When Schedule Risk Analysis (SRA) based on the Monte Carlo (MC) Method of randomness simulation was first developed, the tools available imposed practical limits on the size of schedule. Schedules were limited to around 100 activities or less. Since then, software and hardware advances have enabled SRA models to become much larger. How big should such models be? And does it matter?

To answer these questions, we need to deal with two effects with opposite implications.

First, we consider a familiar concept in project development, known as “Fast tracking”. This refers to the desire of project owners to get to market as quickly as possible by overlapping the phases of development of projects, such as Design, Procurement and Construction. At a high level, this seems quite feasible, but when detailed planning is performed, the difficulties of carrying this through become more evident. We can see this by taking a simple schedule model:

When this is translated to project schedules, it is known as the Merge Bias Effect (MBE). It means that where there are logic nodes in a project schedule, where several strands of schedule logic come together in a milestone, the probability of achieving that milestone by the planned date is the product of the probability of each logic strand being completed by the planned date.

As more strands of schedule logic are joined into milestones representing completion of portions and phases of work, these logic nodes become points resisting the schedule finishing earlier because each logic strand would have to finish earlier simultaneously, which rapidly becomes less and less likely as more detail is added to the project schedule and more intermediate milestones are created.

If a detailed project schedule represents the way that the project will be executed, summarising it will remove details including intermediate milestones that have to be achieved. This will remove logical resistance points to earlier completion such that a MC simulation model built on a summary schedule is likely to produce a falsely optimistic forecast of the likelihood of finishing the project by the planned date.

Second, producing realistic results from MC simulations with progressively larger schedules is increasingly difficult.

Simulations consist of hundreds or even thousands of iterations of critical path calculations, randomly substituting sampled durations from each task’s duration probability distribution. The software selects the duration for each iteration so that it matches the probability distribution defined for that task. So more samples are taken around the Most Likely values than toward the extremities of the ranges.

An inherent assumption of MC simulation software is that each element in the model is completely independent of every other. This assumption is relatively safe when using a highly summarised model but progressively breaks down as the number of activities increases. So one task representing all earthworks will be relatively independent of another task representing all civil works and a further task for structural steel installation, etc.

But if there are 50 tasks describing in detail earthworks to create, for example, access roads, the ROM Pad and the building of tailings dam walls, some, perhaps all will be related to each other to varying degrees. We tell the software the extent to which tasks are related to each other through duration correlation: 100% represents a perfect positive relationship, 0% represents complete independence and -100% represents a perfect inverse relationship. Normally the relationship is positive and somewhere between weak (say 10-30%), moderate (say 30-60%) and strong (say 60-90%).

If we did not tell the software the extent to which tasks are related, it would choose short durations in combination with long ones and vice-versa, such that they would tend to cancel each other out over the iterations and lead to the calculation of very narrow distributions (e.g., a 2 week range between the P10 (10% probable) project finish date and the P90 date in a 2 year project (less than a 2% range)).

This effect is due to the rising impact of the Central Limit Theorem (CLT). As the number of activities increases in the SRA model and the average task duration decreases, the distribution produced tends to a normal distribution and narrows in its spread. The narrowness of the distribution is called “kurtosis”*.

Countering this to produce realistically spread distributions becomes increasingly difficult as the size of the schedule increases. Strategies for countering the CLT tendency include the application of risk factors and increasingly sophisticated correlation models.

To produce the most realistic SRA modelling requires the use of the largest practical SRA schedule model, preferably based on a Level 3 Master Control Schedule for the project and the application of comprehensive correlated risk factors and carefully developed duration correlation models. Achieving this is much harder than producing summarised schedule models, but the results are much more likely to reveal useful information on what is driving the schedule and how best to manage the drivers to maximise project objectives.

Using Oracle’s Primavera Risk Analysis (PRA) and a fast PC, it is practical to analyse a schedule with up to about 5,000 activities in it. Around 1,000 activities is more practical to handle. Provided true critical paths can be identified and the schedule realistically represents the planned project strategy and scope, this size schedule should produce good and useful results.

There is an Australian way of thinking about kurtosis in this context. In SRA output distributions, we wish to follow the platypus rather than the boxing kangaroos.

RIMPL is an innovative risk management and project controls services business. We provide risk management and analysis consulting services as well as skilled project controls and strategic project advisory services and personnel. RIMPL continues services formerly provided by Crescent PSS Pty Ltd (1996 - 2008) and Hyder Consulting Pty Ltd's Project Management Group (2008 - 2012). Through seven years of software methodology development, the risk management team at RIMPL offers the following services:

• The most sophisticated and realistic Schedule Risk Analysis (SRA) modelling available in Australia

• The only true Integrated Cost & Schedule Risk Analysis (IRA) modelling developed in Australia, to assess project contingencies, profitability and viability

• Licensing of the methodology and software suite to approved clients

• Training in SRA using Oracle's Primavera Risk Analysis (PRA) and optionally also using RIMPL's risk analysis software to enhance the modelling realism of PRA

• Sale of and training in selected RIMPL qualitative and quantitative risk analysis applications

Independent of our risk management services, we also offer development of database applications to client requirements.