A typical approach to project planning is to break down the problem into smaller tasks, estimate them individually, and then sum up the estimations in order to come with total project duration. The hidden assumption is that variability of estimations is symmetric and cancels out.
This might be true if we keep estimates to ourselves. But if we use estimates also for project management (i.e. expecting people to keep their "promises") human psychology kicks in and actual variability of task duration looks more like this.
Or put in words, tasks seldom finish early while they are often late.
This phenomenon is due to Parkinson's law which states that
work expands so as to fill the time available for its completion.
One interesting corollary that follows from this observation is that estimates are not only a waste, but actually harmful since they prolong the project duration!
An estimate protects people from project pressure at first, but then returns as a boomerang only to strike people in their heads once the task is taking longer than estimated. My suggestion to project managers would therefore be to abandon estimations altogether and come with better ways to plan and manage projects.