Tips on using this forum..

(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.

Why Schedule Risk results for P50 and P80 are very close?

9 replies [Last post]
Sharyar San
User offline. Last seen 27 weeks 4 days ago. Offline
Joined: 16 Sep 2001
Posts: 11
Groups: None
I am reaching out here because our client had a question and I had no answer for it. Here is our case: We have a fairly large schedule (7000 activities) in P6 and we did duration uncertainty ranging for 454 activities near to critical path (TF<2weeks) also remaining duration more than 2 weeks. Project duration is 2.5 years. Results off of PRA risk analysis show 50% and 80% are only a week apart. Now client says that they use to see at least a month difference between P50 and P80. Any thoughts why this is happening? I appreciate your help.

Replies

AKEEM OWOADE
User offline. Last seen 3 years 25 weeks ago. Offline
Joined: 13 May 2013
Posts: 3
Groups: None

Hi Shanyar,

Pls,i just started using Pertmaster but i am having a problem in creating What-If Analysis.Pls can you help out on this.Thanks.

Stephen Devaux
User offline. Last seen 13 hours 25 min ago. Offline
Joined: 23 Mar 2005
Posts: 607

Hi, Rafael.

"...(T)he wrong approach is mostly due to how contracting is done."

I couldn't agree more.

"...(T)hat the schedule can be used against the contractor is a fact we cannot ignore."

Contracts MUST align customer value drivers with contractor value drivers. Anywhere they aren't aligned will likely lead to the principal/agent problem and place the contractor at moral hazard. Thus customers MUST learn how to write project contracts that, instead of mandating some totally artificial artifact called a "deadline", accurately reflect their business value and provide incentives to the contractor to maximize the customer's business value. And, unfortunately, most customers are woefully ignorant of how to do this. (I hope some of them will read my book!)

"This is being fixed with new contracts allowing the contractor to have an explicit schedule reserve for his own that allows for transparent management of his schedule. This I believe is a movement on the right direction in the way we contract."

It is definitely a movement in the right direction -- but it's not enough, because time almost always has value for the customer and schedule reserve adds time. Instead of a deadline protected by schedule reserve, the contract needs to stipulate a sliding scale of payment based on earlier or later delivery, as it impacts customer value. Then everyone will be on the same page of a win-win contract.

And apologies to Sharyar San for slowly highjacking this thread. But I do think what we've discussed is important.

Fraternally in project management,

Steve the Bajan

Rafael Davila
User offline. Last seen 4 hours 21 min ago. Offline
Joined: 1 Mar 2004
Posts: 4763

Steven,

I guess we agree on 99% and the 1% is on not so relevant issues. My statements in this discussion I see in no conflict with yours, on the contrary. Although some of my statements might look radical like the issue on tweaking the schedule I really mean it, as well as all my clients, none is a fool and understand the need to have two schedules, but the wrong approach is mostly due to how contracting is done. 

Even probabilistic theory advocate for different schedules, the contractor schedule with some reserves for the contractor actions within contract time, while the client schedule with additional reserves for the additional impact of his actions. This is how it should work in theory but that the schedule can be used against the contractor is a fact we cannot ignore. This is being fixed with new contracts allowing the contractor to have an explicit schedule reserve for his own that allows for transparent management of his schedule. This I believe is a movement on the right direction in the way we contract.  

I must agree with you that the contract practice must still be further improved, especially in government contracts where I see most of the poor performance, usually delays by the designer/owner no body is accountable and paid with taxpayers money. My experience in private jobs have always been excellent. I never had a client placing a claim on private jobs while in public works the story is quite different. 

Best regards,

Rafael 

Stephen Devaux
User offline. Last seen 13 hours 25 min ago. Offline
Joined: 23 Mar 2005
Posts: 607

Hi, Rafael. As usual, we agree on far, far more than we disagree.

"Probabilistic scheduling is more of a scientific game strategy than a pure science because it is naive to believe we will get accurate and precise distributions,..."

