The first task is to "modify standard parts." It is estimated to take sixteen days elapsed time, and is being done by the resources in the Engineering Drawings department.

Every theory or methodology is based upon assumptions. Critical Chain is no exception. One of the assumptions in Critical Chain is that a task time, such as the first task described above, is not a deterministic number. It is an estimate. What this means is that any task that is part of a project cannot be predetermined to take an exact amount of time. By the very nature of projects, each task is unique and somewhat unpredictable in terms of how long it will take. Even if the same person were to do an almost identical task as before, the time to perform the task can vary widely. Is the person's mental state the same? Is he using the identical computer and software as before? Are all of the conditions identical (team members, bosses, external contractors, communications facilities, etc.)?

In this section, we will explore how the task time varies in execution, and how this impacts the estimate. In most companies, people take pride in reliable estimates. It is not good for anyone's career, especially an engineer's career, to repeatedly be way off on estimates. Considering that an engineer can never predict exactly what kinds of problems he or she may run into, and considering that this project is not the only work that the engineer is doing, the engineer provides an estimate that takes all of these considerations into account.

The engineer knows that this is not the only project the company is working on. Some work is far more urgent than other work, and the urgency itself is unpredictable. What happens is that a contractor or a community decides to build towers to offer new or improved communications services within the community. Many residents who live in the community consider the towers ugly and do not want them erected near where they live, so locations for a tower are limited. Further, the approval of a particular location requires preliminary surveys, including a geological survey to check soil conditions, and some political investigation to determine which locations are likely to be approved by the zoning commission.

Therefore, it is quite unpredictable when a location will be found and approved. Once a location is approved, the contractor often wants the project work done as soon as possible, before people change their minds.

This is all part of the engineer's uncertainty when providing an estimate. In this case, assume that the task could be realistically done in ten days if the engineer could dedicate himself to this one task. However, this rarely happens. If other project work is authorized, the engineer will either be working substantial overtime to meet the ten-day estimate or will be late.

Therefore, the engineer does not give an estimate of ten days. The engineer protects his reputation by committing to a date that gives him flexibility. In other words, there is a difference between the due-date he would commit to if dedicated to the one task (a "dedicated" estimate) versus the due date he actually commits to ("elapsed time" estimate). Another way to protect his reputation is to add safety or padding time directly into the task estimate. This is more common where the amount of individual effort on tasks is closely watched or tracked.

So far, we have discussed only the first task, which has relatively little variability. The engineers are taking existing drawings and doing minor modifications. The amount of protection on this type of task is likely to be small.

Now let us consider the last task in the project, which we have called "testing the finished product." This task usually has a very high degree of uncertainty. This is where all of the previous tasks must come together. If any one of the previous tasks is not finished on time, this task cannot start on time. If any one of the previous tasks was done without perfect quality, this task will discover the problem. Fixing the problem might require sending parts back to an outside vendor and waiting for rework, or trying to free up resources inside the organization to do rework. Under most circumstances, resources in organizations are very heavily scheduled, and it is not easy to free them up.

There are two types of dependencies that can cause a task to be late. One type is called a "logical task dependency." For example, you cannot purchase raw materials (task 2 in Figure 22-1) until the engineering drawings are ready (task 1). Another type of dependency is called "resource dependency." For example, assume that there is only one person who is capable of doing the engineering drawings. The same engineering drawing resource is used in two places—once to modify existing drawings and once to complete new drawings. The chance of having new drawings available for the subcontractors to manufacture (task 6 in Figure 22-1) depends not only on those new drawings being completed on time, but also on the previous task for the engineering drawing resource being completed on time so that the same engineer(s) can work on the next task.

Resource dependencies, compounded with task dependencies, further decrease the probability that a task will finish on time. At the final stages of a project (usually involving some type of integration or testing), there are usually a lot of task and resource dependencies. This implies a high degree of uncertainty in these tasks. A large amount of protection is usually put into these types of tasks.

In order to understand how much protection is included in an estimate, consider the task duration profile shown in Figure 22-2. The exact nature of this profile for any task in any organization can never be known. However, the general shape of the curve is commonly observed in any organization.

