Vous êtes sur la page 1sur 7

Downs, Peter/GLA

From:
Sent:
To:

Downs, Peter/SNG
06 February 2013 15:02
Arquilla, Bob/DEN; Avant, Michael/BOS; Bacon, Denis/PHX; Carver, Dennis/PDX;
Clark, Gordon/GLA; Downs, Peter/SNG; Fraser, Gordon/SNG; Gibson, David/KNV;
Kerr, Jim/GLA; McDowell, Jerry/SPB; Meier, Tim/PDX; Morris, Rhen/PDX; Myers,
Catherine/PDX; Palmer, John/AUS; Smith, Randy/PDX; Stack, David/CLT; Aguilar,
Daniel/PHX; Albertson, Andy/CVO; Amezquita, Gabriel/MXC; Anderson, Sig/PDX;
Angelino, Raymond/WDE; Ansari, Noman/SNG; Lisanti, Anthony/NJO; Baca,
Steve/RON; Bagwell, Gene/SPB; Banaszek, Daryl/RON; Barajas, Edward/PDX;
Barraza, Jose/SFE; Bash, Troy/PDX; Benovengo, Ed/NJO; Billings, Jonathan/DUB;
Bingham, Kevin/PHX; Bouchard, Carl/CIN; Bradaschia, Roberto/SAO; Brady,
Patrick/CIN; Brinster, Kenneth/RON; Brown, Nora/GVN; Buchan, Paul/SHA; Burch,
Daniel/PSD; Burch, Robert Frederick/SNG; Burckhardt, Rainer/KSG; Burton,
Kevin/SPB; Caballero, Jose/MXC; Cabanillas, Edgar/MXC; Cadena, Pedro/MXC; Cai,
Liang/SHA; Carr, Jim/AUS; Castro, Miguel/MXC; Chan, Lina/SNG; Chee, Allan/SNG;
Chen, Kwanghung/SHA; Chou, Wesley/SHA; Clemente, Adolfo/MXC; Coles,
Arthur/NYC; Colton, Justin; Commins, Kevin/RON; Contreras, Luis/MXC; Cook,
Steve/PDX; Cooley, Steven/IND; Cooper, John/SPB; Cordova, Armando/PHX;
Cordova, Raymundo/MXC; Cosgrove, Kevin/CIN; Cristobal, Kristoffer Krane/PTJ;
Cruz, Abel/BUA; Cullison, Jeffrey/RON; Dadi, Jorge/BUA; Daniels, Abhijeet/RON;
Decker, Guy/RNC; Del Angel, Miguel/MXC; Delacruz, Michael/RON; Ding, Lan/SHA;
Donohue, Ronan/DUB; Dowell, Dan/PDX; Drenth, David /CIN; English,
Matthew/RON; Espinoza, Saul/MXC; Evans, Dan/RNC; Fagel, Winston Bartolome
Mique/SNG; Farrell, John/PDX; Flagella, Bob/PDX; Flores, Alberto/MXC; Forbeck,
Donald/PGH; Francis Xavier, Lincoln/SNG; Frazier, Scott/PDX; Frech, Steve/ATL;
Fulbright, Steven/RON; Fullam, Paul/DUB; Galtieri, Nestor/BUA; Gavel, Greg/PHX;
Gil, Hector/BUA; Gil, Roman/KRA; Gilyard, Frank/RNC; Gimenez, Ricardo/BUA;
Glaese, Matthias/KSG; Godfrey, James/PDX; Grant, Doug/PDX; Gray, Michael/SNG;
Gregoratti, Angel/SFE; Grignola, Gustavo/SFE; Guadarrama, Ernesto/MXC; Guinea,
Elisa/MXC; Gutierrez, Alejandro/MXC; Gutierrez, Mario/MXC; Hailey, Jim/RON;
Hamburg, Mark /CIN; Hansen, Merrill/LHI; Harry, Heath/PGH; Haselton, David/PGH;
He, ShuGuang/SHA; Heberling, Dale/PDX; Heng, See Meng/SNG; Hering,
Jorge/BUA; Hollo, Joseph/PGH; Howard, Bill/SPB; Hubbell, Sid/PDX; Huddleston,
Dennis/SYR; Huerta, Filiberto/MXC; Huerta, J. Manuel/MXC; Hyland, Kenneth/RNC;
Jackson, Vernon/SNG; Jian, Feng/SHA; Kaferle, Jack/NJO; Karles, Alejandro/BUA;
Kenny, Pierce/PDX; Kibbee, Keith/PDX; Kindermann, Kurt/PSD; Kirkendall,
Robert/PDX; Koh, Meng Kwang/SNG; Kolari, Cynthia/BUA; Kubisiak, Michael/RON;
Lagamba, Deborah/PGH; Ledesma, Ruben/SFE; Lee, Lynette/SNG; Lee,
Malcolm/SNG; Lee, Sanders/SPB; Lee, William/RON; Lenz, Dick /CIN; Lerin,
Susana/BUA; Lewis, Jeffrey/CIN; Li, Annie/SHA; Little, Patrick/RIC; Liu, Jasson/SHA;
Liu, Xianhu/SHA; Loeb, Jeffrey/PDX; Lopez, Rodolfo/MXC; Luevano, Marc/RON; Lv,
JiJin/SHA; Macaskill, Ian/GLA; Main, Kenneth/KNV; Malmros, Stephen/RNC;
Manjarrez, Edson/MXC; Marini, Andre/SAO; Marmolejo, Jorge/MXC; Matsukado,
Bill/PHX; Mays, Gary/PDX; McAlexander, Lyxa/PHX; McCabe, Denise/PHX;
McCormick, Phil/PDX; McGarity, Tim/CIN; McGarry, Neil/PHX; McGrane, Terry/DUB;
McInally, David/GLA; McIntire, Daniel/RON; McIntire, Frank/SPB; McRoberts,
Peter/GLA; Mendizabal, Francisco/MXC; Mendoza, Romulo Cruz/SNG; Mezhov,
Grigory/MOS; Miller, Marion/SPB; Mitchell, Larry W/MKE; Mogannam, John/SJC;
Monteferrante, Ron/PDX; Moorman, Jeff/CIN; Morris, Rhen/PDX; Moussan,
Bassam/SHA; Murphy, Tom/CIN; Murphy, Vincent/PDX; Mydlo, Andrzej/KRA; Win,
Aung Myint/SNG; Nadisin, Bala Krishnan/SNG; Newlands, David/SNG; Nichols,
John/PGH; Nickel, Patrick/PSD; Nogales, Gaston/CIN; Nolen, Mark/RON; Novak,
Jeffrey/PSD; Nuernberg, John/CIN; O'Halloran, Michael/PDX; O'Kelly, Rory/DUB;
1

