Saturday, August 28, 2021

The worst error in scheduling

The worst error in scheduling is to have a milestone depend upon two or more tasks scheduled (planned) to finish at the same time.
Here's a footnote to that witticism: It's assumed the two tasks are independent, meaning:
  • They don't share resources
  • They don't have the same vulnerabilities to a common risk
  • The progress, or not, in one does not affect progress in the other
So, what's the big error here?
  • First, as regards milestone success , each of the tasks leading into the milestone is a risk to success (success means: it is achieved on time)
  • Second, total risk is the product of all the input risks (as seen by the milestone) . 
  • So, whereas each task coming into the milestone may not be too risky, say by example 90/10 (*), three tasks of 90/10 each would present a risk to the milestone such that success is reduced to about 73/27
(*) 90 successes out of 100, or 90% chance the task will finish on time, or early.

What are you going to do about this?
Bring on the time buffers! (**)
  • You might be able to add a buffer on one or more of the input tasks to raise the success of that task to 99/01, or so
  • You might be able to add a buffer following the milestone, such that any late success is absorbed by the buffer (This tactic is called "shift right" by schedule architects)
  • You might be able to reorganize the schedule to eliminate this milestone, or one or more of its input tasks.
(**) A scheduled event of zero scope, but a specific amount of time, aka: a zero-scope time box.

Buy them at any online book retailer!


  1. The worst error is to plan to have a "event" complete with multiple Finish to Start constraints and have NO margin in front of that event.
    It's not the FS constraints that are the problem, its the lack of protection for the aleatory uncertanties to the left of the Event. See the GAO guide

  2. Glen: your comment is spot on; I raise your point about the milestone buffer in the second of three points at the bottom of my original posting.

    Of course, once the matter of schedule architecture is addressed, then comes the more messy issue of schedule management. Who sets up the buffers, and who gets to say how large are the buffers? PMO or work-package managers?

    Eliyahu Goldratt discusses some of this in his book on Critical Chain scheduling, where he makes the case that critical path milestone buffers belong to the PMO. Perhaps that is the way it should be, but ....


Please add your value with a comment