none
How to add work/duration uncertainty to a task/project? RRS feed

  • Question

  • Hi all.  Is it possible with Project 2013 to specify min and max work required (and duration) for a particular task?  One could also describe these figures as best/worst case scenarios for work required (and duration).  

    I ask because I frequently prepare estimates for clients that indicate how much work a project will require, and how long it will take.  I typically give them a range for each (e.g. 80 - 100 hours of work and 2 - 3 weeks in duration).  Other than maintaining three separate Project files (one each for best/worst/expected case scenarios), I was wondering if Project had any capability to manage and account for such uncertainty.

    Thanks in advance.

    Monday, October 17, 2016 8:28 AM

Answers

  • The answer to my original question seems to be to create three separate baselines:  one for min estimates, one for max estimates, and one for expected estimates.  That way you can display and edit them all on one table.

    But as others have pointed out, this may not be the most sound methodology to account for risk in my project.  I will look into the Monte Carlo plugin mentioned and see if I can get it working, if anything to compare the two methods.

    • Marked as answer by cagross99 Tuesday, October 18, 2016 3:56 AM
    Tuesday, October 18, 2016 3:56 AM

All replies

  • You can record that information in the custom fields, e.g. Duration1, Duration2, etc. for min and max.  

    That being said, what you are probably getting at without asking is how to do Monte Carlo Simulation of project schedules.  That is possible using commercial add-in products to Project that add Monte Carlo functionality to Project, or connect to Project files to do it.


    --rms www.rmschneider.com

    Monday, October 17, 2016 9:25 AM
  • That capability is not built in. There are 3rd party tools for MSP which perform rigorous monte carlo analysis based on ranges of estimates.

    However, once you have a plan, and it is full of one set of estimates and assumptions, then the answer is the answer, subject to all of those estimates and assumptions,.

    So can have a Plan B and Plan C and so forth, each one with a different set of estimates and assumptions, and whatever the answer is for each one that is the answer.

    Standard risk analysis would require the consideration of a range of possible outcomes.

    Any help?

    Monday, October 17, 2016 9:51 AM
  • OK thanks for this.  From your instructions, I see how I can use custom fields to set the duration min and max for a particular task.  But is there a way to do this for work hours?  For example, a task may require 4 to 6 hours of work, but have an overall duration of 2 - 4 days.  Can I specify all that information for a particular task?

    The goal would be to then aggregate each properly and get an overall min and max for the entire project.  Doing so for the work would be trivial--since there is only one resource, I would just need to sum the min work, and sum the max work.  Trickier though might be calculating the min/max duration for the entire project, given the min/max duration for each task.

    >> That being said, what you are probably getting at without asking is how to do Monte Carlo Simulation of project schedules.  

    I did a little reading on Monte Carlo simulations in PM, and I'm not yet sure if this is what I want :-) I read this very basic illustration of the technique.  As I understand, it replaces the min/max aggregate figures for the project with more accurate min/max statistical figures it pulls from a probability distribution.  But the upper and lower bounds of this distribution still correspond to the min/max aggregate figures, correct?  As an initial step though, it would be very helpful to me if Project could calculate just these upper and lower bounds for me, if possible.  What would actually be even more illuminating would be to then calculate the Monte Carlo figures, compare the two, then compare them to my actual figures when the project is finished.

    This is my first time using Microsoft Project, and I've chosen this particular project as more of a learning exercise than anything, since it is relatively simple.  It has only a single resource (me).  The total hours required are ~110 and the total duration is ~60 days.  However it's not as basic as the example project in my previous link.  Many tasks in my project have lag of varying amounts, and many (but not all) have finish-to-start links.

    Thanks!

    Monday, October 17, 2016 10:40 AM
  • Thanks for this.  I posted a reply above (to @Rob Schneider), but couldn't figure out to also call you out in the text.  Much of what I wrote was a reply to you as well.  Specifically, I'm not sure if I want a Monte Carlo simulation just yet.  I think I'd just like to specify a max and min for each task, then have Project aggregate up all these maxes and mins for me into a total max/min for the Project.  I think what you're saying is that I should just maintain three separate baselines--max, min, and expected?
    Monday, October 17, 2016 10:45 AM
  • You mention this is the first time you are using Microsoft Project.  I think you expect too much about things it is not really designed to do. It facilitates doing what you want (Monte Carlo simulation), but does not do it.

