Critical Chain with buffer

When we have to fit in a time or when we have to pay for something where there is uncertainty, we automatically think that we must have a margin. When we travel by train, we make sure to go to the station, with a sufficient time margin, to withstand a little worse traffic than usual. If we are going to pay for dinner, where we know that you can only pay in cash, we make sure to include what we think it will cost plus a margin. When we are going to dimension the margin or the buffer, we quickly realize that it is not smart to try to break down the buffer into sub-buffers. This requires a larger buffer. One buffer for starters, one for the main course, and one for drinks require a larger total buffer. With a common buffer, you can take advantage of the fact that it is unlikely that all 3 will be more expensive than calculated. The simplest and most resource-efficient is to gather the margin in a buffer that protects the entire commitment. If the starters became 40% more expensive than expected, the main courses might be 15% cheaper, while the wine might be 10% more expensive. With separate buffers, you probably have to have a 50% margin to be safe, but with a common buffer, half may be enough, around 25%.

When Eliyahu introduced buffers in his project methodology, he started from this, which is also called the law of large numbers. The rule of thumb was formulated as that the variation of the individual activity is such that a buffer is needed that is 50% of the Critical Chain, i.e. the project's unbuffered extent, to protect it against the typical variation of the project activities. To force the margin away from the individual activity estimate, Eliyahu introduced the rule that if an estimate is made with a margin, it must be divided into half and half the difference must be added back for the new common buffer. In this way, you get plans that are both shorter and better protected against variation than the original.

The buffer must be inserted so that it protects the project's deliveries to the customer.

By removing the local buffer in the estimates and adding half back to the buffer, you get a plan that is both shorter than the original and more robust to variation.