Schedule distribution

[thinkingbox]

Build a schedule distribution so that for every task we can see that its end date has a certain percentile distribution. This means that rather than projecting a task distribution on an individual schedule, we need to build a schedule distribution with all project- & task-runs. The mistake that we made at the time was to start from the previous average and add the next task, which is wrong, because it destroys variability (hence the likelihood that a project may fail). See clt.scala.

References:

• http://www.math.uah.edu/stat/sample/CLT.html

[thinkingbox]

To clarify, given a project made of a bunch of tasks and related estimates, we run multiple iterations and we obtain a project-run distribution and its related task-runs distributions. A schedule is a projection of a project-run and its task-runs on a calendar. A schedule distribution is a distribution of all iterations calculated previously. The likelihood of a project to be late is independent of the schedule, provided that when choosing strategy 2 the desired times are calculated appropriately (in strategy 1 a project is never late because the the desired time is defined as the necessary time to complete a task => there is no schedule compression).