(1) Explain your problem, don't simply post "This isn't working". What were you doing when you faced the problem? What have you tried to resolve - did you look for a solution using "Search" ? Has it happened just once or several times?
(2) It's also good to get feedback when a solution is found, return to the original post to explain how it was resolved so that more people can also use the results.
Hey folks. I've recently ben made a junior planner and trucked of for P6 training to manage an existing project plan. Unfortunately I don't have anyone of seniority or technical experience to field amateur hour questions with, thankfully I've found this website and hope to get some answers.
My current pressing query (that I first referred to the wiki to try and answer and googled furiously to no avail) is a decent definition of what a P10, P50 or P90 schedule actually are. I know they relate to schedule risk analysis but I couldn't give a decent description if someone asked me. I've heard managers throw the terms around a lot and am increasingly anxious that I don't know what they're referring to.
And assistance or advice would be greatly appreciated.
Printer-friendly version
Send to friend
Oumar,
In future, please post a new topic rather than resurecting a long-dead one.
Tu run monte carlo analysis, you need:
a schedule
a risk register
monte carlo software
some training on how to use the monte carlo software
There is also lots of informative threads in the "schedule risk analysis forum", such as the below:
Cheers,
G
i want to know how to do monte carlo sumilation
i want to estimate the reserve for P10, P50 and P90
what kind of data do i need to have to generate
thank you
Vladimir,
"simulation are not valid if resources are assigned different way" no doubt about it. It makes no sense to use Pertmaster for Monte Carlo on a schedule that is managed using different software, that will yield different resource leveled schedules with different probabilities of occurrence. Too much of a "marronazo", Spanish for hitting with a hammer to make it fit no matter what.
I would never trade good resource leveling because of a probabilistic model. There is a weight ratio of 1 to 100 that points toward good resource leveling algorithms and a ratio of 1 to 1,000 with regard to functionalities such as true shift model and partial assignments, if I combine the ratios the global ratio increases, if I include other functionalities the global ratio increases even further.
As I said before I do not believe any probabilistic model is precise because it is based on many assumptions, not only because of assumptions regarding the assignment of probability distributions but also because of an even greater assumption that on delayed project no recovery action will be done, this is not how it works. As a project is delayed management will take some action and the model will be wrong. Still the "marronazo" is not justified, that is too much to add.
To me they represent trends mostly based on a fixed model you shall still apply what-if scenarios, real life is not that simple. Again I believe these are still of value.
I can see that because of todays computers Monte Carlo is becoming within reach of the average personal computer and the functionality shall be encouraged. The more people get into accepting the tool and its limitation it will be easier for those of us that believe on its value to use it.
In practical terms I see no much difference between SPDM and Monte Carlo as the computing capacity of new computers is closing the gap. Yes resource leveling algorithm and other resource modeling functionalities matter, and become more relevant as the gap in computation time between probabilistic methodologies is reduced.
The theory by itself justifies the need to target for early finish in the hope of finishing on time. At home we are required to use in our contract baseline schedules all available contract time. This prevent us to plan true resource loading and activity duration targets without all the misleading noise it brings using buffered activity durations.
For example if an activity is scheduled 10 days without rain delay it will be projected to finish earlier than if scheduled for 14 days on a rainy season. But what if it does not rains and you schedule successors based on this data? You also need to plan resources adequately for the activity to be performed in 10 days, not in 14 days.
You have no idea how difficult is is on this side of the Atlantic to convince owners that the targeting of "early schedule" is needed to conform a deterministic schedule finish with adequate probabilities of success. If the contract finish date is set equal to the deterministic schedule then the probability of success will be low.
Our practice is self defeating. Makes no sense to compute probabilities of success if you are not allowed to target for an early schedule, if you cannot implement the most basic options for reducing uncertainty and the level of risk. Makes no sense to care about resource loading if you are not allowed to plan for real activities duration but for distorted activities duration and their equally distorted resource loading. Perhaps this is why our scheduling culture is so low.
Best regards,
Rafael
Rafael,
the main problem is the difference of resource leveling algorithms of Pertmaster and other softwares.
Imagine that you use P6 for scheduling and Pertmaster for risk analysis. Pertmaster creates probability distributions basing on own algorithms but you manage different schedule. The results of Pertmaster simulation are not valid if resources are assigned different way.
Best Regards,
Vladimir
Ronald,
I was skeptical about Pertmaster and asked a friend, he told me Pertmaster can be set to perform resource leveling after each Monte Carlo iteration. He sent me the following screen.
I still was in doubt about how it was performed and ask him to run a sample job where resource availability was zero for a period of time a few days after deterministic finish date. It came out as it should be, the finish date distribution displayed finish occurrences before the period and some after, none within the period.
Maybe it cannot perform resource loading and leveling as good as other software but proved me wrong about Monte Carlo running on a laptop within reasonable time, seems like today computing power is making true Monte Carlo simulation available on the laptop computer. But if the algorithms are not as optimal as can be in any case will be on the safe side for buffer determination purposes.