I completely agree. My complaint is not with those who, like you, understand this, but with those who pretend that Monte Carlo systems (which they often sell for lots of money!) in fact provide some kind of accurate and precise prediction. In fact, as we have discussed many times over the years on PP, they are fraught with myriad inaccuracies and guesses (the distribution shape and the fixed lag values being but two) that the marketers never mention. 

"...still as a methodology is more useful than the worst approach that assumes 100% deterministic values."

We can agree that anyone who assumes the accuracy of deterministic estimates is an even bigger fool (or fraud!). And I suppose there are people like that out there. But in my experience, those using determistic estimates almost always are keenly aware that the actual duration can change. They know that a 10D activity with 6D of float might take 20D. What I want them to understand is that if that happens, it just went from having 6D of float to having 4D of drag -- and that if the drag cost is $20,000/D, that activity is now costing $80,000 more to perform! That info is as important a benefit of drag calculation as in the upfront optimization process. (I hope Vladimir's salespeople are demoing Spider's functionality in this way.)

I just think that the "scientific" veneer of the probabilistic approach and its snake-oil salesmen ("It's based in the mathematics of Abraham de Moivre and Frederick Gauss, you know!") has given the "probabilistics" unwarranted confidence in Monte Carlo forecasts about whose underlying mathematics and functions they know nothing.

In my new book (due out in September) I mention that I attended a presentation at the 2009 PMI College of Scheduling by one of the big "gurus" of that organization. He talked about the great results he got by using a Monte Carlo system for his estimates. I asked him if he didn't find the effort of choosing a distribution shape for each of the 10,000 activities enormously laborious. He gave me the reply I expected: "Oh, I run it on one of the default distributions." Which one, I asked. "Oh, I think the triangular, but it doesn't make much difference." Not much difference? 8%-12% change in duration estimate is "not much difference"?? He had no idea because he'd never bothered to check out the mathematics. And remember, he is regarded as a scheduling authority! (What is the drag cost of an extra 8% - 12% on one of his project durations, I wonder?)

"Deterministic durations are even easier to tweak."

I disagree -- they are identically easy to tweak! Once you decide you need a 10% or 25% or whatever management reserve, you're going to tweak until you have it. ("You know, this activity really could wind up taking 50D -- let's make that the pessimistic instead of 20D. And let's make this one 35D.") I will concede that deterministic tweaking is less time consuming -- but surely that's an advantage, considering you're going to get where you want to be either way.

"...(U)nfortunately the contractual trend is for the plan to be used as a claim tool rather than a managing tool.  Whether you like it or not these two come into conflict and the contractor has no other option than to tweak it, otherwise he would be a fool, at home would be called a "pendejo", Spanish for pubic hair."

Oh, I understand, believe me. The state of contract project management is a disgrace to the human race (and I have a lot to say about that in my book!). Until we get away from those absurd anachronisms misnamed "deadlines", contractors will have to continue to play games and use techniques that are either useless or counter-productive.  I have had my clients use Monte Carlos because they are often the only way to persuade their customers of the need for significant schedule reserve, or that a certain desired date could probably not be achieved. It works very well for persuading the unpersuadable. ("You wouldn't want to contradict Gauss, would you?")

Fraternally in project management,

Steve the Bajan

Rafael Davila
User offline. Last seen 4 hours 21 min ago. Offline
Joined: 1 Mar 2004
Posts: 4763

Deterministic durations are even easier to tweak.

There is no such thing as true probabilistic distributions on real life construction jobs because simply there is not enough statistical data on every risk element particular to each job.  Risk analysis is more about trends and sensibility analysis; about the possibilities of delays and changes in critical path chain. Better than the deterministic approach that only forecasts a single dominant critical chain at each update while probabilistic methods can warn you of unsuspected activities that can become critical with higher probabilities than current critical activities. 

Probabilistic scheduling is more of a scientific game strategy than a pure science because it is naive to believe we will get accurate and precise distributions, still as a methodology is more useful than the worst approach that assumes 100% deterministic values. We got to plan for a stochastic future, not the deterministic past, for this we need probabilistic methods. 

