You may want to go in a different direction. Project will calculate duration
based on resource assignment units and work assigned to the task. You may
want to read the FAQ about differences about tasks types: fixed work, fixed
time, fixed units. It also gets more complicated by "Effort Driven" which
determines if adding resources will increase work or divide the work
remaining over the new resource count.
In reality, PERT analysis with Project really requires some additional add
ins. Your project will not really take the Optimistic or Pessimistic
times.... Project will give you 3 Gantt Charts. However, the reality says
that each task will vary between it min and max duration. Just because task1
took "optimistic" time, uncertainty tells us it is unlikely the next task
will also take "optimistic time". Enter the add on for Project called Risk+
which is owned by Deltek.
Risk+ allows you to do a Monte Carlo simulation of your schedule to predict
a probability distribution of finish dates for each task and the project
itself. This is done by randomly selecting durations for all the tasks that
are between Optimistic and Pessimistic. Repeat thousands of times and you
will begin to see how the project will probably shake out.
That said, you still need a plan to generate the three duration estimates.
If you say a task will take 20 days, then how do you determine the optimistic
time is 15 days or 12 days or something else? How do you know the
pessimistic time is no longer than 45 days or 25 days?
Experience is the answer to that. Look at past performance on similar
tasks. Think about how complex the task is, are the specifications settled,
how likely is it your customer will change the requirements on that task? How
senior is your team? How familiar are they with the tools being used? Are
there better tools available to help you do it more quickly? All these items
will help you decide the range. Consider a jury of peers to argue it out in
a scheduling meeting. For sure, one guy estimating the task three magic
durations is not the correct way to go... that one opinion is biased. Seek
opinions from multiple Subject mattter experts, research online, look at
historical performance. You can also look at cost data... if you have
historical costs on a specific project you can probably guestimate hours or
durations from the cost data.
There are also multipliers available in some of the text books. One example
may come from a book called Code Complete, or some of the other books on
Software Engineering if your project is SW related.
--
If this post was helpful, please consider rating it.
Jim
Visit
http://project.mvps.org/ for FAQs and more information
about Microsoft Project
