Wednesday, January 30, 2013 3:38 PM
I’m generating some schedule metrics using MSProject data and was wondering if anyone might have any insight on a particular metric I am trying to create. I am trying to answer the general question of how the project is progressing on the tasks. No resources are loaded into the schedule. I’d like to see if we are slipping tasks to the point where there is not sufficient time remaining to recover --- a bow wave effect. The analogy would be everyone waiting until the last minute to board the plane. If this happens, the plane will be late departing. I am envisioning a trend chart but can’t seem to figure the right way to go about configuring MSProject data. Any insights or other resources to pursue would be appreciated.
Wednesday, January 30, 2013 4:04 PM
No resources are assigned? So how are the tasks being completed? It sounds like you are working with a "schedule" that is for show only so metrics can pretty much be spun any way you want.
Without resources the only method of showing "progress" is via the Percent Complete field. Is there a reason why the progress lines shown in the Gantt bars isn't sufficient? That definitely shows how tasks are "progressing".
Another good metric in you case is total slack. By monitoring the Total Slack field, you can see if your are losing or gaining ground, so to speak, on reaching the end goal. If the numbers are negative, you're in trouble. If the numbers are decreasing, you're headed for trouble. If the numbers remain relatively constant or increase, you're gaining ground.
You could also look at duration variance assuming your set a baseline and there are other metrics that already exist.
My two cents.
Wednesday, January 30, 2013 6:18 PM
Thanks for the reply John. I have people assigned using the text field to assist in reporting. So, when I mention no resources, I'm referring to not doing resource-level scheduling. I am doing critical path analysis and have other metrics using total slack as a method to assign criticality. I also have metrics to check for tasks that are missing the start or finish dates (against the baseline as well as the Status date). My dilemmna is how to derive a metric that provides insight to when enough tasks are slipping that it shows less of a likelihood that all remaining tasks can be completed in the time remaining.
Wednesday, January 30, 2013 7:46 PM
The metric for "...insight to when enough tasks are slippping..." is of course subjective. It sounds like you want to work with probabilities. In Project one method of doing probability analysis is via Monte Carlo simulation. As far as I know, (I don't have Project 2013), such simulations are not a part of Project's functionality but it can be added via third party add-ins. You might want to check out the vast array of companion products for Project at: http://project.mvps.org/comprods.htm
You might also be interested in Jack Dahlgren's Blackjack Monte Carlo simulator at: http://masamiki.com/project/macros.htm. He wrote it several years ago so I don't know how up to date it is but it may work fine as is.
Hope this helps.
Wednesday, January 30, 2013 10:18 PM
I'm not sure the Monte Carlo simulation would help much in this case. It depends on a having a robust schedule with tasks linked - and on having some info on the probabilities for task durations. If you already have the robust schedule with task dependencies, you already have the information needed to deduce whether the schedule is achievable or not. If a critical path task is slipping, then the schedule is slipping.
Or is the question about the accuracy of estimating? If one task estimate turns out to have been optimistic (the task is actually slipping), maybe other estimates are also over-optimistic. One can imagine a metric based on the average schedule variance, perhaps as a percentage of task duration. If actual task durations consistently run 10% over their baseline duration, maybe all remaining durations should be increased by 10%.
Friday, February 01, 2013 8:45 PM
I agree that it's not a Monte Carlo simulation that will produce the result I'm looking for. I've actually already run those with @ Risk SW on my schedule. I'm actually trying to devise a metric (perhaps using one or more of the variety of fields like Variance, Total Slack, etc.) that will serve as a relative indicator on feasibility of completing all the work in the schedule when individuals continually slip their tasks. I do active critical path analysis. However, it's possible that the project end date does not change yet the likelihood of achieving the schedule may go down because tasks slip and now there is less time (compared to the previous status period) to get the remaining tasks done. Poor estimates can certainly be one of the root causes. I am envisioning some use of MSProject fields/data calculated each status period so that I can see if the schedule do-ability is being compromised --- looking for a relative trend and not an absolute metric, to supplement the CP output.
I'm wondering if I have to look at past history to see the rate at which tasks were completed and then applying this to future tasks (still to be completed).....
Saturday, February 02, 2013 2:23 AMModerator
I think your only chance is to make sure you have a baseline and update Actual Duration and Remaining Duration. Then you can calculate the schedule accuracy for each task based on the formula iif([PercentComplete]=100,[Actual Duration]/[Baseline Duration],1)
You can the calculate the predicted finish date for each summary task using an average value for the summary task.
Or you can calculate a timing accuracy value using a similar formula but based on finish variance.
Neither of these is as accurate as assigning Resources, predicting and tracking actual work and using Earned Value calculations to calculate the new estimated finish date.
What ever the method used, its accuracy will be proportional to the accuracy of your scheduling data.
Rod Gill Project Management
Saturday, February 02, 2013 2:26 AM
Way back when I developed a predictor algorithm on our project to show management what was likely to happen given the current trend. Understandably, it wasn't very popular with those responsible for completing the effort. So, yes, you could look into the bowels of the plan and develop trend metrics but as they say in the investment world, "past performance is no indication of future returns". I guess you have to ask yourself, is trend analysis any better than critical path analysis and good project management. Piling on more metrics may not be value added; just make sure you strike a balance between metrics and proactive management.
Just my thoughts.