Ask Your Mates Open Forum

Monte Carlo - Analysis

  • 1.  Monte Carlo - Analysis

    Posted 24 February 2020 17:54
      |   view attached
    Morning All
    Im learning the Monte Carlo analysis currently and using the @risk plug-in feature.
    The question i have is that im using this with a contractors risk register information.
    For example we have a element that may occur during the construction period, we have been provided a cost impact range which i have made as the distribution (RiskPert) and then there is a % chance of this event occurring.
    Then to determine the risk allowance we have a likelihood of the event occurring. This is a RiskBinomial of the %. I then have a RiskMakeInput for the risk allowance.
    I then take the probable cost plus the risk allowance to arrive at the project budget...
    Can anyone please advise if this is the correct process to determine my project budget or if i should be taking a different approach.
    Many thanks, example attached and i have tried to show workings below:

    Description  Total Forecast Cost   Worst    Expected   Best   Probable Cost  Likelihood Simulated Occurrence Risk Occurs (Yes/No) Risk Allowance
    1 H&S Issues $75,000.00  $100,000.00 $75,000.00 $50,000.00 $75,000.00 10% 1 Yes #NAME?

    Attachment(s)

    BlogPageSpacerBlank


  • 2.  RE: Monte Carlo - Analysis

    Posted 02 March 2020 17:15
    Please - Anyone have any feedback around the Monte Carlo?
    Would appreciate any information.

    Cheers
    Sam
    BlogPageSpacerBlank


  • 3.  RE: Monte Carlo - Analysis

    Posted 03 March 2020 21:19
    Hi Sam,

    My understanding is that @risk is a purchasable plugin for excel?
    ​I've never used it, and my assumption would be that the vast majority of people here haven't either.

    My advice:
    • If you want generic help with using risk assessments to calculate project budget - Rephrase your question to remove all references to @risk commands so that someone with no understanding of how the niche program works can understand or,
    •  If you want specific advice on whether you've used the @risk plugin correctly - Contact @risk directly for help or try an @risk specialised forum.
    Regards,
    Ryan
    ​​​
    BlogPageSpacerBlank


  • 4.  RE: Monte Carlo - Analysis

    Posted 03 March 2020 22:38
    Hi Sam,

    I've had limited experience with ​@RISK​ but it's something I'd like to get better at.

    Keep in mind that I am by no means an expert, I can say that yes your approach is pretty well how I've produced project budget estimates before. Typically we develop a bottom-up estimate for the full project, which we use as the base estimate. We then model risks and their impact as costs incurred on top of that base estimate. We take the P50 result of the monte carlo simulation as the budget estimate, and the P90 as the budget estimate including contingency (i.e. the difference between the P50 and P90 is the estimated contingency value).

    We model the risks pretty well in the way you've described; a distribution applied to both the likelihood of occurrence and the cost impact if the risk is realised. We usually use Pert, or sometimes LogNorm distributions for continuous distributions, depending on how long the upper tail of the cost risk is (i.e. could the risk be double the "most likely" value, or a whole order of magnitude bigger). Bernoulli or Binomial are what we would select for discrete distribution. We don't really know enough about probability to deviate from these. The two categories of risk we model are:

    • The inherent risk (the general uncertainty in the unit rates we've used), which have a 100% likelihood, but varying cost impact
    • The contingent or "event based" risks, which may or may not occur, and may even occur multiple times. Examples would be inclement weather, safety incident, resource shortage, etc. I think your selection of the binomial distribution for the likelihood works well, as it could happen multiple times. I'd be wary of a general declining likelihood depending on the risk identified, as it could be that for a 12 month project, the chance of multiple wet weather events is more likely than just one.

    That's about the best info I can offer. I do think Ryan makes a good point and would also suggest you ask advice from the @RISK​ developer, Palisade, for more tailored advice. Their support staff are excellent resources for understanding both their software and probabilistic simulations in general. Rishi was one guy who gave us some training, and he is a genius.

    Good luck!
    Nick
    BlogPageSpacerBlank


  • 5.  RE: Monte Carlo - Analysis

    Posted 04 March 2020 16:50
    Hi Sam van der Leij,
    I think this is a question which calls for an in depth answer, from someone whose competence on the field you can check and where you're likely to get additional comments if crap is being anwered.
    I therefore advise you to adress yourself to specialised fora for reliability specialists. In the Netherlands we have the NVRB with their site www.nvrb.nl, where you might adress yourself to with questions like this.
    I hope this helps.
    Best regards,​

    ------------------------------
    Met vriendelijke groet / With kind regards / Mit freundlichem Gruß / مع أطيب التحيات / Saludos cordiales


    Geert Henk Wijnants
    Principal Consultant

    Stork Asset Management Technology
    Van Deventerlaan 121, 3528 AG Utrecht
    P.O. Box 2776, 3500 GT, Utrecht
    The Netherlands

    Mobile: +31 (0)6 13357246
    Email: geerthenk.wijnants@stork.com
    Website: www.stork.comNetherlands
    ------------------------------

    BlogPageSpacerBlank


  • 6.  RE: Monte Carlo - Analysis

    Posted 29 days ago
    ​Although I have no experience of using this specific add-in, the basis of a Monte Carlo Simulation is that you look at the likelihood of the risk occurring (as a distribution) and the impact (in this case cost or delay) caused by the risk occurring (also as a distribution). What happens then basically is that multiple tests are performed choosing risk and impact choosing values (defined by the distributions) and a new composite profile for the project developed. Just be aware that if you have two binomial type distributions the mean value will almost definitely be mean likelihood and mean impact. Logic also dictates that if there is a 50% chance that something occurs the impact will be half of the mean (assuming that if it doesn't happen it costs nothing). The Monte Carlo approach is generally only worth doing if you have many different risks and properly defined impact profiles, otherwise just use a historic percentage based on projects of the same type.
    BlogPageSpacerBlank


  • 7.  RE: Monte Carlo - Analysis

    Posted 03 March 2020 22:37
    I have now looked at your closely, it runs fine. What is your query. Note that you do need to select the cell you want to iterate - which in your case is the total. Make sure random values are selected in settings. Good luck..
    BlogPageSpacerBlank


  • 8.  RE: Monte Carlo - Analysis

    Posted 27 days ago
    Thank you all for your assistance and information - Greatly appreciated!
    BlogPageSpacerBlank