In a Critical Chain project, management accepts that task times are not deterministic. This means that during planning, a person cannot state that the task will take exactly 3.2 days,
100 80 60 40 20 0 P R O B A B I L I T Y (%)
Duration (Days)
FIGURE 22–3. Observed duration of a task during execution.
for example. Task times are guesses. Therefore, it is perfectly normal for task times to take longer than estimated.
Management does not worry about whether or not a task finishes on time. They focus on finishing the project on time. In Critical Chain environments, organizations quote an- nual project on-time percentages typically above 95 percent. This section examines the way in which projects are protected to achieve these results.
W. Edwards Deming taught management the importance of having a system that stays in control. By this, Deming refers to the “predictability” of a system, relative to its goals.
In describing such a system, Deming refers to two types of “problems” that any system can encounter. He calls these problems “variations.”6Every system is subject to a type of prob- lem called “common cause” variation. These types of problems are absolutely normal, and management should do nothing about them. In fact, if a manager tries to do something about a common cause variation, it often causes the system to go into chaos.
For example, in any organization, it is perfectly natural and expected that some peo- ple occasionally will not show up to work on time or will report in sick. It is perfectly nat- ural for machines to break down from time to time. A person, designing the procedures for an organization, should expect these “common cause variations.” Deming insists that the correct procedure, when a common cause variation is encountered, is to do nothing.
In any system, “common cause” variation (in which managers should expect the vari- ations and take no action) and “special cause” variation (in which managers always take action) must be defined.
To arrive at a project plan that will stay in control throughout the entire execution, the following steps are necessary:
● Take all forms of padding out of the task estimates. This can be done through moving from “elapsed time estimates” to “dedicated time estimates.” Another way to remove padding is to educate the organization on the Critical Chain methodol- ogy, and give this task to the project managers. This step removes any chance of
“Student Syndrome.”
● Resource level the project. Do not schedule the project assuming that resource contention will magically take care of itself. In Critical Chain, resource contention is resolved up front.
● Don’t measure people on completing their tasks on time or on the accuracy of their estimates. If management wants to reward team members, the reward should be based on finishing the project on time or early.
● Allow people to work on a Critical Chain task in a “dedicated” manner. This is part of the second focusing step, “Exploit the system’s constraint.” Dedicated be- havior implies that once a person begins to work on a task, he will work only on that task until either it is complete, or he has progressed far enough on the task to turn it over to the next resource to work on it. Further, dedicated behavior implies that if the task is turned over to the next resource, but will be returned to the current
Protection in a Critical Chain Project 843
6. An excellent explanation is contained in Donald J. Wheeler,Understanding Variation, The Key to Managing Chaos,2nd ed. (Knoxville, TN: SPC Press, 2000).
resource within a very short period of time for more work, the current resource will wait for the return of the task, rather than begin work on a new task. This is true if the new tasks will take up substantially more time than waiting for the cur- rent task to be returned.
● Change the organization’s resource management approach to implement the par- adigm that resources on the Critical Chain have more flexibility to accept work earlier than expected. This is usually accomplished through an early warning system, called the resource buffer. The resource buffer does not take up any time on a schedule. It simply acts as an alarm clock, warning resources that they are X days away from receiving a new task.
● Implement a project buffer to protect the project’s Critical Chain. It sits at the end of a project and is calculated as a percentage of the length of the Critical Chain, typically 30–50 percent. This buffer is protecting the collection of tasks on the Critical Chain from any common cause variation.
● Implement feeding buffers on each feeding path, to protect the Critical Chain from variances on any feeding path.
We will use the same example as shown in Figure 22–1 to illustrate how this works.
Figure 22–4 shows what the project looks like, after the protection has been removed from each task. For example, the first task, “Modify Standard Parts,” was changed from the sixteen-day estimate shown in Figure 22–1 to a ten-day estimate. The task “Design Custom Parts” was changed from a forty-day estimate to a fifteen-day estimate.
The next step is to perform resource leveling. Since the same resources from the Engineering Drawings area and the Procurement area will be working on all of these tasks, we will not allow them to do two tasks at the same time.
Test Finished
Product Modify
Standard Parts E.
D.
10
Original Drawings E.
D.
30
Test Purchased
Parts I.
E.
10
I.
E.
20 Purchase
Raw Materials P
15
Purchase Available
Parts P
30
Fabricate Poles and Base F
10 Design
Custom Parts D.
E.
15
S 10
Manu- facture
FIGURE 22-4. Build a tower project with dedicated effort (no padding).
Figure 22–5 shows the project after resources are leveled.
The next step is to identify the Critical Chain—the longest chain of dependent events, where the dependency is either task or resource related. Looking at Figure 22–5, we can identify several possible paths. In considering which path is the most critical, Critical Chain looks at two possible paths for every task.
One path asks, “What is the next task that is logically dependent on this task?” For ex- ample, after the second task, “Purchase Raw Materials” in Figure 22–5, the next task that is logically dependent on purchasing raw materials is task 3, Fabricating the Poles and Base. You cannot fabricate without purchasing.
Another path asks, “What is the next task that uses the same resource as the current task where the dependency is therefore a resource dependency?” In the above example, af- ter “Purchase raw materials,” which uses the Procurement resource, the next task using Procurement is “Purchase available Parts.”
Based on this analysis, Critical Chain examines the duration of each of the possible paths and picks out the longest path based on task or resource dependency. In this exam- ple, Figure 22–6 identifies the Critical Chain, shown in shaded format.
To finish the Critical Chain schedule, the project and feeding buffers must be inserted.
Buffers are illustrated in Figure 22–7 with bold text. The buffers were calculated at 40 per- cent of the length of the paths.
Every task is now attached either to a feeding buffer or to the project buffer. In this ex- ample, there are three paths that feed into the Critical Chain. Therefore, there are three feeding buffers. The task “Design Custom Parts” feeds the Critical Chain directly and also feeds another task, “Original Drawings” on a feeding path.
With the protection now accumulated, the effect of individual task time deviations during execution are insulated by the buffers. In effect, the buffers act as shock absorbers.
Protection in a Critical Chain Project 845
Test Finished
Product Modify
Standard Parts E.
D.
10
Original Drawings E.
D.
30
Test Purchased
Parts I.
E.
10
I.
E.
20 Purchase
Raw Materials P
15
Purchase Available
Parts P
30
Fabricate Poles and Base F
10 Design
Custom Parts D.
E.
15
S 10
Manu- facture
FIGURE 22-5. Resource-leveled project.
Test Finished
Product Modify
Standard Parts E.
D.
10
Original Drawings E.
D.
30
Test Purchased
Parts I.
E.
10
I.
E.
2 Purchase
Raw Materials P
15
Purchase Available
Parts P
30
Fabricate Poles and Base F
10 Design
Custom Parts D.
E.
15
S 10
Manu- facture
FIGURE 22–6. The Critical Chain identified.
Test Finished
Product Modify
Standard Parts E.
D.
10
Original Drawings E.
D.
30
Purchase Available
Parts P
30
I.
E.
20 Purchase
Raw Materials P
15
Fabricate Poles and Base F
10
Design Custom
Parts D.
E.
15
S 10
Manu- facture
Test Purchased
Parts I.
E.
10
Feeding Buffer
Project Buffer Feeding
Buffer
Feeding Buffer
FIGURE 22–7. Project and feeding buffers.