    Work is normally computed (or entered) and does not have Work1, Work2 custom fields. But you can define yourself custom number fields named, say, "Work min", "Work Max" and store that information.  

    But (a big) However, you have to develop yourself the algorithms and methodology for how you sum up and otherwise use those numbers.

    I think what you are intuitively trying to do is solved with Monte Carlo simulation and Microsoft Project using both in a way as designed to get the proper answers. But you probably need to spend a little time learning and understanding that. Building all this on your own requires learning and understanding also but there is risk you'll end up at a different place than current state of the art.


    --rms www.rmschneider.com

    Monday, October 17, 2016 11:27 AM
  • OK thanks.

    >> I think you expect too much about things it is not really designed to do.

    I'm not expecting anything, I'm just asking.  But I acquiesce that I may not understand some of the principles well enough.  I will go back and make sure I understand work and duration properly.  Then I will revisit this question.

    Monday, October 17, 2016 11:49 AM
  • The answer to my original question seems to be to create three separate baselines:  one for min estimates, one for max estimates, and one for expected estimates.  That way you can display and edit them all on one table.

    But as others have pointed out, this may not be the most sound methodology to account for risk in my project.  I will look into the Monte Carlo plugin mentioned and see if I can get it working, if anything to compare the two methods.

    • Marked as answer by cagross99 Tuesday, October 18, 2016 3:56 AM
    Tuesday, October 18, 2016 3:56 AM
  • I will support Rob's comment about what I think he is saying about getting a firm grip on the basics before getting more ambitious. Perhaps a crash course in the critical path method and statistics.

    You make the comment that you want to estimate the duration and the work. Not really the way MSP is designed to be used. You can save yourself the trouble of having to do this if you start with a duration estimate, assign the resources, and allow the work to be calculated. It's an output, not an input.

    So you estimate a task will take 10 days. It could be 8 or maybe 12.
    You assign Bob at 50%, but he may really be available 40% - 60%.
    The work is calculated to be 40 hours, and it will be uniformly distributed as 4 hours/day (the first 4 hours) over the 10 days, but that is just one outcome out of a range of possible (or probable) outcomes.

    Even when making single value estimates we know that that is a just a short cut way to cut through the inherent uncertainty, and get to a plan which is practical and useful, even if we know that everything estimated can and probably will turn out different..

    Wen we make single value (deterministic) estimates we know that they are really a number selected from a range of possible values that we might have in mind. We just ignore that and go ahead and choose the single value estimates which feel about right and build the project plan and get an answer, which is one answer more than you have otherwise.

    Tuesday, October 18, 2016 11:20 PM
  • >> You make the comment that you want to estimate the duration and the work.

    If I conveyed that, my mistake.  I am definitely interested in estimating both work and duration.  

    It's true that I am a beginner at project management and project management software.  But I think I understand everything  that has been said in this thread.  I asked if Project could calculate uncertainty using <method A>.  Replies indicated that <method A> is undesirable, and that I should instead use the Monte Carlo method to assess uncertainty, since it is more realistic and more accurate.   I did a little reading on the Monte Carlo method and I see how it is indeed a more realistic and favorable method of determining uncertainty.  

    I also understand your simplified scenario involving Bob and his 10 day task, and that Project gives us one (very useful) answer out of an infinite number possibilities.  In light of that, how would you suggest I calculate uncertainty for clients when developing projects?  I'm a beginning freelance web designer, and give my clients a range of work and duration before starting a project.  Would the Monte Carlo method still be the best solution for me?  Or is that not something I should try to carry out with MS Project--e.g. perhaps I learn to quickly estimate that range myself, by eye, as I gain more experience.  The projects in-question are relatively low work (~100-200 hours) and low duration (1-2 months) projects with a singe resource (me).  I am indeed interested in giving clients the most accurate estimates I can.  But at the same time, if I'm off by 10-15%, it's typically not the end of the world.

    Thanks again.

    Wednesday, October 19, 2016 5:09 PM
  • Dealing with the uncertainty of future plans in a rigorous statistical way means that you have to very good with statistics (well, good enough) and then get all the tools and learn to use them.

    Definitely justified if you have to calculate the probablity of getting the oil rig built to a point where it can be launched before the sea freezes over.

    However, in your case you can just rely on your experience.

    Wednesday, October 19, 2016 11:14 PM
  • OK thanks so much for all that.
    Thursday, October 20, 2016 12:24 AM