This curve shows, on the vertical axis, the probability of a task finishing within the duration shown on the bottom of the graph. There are three points that we can identify immediately on any such curve. For any task that is expected to take ten days, for example, we know that there is a zero percent probability that this task will finish in zero days. The minimum time duration is some number greater than zero. This starting point would appear at the left side of the curve. Chances are very slim that this task will finish in the shortest duration.

For this task, there is another number on the curve, which is the longest amount of time we might expect a ten-day task to take. This number might be 2.5 or 3 times the normal expectation. In other words, in rare cases, a ten-day task might take us twenty-five or thirty days. The probability of such an occurrence is also rare, and appears on the right side of the curve.

There is another number, which is the most likely time. This number is somewhere between the lowest duration and the highest duration. The peak in the curve represents this most likely duration.

The curve shown in Figure 22-2 is skewed to the right. If correct, it suggests that project tasks are likely to take much longer than expected, and less likely to take a shorter time. Critical Chain claims this is true for several reasons, explained in detail in Dr. Goldratt's book, Critical Chain.5

One reason that tasks typically do not finish in shorter durations than planned is the effect of dependencies. Some of the work that is needed to finish the task does not show up on time, either because a resource was tied up on a previous task or because a previous task did not finish on time.

Another reason that tasks are often skewed to the late side is multitasking, which is discussed more extensively below. Briefly, multitasking exists when an organization has a project resource that is assigned to do multiple tasks at the same time. Organizations often have more projects than they have resources to dedicate to these projects. As a result, people often start their tasks later than planned. The multiple tasks may be on the same project or on different projects.

People sometimes start tasks late due to "Student Syndrome." This behavior is similar to what happens in a classroom when a professor announces that there will be a test tomorrow on all of the material covered during the semester. The students complain bitterly that this is not fair, since they did not have enough time to study for the test. A kind professor might change the test date, moving it back two weeks. Do the students rush out of the class and begin studying immediately? Not typically. Their attitude usually is, "We have plenty of time. Why rush?" They begin studying the night before the test.

When people give elapsed time estimates or estimates with padding, they realize that there is no urgency to start working on this task. Since they are multitasked, they pick the most urgent tasks to work on, and delay the start of work on other tasks.

This behavior relates directly to the skewed curve. Often, the major difficulties in completing a task are not discovered until the person is very involved in the task. This usu-

5. Eliyahu M. Goldratt, Critical Chain (Great Barrington, MA: North River Press, 1997).

Duration (Days) FIGURE 22-2. Profile of the duration of a task.

ally happens in the latter part of working on the task. At this stage, it may be too late to finish the task within the time estimate. Student Syndrome wastes the protection that was embedded in the original task time estimate. As a result, the actual time to complete the task moves toward the right side of the skewed curve.

The reason that the concept of the skewed curve is important, relative to task time estimates, is that it impacts the estimate that a person will give. If somewhere toward the middle of the curve, the task has a 50/50 chance of being late, this is not the likely estimate that a person will provide. An engineer who wants to be considered reliable will not take a 50/50 chance of being late. To be considered reliable (i.e., 80 percent chance or better), he will estimate toward the right side of the curve. On a skewed curve, the 80 percent chance of meeting the estimate is often 2-2.5 times the 50/50 duration, according to the statistical distribution of a skewed curve.

Therefore, on tasks that are considered by team members to be more risky or more difficult to predict, a ten-day task may well be estimated at twenty to twenty-five days. However, even with such an estimate, the protection is often wasted due to Student Syndrome and multitasking.

These task estimates are not created in a vacuum. Often, there is extensive discussion about the estimates, with managers pushing back on what they consider to be unreasonably protected estimates. On the other hand, team members expecting management to push back often inflate their estimates to take this into account. The end result is that estimates often are distorted based on the individual experiences and negotiating skills of management and team members. Further, two estimates for the same task, done by different people, will likely vary widely.

Project Management Made Easy

Project Management Made Easy

What you need to know about… Project Management Made Easy! Project management consists of more than just a large building project and can encompass small projects as well. No matter what the size of your project, you need to have some sort of project management. How you manage your project has everything to do with its outcome.

Get My Free Ebook

Post a comment