To:

Subject:

Oliveras, Max/SJN; Olson, Eric/SHV; Orozco, Gustavo/MXC; Ortiz, Isidro/MXC;


Parsons, James/SPB; Patterson, Steven/LAC; Pawlak, Tim/PHX; Peloquin, Larry/SYR;
Perez, Yvette/CIN; Perretta, Miguel/SFE; Petarnella, Chrystiane/SAO; Phubundith,
Susidda/BNK; Plaag, Hal/KNV; Powder, Mark/PGH; Powell, Mark/CIN; Pribish,
Buddy/PGH; Profitt, Del/PDX; Punch, Brandon/SPB; Querijero, Apolinar/SNG;
Quisenberry, Jeff/SPB; Raketich, Jacob/PGH; Ramirez, Rita/MXC; Reidy, Bill/SHA;
Reinert, Mark/PHX; Reitmeier, Hal/PDX; Robles, Gerardo/MXC; Robles, Jesus/MXC;
Rojo, Alfonso/MXC; Ruiz, Oscar/MXC; Ryan, Sonya/CIN; Sanchez, Bulmaro/MXC;
Santos, Juan/MXC; Satibi, Roslan/SNG; Savage, Kenneth/SNG; Savage, Robert/SNG;
Schaefer, John/SHA; Schmidt, Robert/SEA; Schuetz, Paul/PGH; Scott, Ron/KNV;
Scribner, Shane/RON; Sears, Teresa/RNC; Serafin, Grzegorz/KRA; Shastry,
Vasudeva/SNG; Shidemantle, David/PGH; Sierra, Teofilo/MXC; Sing Jr,
Angelino/SNG; Sloan, Martin/DUB; Smith, Clark/SPB; Smith, Kent/SPB; Smith,
Troy/SNG; Solberg, Andy/RNO; Soriano, Eric/PTJ; Staley, Ernie/PDX; Stamm,
Dennis/ATL; Strange, Mike/DFW; Strickland, John/PDX; Strohlein, Brett/PDX; Sujarit,
Supachart/BNK; Sunshine, Steve/KRA; Sutton, Sam/PGH; Swanson, Kurt /IDA;
Switzer, John/RON; Tao, ChongJun/SHA; Teh, Hong Tiong/SNG; Terrada,
Antonio/SJN; Therrien, Harry/RON; Thomas, Mike/PGH; Thomson, Allan/GLA; Tian,
Hua/SHA; Tomlin, Joel/PDX; Torioni, Jose/BUA; Tourn, Jorge/SFE; Trigiani,
Alberto/BUA; Tuckfield, Richard/RON; Tyksinski, Tom/DEN; Veerachamy, Baskara
Venkatesh/SNG; Vega, Francisco/MXC; Velazquez, Irma/MXC; Velazquez,
Mauro/SFE; Verdeja, Ruben/BNA; Vidal, Javier/SFE; Voegels, Alfred/PDX; Voloshen,
Michael/PGH; Vonder, Paul/NJO; Walker, Michael/TCA; Walsh, Michael/PDX; Wang,
JianHui/SHA; Wang, Michael/TAI; Wang, Wei/TAI; Warwick, Trent/RON; Watson,
David/COS; Webb, Chris/RON; Weir, Robert/SPB; West, Evelyn/PDX; Wheeler,
Terry/PDX; White, Reed/RON; Wilhoyte, Jerry/SPB; Witosky, Robert/PHX; Wolak,
Grzegorz/KRA; Xia, YueHua/SHA; Xu, Graham/SHA; Yancey, Ginger/RIC; Yang,
May/SHA; Yap, Hon Khiaw/SNG; Yelton, Richard/LSG; Yip, Kwok Ka/SNG; Youngbar,
Adam/PDX; Zhang, Shuo/SHA; Zhao, XinYu Jeff/SHA; Zuchara, Dorota/KRA; Frescia,
Fabio/THT; Allen, Paul/SPB; Clark, Gordon/GLA; Kerr, Jim/GLA; Rodriguez,
Humberto/MXC; Scott, Robbie/SPB
IAT Monthly Risk Topic: Taking the mystery out of Monte Carlo