The real value of the schedule is for the one who will use it, unfortunately the contractual trend is for the plan to be used as a claim tool rather than a managing tool.  Whether you like it or not these two come into conflict and the contractor has no other option than to tweak it, otherwise he would be a fool, at home would be called a "pendejo", Spanish for pubic hair. 

As in any deterministic schedule that is tweaked to serve the whims of client you can consider tweaking the distributions and use your own probabilistic schedule to do your management. 

Stephen Devaux
User offline. Last seen 13 hours 25 min ago. Offline
Joined: 23 Mar 2005
Posts: 607

Sharyar San wrote:

"I used triangle shape for min/max ranging ."

Of course, if you think the triangular shape is forecasting the project schedule to be too long, you can just switch to the Beta default! In my experience, that right there should pull in your 80% forecast duration by between 8% and 12%! And if you think that result is a little too short, use triangular for half the tasks and Beta for the other half!

That's the great beauty of Monte Carlo simulations: whatever your gut tells you is the amount of time you're going to need, it's usually pretty easy to do a few tweaks here and there and get the system to forecast that duration.

Fraternally in project management,

Steve the Bajan

Rafael Davila
User offline. Last seen 4 hours 21 min ago. Offline
Joined: 1 Mar 2004
Posts: 4763

At the moment I do not have nay job for which Risk Analysis is required. I do not keep old files except those I use/create as reference to showcase methodologies. The following is a distribution form one of my reference jobs and the difference between P50 an P80 is 4 days, less than 10 days.

 photo SampleDistribution01_zps2e25648c.jpg

 photo SampleDistribution02_zps05eeccfe.jpg

The following is a distribution I got from the internet and it shows a difference between P50 and P80 larger than 10 days.

 photo SampleDistribution03_zps22e6161f.jpg

You can get from the internet distributions with a difference between P50 and P80 that is less than 10 days as well as distributions with a difference larger than 10 days. 

I do not see anything wrong with your estimated distributions, I am more risk adverse and plan for wider distributions toward pessimistic durations. This requires to target and plan for a tighter schedule in the hope of meeting dateline. 

Sharyar San
User offline. Last seen 27 weeks 4 days ago. Offline
Joined: 16 Sep 2001
Posts: 11
Groups: None
Rafael, Thank you for your reply. I used triangle shape for min/max ranging . With total of 450 activities ranged for SRA, 60% of them has min/max of +/-10% of remaining duration and the rest skewed to the right and left +/-30% of duration . We didn't want to skew more than 30% because our original duration can be questionable in that case. Plus we use Risk Register in case of event that they may happen and cause delay for project. I used both Monte Carlo and Latin hyper cube algorithms and results were similar. I'm curious how was P50 and P80 of your project are they close?
Rafael Davila
User offline. Last seen 4 hours 21 min ago. Offline
Joined: 1 Mar 2004
Posts: 4763

If all your individual activities durations have a narrow distribution then the difference between P50 an P80 will be narrow and wider if their distribution is wider; but in reality usually it is more complicated than this. Each schedule is different, the relationships are complicated and even calendars and holidays enter into play and the only way to figure out the real difference between P50 and P80 is though Monte Carlo or similar statistical approaches. 

Keep an open mind, it is possible that time distribution on activities with TF < 2 weeks and RD > 2 weeks is too restrictive and some impact is being missed. Also double check on your distributions.

In order to avoid missing unsuspected possible critical paths I usually apply some distribution to all activities, I use formulas [named Global Change in P6] to set my initial distributions something like optimistic duration = 0.70 deterministic duration, most probable = deterministic duration and pessimistic = 1.5 deterministic duration, yes a non uniform distribution has been my experience. Then I make some adjustments like for example some activities duration like concrete curing I set to fixed duration and no variance while for others I use some judgment. 

Double check your model and stand for it if you feel the model is correct, ask the client to show you where your model is wrong.  If you did it right why insisting on forcing to an unreal result?.