Still there are other limitations with Monte Carlo Simulation that makes it not as precise as many believe. It is a pity at home we are not allowed to submit early finish baselines in the hope of finishing on time, we are required to submit a schedule that uses all contract time and reserves no float in the form of a Contractor Buffer, they still insists all float belongs to whoever uses it first and that this requirement cannot be waived through the use of buffers.
Best regards,
Rafael
Mike,
This is the eternal problem with the use of acronisms. My guess would be it means the date or project duration for which the cumulative probability of completion date is 10%, 50% and 90% respectively.
To me the most important lesson from scheduling probability theory is the fact that the deterministic schedule is NEVER the most probable, that you got to plan for earlier completion in order to increase the probability of success to the most probable target date and that this value for the most probable target date might still have low probability of success.
In the above example the deterministic schedule have a 13% probability of success, a very low probability. If the contract end date is 02/Apr/02 the way to increase the probability of finishing within contract time is to schedule for an earlier deterministic schedule finish date. The theory is not there just to show you what your probabilities of success are but to help you to take control on increasing the probabilities by focusing on activities that will make the difference as this might come at high cost.
That the theory of probabilities is not exact and is based on some assumptions does not means it is useless, just not as precise as some believe. It can guide you as to how much buffer is needed for a deterministic plan have a reasonable probability of success. It can help you identify what other near critical path have high probability of becoming critical by providing some metrics for this purpose such as the criticality index.
Unfortunately the stupidity of the American system does not allows you to plan for earlier completion and require you to model using a wrong deterministic plan based on using 100% of available contract time. What an irony, here we can compute probabilities but we cannot use the findings in the correct way. The correct application of these findings is to target for an earlier deterministic plan, inflating activities duration is misleading and mathematically wrong.
Best regards,
Rafael
Hi All
So what does P10 - P50 - P90 mean.
I have not heard of the terms either.
As you know Rafael I do not use resource levelling software for live projects but I can see its use in a risk evaluation operation as shortage of available resources is a major risk factor.
I have never had to use monte carlo but it is an add on for PowerProject if needed.
Best regards
Mike Testro
Ronald,
“Does any existing Monte Carlo analysis software out there perform resource leveling during the analysis?”
I agree with David Hulett's argument but wonder about Deltek or any other commercial software doing the right thing. In large networks it would take hours to run the simulation on a common computer, this is not practical in commercial applications.
Consider the following statements.
Statement no 1: Any software that performs Monte Carlo without resource leveling is wrong as the delay on activities duration might cause delays on the schedule because of resource availability in addition to the activities duration increase or reductions.
Statement no 2: Theoretical Monte Carlo simulation assuming no management action can be questioned because the simulations simply carry through each iteration unintelligently. In the real world, it is likely that management will take action to recover projects that are severely behind schedule, and some of these actions may (though not always) help bring the project back into an acceptable schedule range.
Statement no 3: Assumed durations distribution can be questioned and labeled as a guess as for each of the hundreds of activities you would need hundred of samples to get a statistical distribution that has statistical value.
I believe Spider Project never implemented Monte Carlo because Vladimir and their team are aware of this, they implemented a methodology that does not uses Monte Carlo but calculates probabilities based on three individually resource leveled jobs. It is not old school PERT that was mathematically wrong nor Monte Carlo. Kind of a practical compromise, in part because of computation time and in part because of the subjective assumptions on any of the methods.
In terms of pure quantitative values even if statement no 1 is correctly addressed because of statements no 2 and no 3, any Monte Carlo Simulation can only be considered a roughed estimate.
For strategic planning these methods are useful but for a quantitative presentation in court I doubt it.
Best regards,
Rafael
Rafael,
What an interesting point! Monte Carlo analysis using resource leveling? It had not even occurred to me before you cited David’s paper. Of course it would take longer to process but the real question is; “Does any existing Monte Carlo analysis software out there perform resource leveling during the analysis?” David Hulett's White Paper was sponsored by Deltec but I could not find any such Monte Carlo CPM risk analysis software on their website.
Aaron,
Welcome to Planning Planet.
The following reference can provide you with a basic introduction of what it might mean.
http://www.coepm.net/wp-content/uploads/Whitepapers/wp_scheduleriskanlys...
Please note that, as stated in the reference:
The risk analysis software package should be able to level resources as it is iterating. This increases run time substantially.
I believe most software packages providing Monte Carlo do not level resources as it is iterating. This you can deduct by the time a complex schedule takes to perform a single resource level versus the time the Monte Carlo simulation takes, if it takes 30 seconds to perform a resource leveling run then 1,000 iterations shall take 30,000 seconds or 500 min or 8.33 hours.
Beware of the false prophets.
Best regards,
Rafael
P.S. Next time try posting issues regarding probabilities and risk at the forum Schedule Risk and Schedule Risk Analysis there you shall get better response.
Email: PPadmin@planningplanet.com
Replies