Several buffer length calculation methods for key chain methods

In the key chain schedule method, after WBS decomposition, task period estimation, and task chain and key chain identification are completed, feeding buffer and project buffer must be added ). The connection buffer can avoid delays in the entire project because tasks on non-critical links cannot be completed on schedule. The project buffer is used to ensure that the project is completed on schedule.

According to goldratt, the author of the key chain method, and some other authors, if the buffer length is set properly, the average project duration can be shortened by 50% over the traditional method, and the project can be completed on schedule. There are roughly three buffer length calculation methods:

SRSS
Task chain length: 1/2
Security Time 1/2 Method

The following is a simple project with only three tasks. Because the same resources are allocated, all three tasks are on the key chain. Three methods are used to calculate the project buffer length, and the results are significantly different.

The SRSS method is obtained based on the mathematical method of probability statistics. It was first proposed by Donald reinertsen. He proposed that after a set of serial processes are merged, the uncertainty of the serial process will be reduced to a part of the total amount. Specifically, for serial activities, the total uncertainty is the square root of the sum of squares of uncertain quantities. SRSS stands for square root of sum of squares ).

A task is a serial activity, and a task chain is a combination of serial activities. The task schedule is uncertain. We often say that the task "Module 1 encoding" takes three days, but in fact, various unexpected situations will lead to task delays, such as hard drive burning and power failure. Therefore, the task schedule is not a fixed value, and it is a random variable that is subject to a certain distribution, accurately speaking, for the task "Module 1 code", if everything goes well, it can be completed in three days. On average, it will be five days. If all kinds of accidents happen, it will take seven days, and the construction period will be uncertain, there is a change scope.

Similarly, the total construction period of the task chain composed of multiple tasks is also uncertain. There is a change scope, which is equal to the square root of the sum of squares of the change ranges of each task. If the change scope is used as the buffer length, the progress plan is made based on the task schedule in the best case. In the worst case, the project just consumes all the buffers, which can greatly shorten the entire project schedule, in addition, the project can be completed on schedule in various situations.

In theory, SRSS is the most precise, and it is the theoretical optimal solution of the problem. Therefore, compared with other methods, SRSS usually gets a much shorter buffer length. In the preceding example, the buffer length calculated using SRSS is only 4 days.

But in fact there are not many people who like to use SRSS. First, SRSS is theoretically optimized, But it assumes that the schedule change of each task is independent, which is sometimes not in line with the actual situation, for example, the same accident may lead to the delay of multiple or all tasks. For example, a power failure may lead to a delay of some time for all ongoing tasks. Therefore, the buffer length calculated by SRSS is usually too short, and the project end date cannot be fully protected.

Another reason is that the input is troublesome. SRSS must use the two-point analysis method or the three-point analysis method (PERT method). If the two-point analysis method is used, you must enter the average duration of each task, in the worst case, you need to enter two construction periods, which is twice the input workload of the traditional scheme. The three-point analysis rule requires three construction periods, which is more troublesome. In the case of heavy work pressure, it is intolerable to enter a lot of trouble.

Therefore, SRSS is mostly used for theoretical analysis to explain why buffering can shorten the time without affecting project completion on schedule.

The other method is the task chain length 1/2 method. The buffer length is 1/2 of the duration of the task chain. Projects with high uncertainty can have a higher proportion than 1/2, while projects with low uncertainty can have a smaller proportion.

The advantage of this method is that it is simple, does not require two-point analysis, and does not require two periods. In addition to the average duration, the data size is the same as that input by the traditional method.

In addition, this method only requires an estimate of the overall uncertainty to directly obtain the buffer length. It is convenient to use without estimating the uncertainty of each task.

Another advantage is flexibility, which can be easily corrected based on actual conditions and experience. For example, for projects with more than 50% of uncertainty, many companies cannot accept a buffer of 200%. In this case, they can change the ratio to 100%, the calculation method does not need to be changed. It is not easy to modify the SRSS method.

Another feature is that it is easy to calculate. In this way, the buffer length can be obtained through manual calculation without the help of software. Because the 1/2 method is simple and flexible, this method is used in many projects. The buffer length obtained by this method is usually much longer than that of the SRSS method, and the protection for the project end time is more adequate, it is more advantageous to the project team. Because the ratio can be changed flexibly, the company, the customer, and the project team can determine a acceptable proportion through negotiation and take into account the interests of all parties.

As in the preceding example, the buffer length calculated by the task chain length 1/2 is 6.5 days, And the SRSS returns 4.12 days.

The last method is the security time 1/2 method. The two-point analysis method is also required. The buffer length is equal to half the total security time of each task. This method is the same as SRSS in terms of input troubles. In most cases, the calculation result is similar to the task chain length of 1/2. The only advantage is that it is easy to calculate, but now PC is very popular. This advantage has little significance. So this method is rarely used, and few people know it.

Among the three methods, I usually think the best method is to use the task chain length 1/2. No matter how detailed the theory is, the deviation from the actual situation will be very large. It must be corrected by experience to obtain accurate results. The 1/2 method is simple, flexible, and easy to modify. Compared with SRSS, it is easier to improve accuracy in practical applications. In addition, it is easy to understand and learn, and it is not easy to make mistakes.

From: http://www.folo.cn/user1/12320/archives/2007/41484.html