act now: put the best people on the critical chain task, clear the pending approval, expedite the material.

Across many projects, the same comparison sets priorities. A shared engineer with tasks from three projects works first on the one whose project is deepest in red. There is no need for a meeting to debate what is urgent. The buffers say it.

Organisations that adopt CCPM typically report substantially shorter lead times and a significant increase in the number of projects completed with the same people.

Projects don’t fail for lack of safety time. They fail because safety time is scattered across every task, where it is quietly wasted.

Mohit Saini

Strategy, supply chain & sales transformation consultant. About · Book a discussion