advice please on guestimating confidence levels

Gary,

If you are using Primavera software at a click of a button you can transfer your job into Primavera Pertmaster. Then you must add your duration distributions for each activity but you might choose a few that automatically determine the distribution form the single deterministic duration. Now you are ready to rumble.

After you run the Monte Carlo Simulation you will get your job deterministic duration as well as the probabilistic distribution. You will find the probabilistic distribution shifted to the right of the deterministic duration. This is due to the fact that at times non critical activities are delayed and become the critical path. That is another reason why many schedules are optimistic.

Believe me the theory is not difficult to understand, because there is no mathematical way to estimate total job distribution for duration you have no other option than use Monte Carlo, this will run your job hundreds or thousands of times with different combinations of individual activities duration based on their individual duration distributions and from this you will get the distribution. Help from within Pertmaster should be enough.

The software should provide you with parameters that will show the probability of non critical activities becoming critical. One of these should be the delivery activity form your external job. I suggest using a single long duration activity for this purpose with a wide spread duration distribution, also a distribution shifted to the right of the initial deterministic duration.

Get your hands dirty with Pertmaster, in no time you will be up and running.

Cordially,
Rafael

Thanks gents for your replies

The problem I have is how best to represent the key external delivery miestone for monte carlo purposes.

The CP to the delivery milestone is:
-Client funding approval process
-Contract Award for EPC Project A
-EPC project A approx 80% complete
-Area released for my project

I was worried that representing this entire path using a single activity would be too simplistic to get sensible results from the monte carlo? Especially as I know so little about them.
My gut feel is that these predecessors have a fairly high risk of overrun, and we would want at least 10% terminal float on our project to mitigate this risk to say a 95% confidence level.
NB: This is not to mitigate risk of LDs, but the risk to the client of late delivery of a project required by it’s regulator.

Thanks,


Gary

Gary,

At your initial posting you expressed you desire not to include details of the schedule as you do not have access to the other project programme, but now I see you have access to more than a single milestone.

If these milestones are FS dependant in tandem maybe the compromise could be to use them as your summary activities, each with a different distribution. This will provide you with the opportunity to update the Monte Carlo run after each of these events happen. As with any schedule you will only get a prediction, a prediction that will vary with time, a prediction that will depend on your client estimates for probable durations.

The resulting distribution will provide you with a rough estimate of how much of terminal float to use, your selection will be based on a probabilistic distribution, not a deterministic number, your final choice will be based on how risk adverse you are.

This is as far as I can go, good luck.

Best regards,
Rafael