Follow Up Flag:
Flag Status:

Follow up
Flagged

Dearall,
ThisistheFebruary2013IATRiskTopic,focusedaroundanexcitingupgradetoouriRAMtoolwhichwillbereleased
laterthisweektheadditionofMonteCarloSimulationtothealreadyricharrayofriskcaptureandanalysis
possibilitiesofferedbythetool.Pleaseforwardthistoothersthatyoufeelmightbenefit.Addressanyquestions
regardingthismaterialoranyotherriskrelatedissuestoyourBusinessGroupRiskManagers.Ideasforfuturerisk
topicsarewelcome.

Bestregards,

I&ATBGRiskManager(International)PeterDowns
I&ATBGRiskmanager(NorthAmerica)TerryWheeler
I&ATMonteCarloGuruJoelTomlin
Monte Carlo Simulation
Is it too much of a generalization to suggest that if you mutter the words Monte Carlo in the context of
ascertaining risk on a project, the target audience is significantly influenced to react positively to that
ascertainment ? Maybe, but it probably matters how many statisticians and theoretical mathematicians are
in your target audience.
2

This in mind we have been somewhat reluctant to bolt-on Monte Carlo [MC] to our iRAM tool without, first
and foremost, a thorough understanding of the theory and then ensuring that we have addressed the real
test: what do the results of the simulation actually mean to me and my project ?.
A brief history of Monte Carlo
It is considered that the MC method was named as a result of the gambling preferences of certain nuclear
scientists working on fission predictions at the end of World War 2. During this time, there were several
names credited with the resurrection of this statistical technique, but it was Stanislaw Ulam that linked
theory with the availability (from wartime efforts) of the first electronic computer, the ENIAC at the
University of Pennsylvania. As at 1947 statistical techniques which had previously fallen into desuetude
due to the length and tediousness of the calculations were now able to be resuscitated. Work continued for
many years in this nuclear arena, the domain of super-computers and a handful of privileged scientists and
mathematicians.
Monte Carlo today
Today even the commonly used version of Microsofts Excel has limited MC capability as a function of inbuilt random number generation [RNG] capability. The Randomness of the random numbers generated to
perform MC simulations has in itself become a continuing debate amongst mathematicians. For our use,
we take advantage of the Excel RNG facility in this exciting iRAM 3.0 upgrade. Elsewhere we see MC
simulation in everyday use in the world of economics, finance and business. There was even an online MC
tool that purported to predict the certainty of the recent US Presidential election !
If you search online for Monte Carlo Simulation using the Google search engine, it returns over 8MM
results; many theoretical papers and many software companies selling MC simulation tools. If some of you
are motivated to a better understanding of the theory, the world wide web is your oyster!
iRAMs Monte Carlo Simulation
First some basics. The iRAM MC simulation is only available for a quantitative log. The simulation
develops a statistical profile of the logged risks from the normal iRAM risk assessment to show possible
project outcomes.
The iRAM MC simulation is set to run virtually 1000 scenarios. The probability assigned to each risk
dictates that risks likelihood of occurring in each run. The cost and schedule impact, when the risk occurs
in the simulation, is determined by a random skewed distribution based upon the values entered in the
log. Natural logarithms are used for these calculations; therefore, for cost entries, the maximum, most
likely and minimum impacts must be 3 different values, an error message will appear if values are the
same or are blank.
The run-time of the Monte Carlo simulation is approximately 3 to 4 minutes. Time will vary depending on
the number of items in the risk assessment. Output from the simulation is displayed on the summary tab
and the graphics tab. It is important to note that this simulation is not updated again unless you re-run the
simulation by pressing the Monte Carlo button in the iRAM maintenance drop-down box.
There are three graphs generated from the Monte Carlo simulation:
1. Probability Plot
The probability plot below displays the profile of the results of the 1000 virtual projects. The y --axis shows
percentage and the x-axis shows cost or schedule. The curve represents percentage of runs yielding
under a cost amount or duration. For instance the curve above shows that 20% of the MC simulations
yielded 73MM (the minimum on the graph will always be the project value entered on the Project
Info tab). Similarly, the graph shows that 50% of the runs yielded -less than 73.428MM, and 100%
yielded less than ~84MM.

2. Distribution Plot
The distribution plot below shows the results of the simulations (total project impact) in a histogram
format. The plot shows the percentage of results between two values. . For instance, the chart below
shows that ~19% runs yielded between 73.000MM and 73.432MM. Such a skewed distribution would be a
result of the majority of risks in the iRAM contextualized with very narrow margins between minimum
most likely maximum values.

3. Tornado Chart
The tornado chart shows the 7 highest impact risks to cost and schedule. The y-axis shows the risk item
number, and the x-axis shows the impact magnitude. Since the impact of each risk differs in each of the
1000 simulations, the magnitude is shown in 3 ways: The average impact of the individual risk in the 1000
4

runs, the 90th percentile impact (90% of the runs fall below a value, the 50% percentile impact (50% of the
runs fall below a value).
Generally, the 50% bar will not appear for risks with less than 50% probability of occurrence. The curves
are sorted by average impact. In the iRAM tool, risks are not assigned potential positive impacts;
therefore, the x-axis will never be negative. The graph below shows that risk 9 has a ~160K average
impact in the 1000 runs.

Interpretation of iRAMs Monte Carlo results


The MC simulation will not yield an exact project impact. Instead, the analysis is to be used to understand
the distribution and range of impacts. For example, the distribution plot above shows a tight grouping of
results below 75MM whereas the plot below has a similar average impact of 74.2MM, but a very different
distribution. Such a normal distribution would be a result of the majority of risks in the iRAM contextualized
with wide margins between minimum most likely maximum values. The total risk and contingency
assumed on the project below would likely be higher than the project above. This is reflected in the
statistical distribution, the curve above has a 50 percentile distribution of $73.4MM whereas the plot below
has a 50 percentile distribution of $74.1MM, nearly $700K higher.
Many papers have been written on the interpretation of MC simulation as applied to risk logs. In general,
many tie project funding, either base or contingency, to the 50th or 90th percentile of the MC
results. However, the MC results have to be interpreted within the context of a number of factors: the
phase of the project, the maturity of understanding of the project scope, the certainty of the risk
contextualization, the overall quality of the risk log, and the like. The MC results should be used within the
framework of all of these factors to understand the probability of risks and focus to better characterize the
highest contribution to magnitude and variability.

Monte Carlo Summary

The MC summary table included in iRAM, as shown above, presents a possible treatment of
cost and time risks applicable to the project budget and schedule. In this summary the project
is recommended to be funded to handle 50th percentile of impact value (the value equal or less
than 50% of the simulated runs). Since this impact is likely, the baseline project funding should
include it. Some element of the difference between the 50th
and 90th percentile is recommended to be added to the project
as contingency.
Opinions vary as to how much of this to add but as a rule we
would advocate no less than 50% as shown here by the data
point labeled bid value. Clearly, the contingency element
needs to be identified in the project finances and proactively
managed in project execution. This is a subject in its own right
and certain forms of contract such as the UKs NEC3 suite,
prescribe such methods. Typically, if risks are appropriately
managed and mitigated during project execution, we would
expect to see the amount of contingency reduce of the life of
the project whilst at the same time, some of the contingency
would be converted to actual project cost.

In conclusion
The validity of the MC analysis depends wholly on the validity of the risk data in the log. This, as well as
the factors outlined above, must be taken into consideration before defining project contingency and when
managing project contingency in execution.

Vous aimerez peut-être aussi