Abstract
Time and cost outcomes of large and complex projects are forecast poorly across all sectors. Over recent years, Monte Carlo (MC) simulation has increasingly been adopted to forecast project time and cost outcomes more realistically. It is recognised that the simultaneous analysis of time and cost impacts makes sense as a modelling objective, due to the well-known relationship of time and money in projects. But most MC practitioners advocate the use of Schedule Risk Analysis (SRA) feeding into Cost Risk Analysis (CRA) because they believe it is too hard to perform Integrated Cost & Schedule Risk Analysis (IRA) realistically. This chapter elaborates an IRA methodology that produces realistic forecasts without relying on questionable assumptions and enables identification and ranking of all sources of cost uncertainty for risk optimisation as part of the process. It also describes an extension of IRA methodology to include assessment of the assets produced by the project as well as the project itself, thus enabling the analysis of business risks as well as project risks.
TopBackground
The author became involved in Schedule Risk Analysis in the early 2000s, seeing it as a different window into project planning and enabling assessment of the likelihood of the project finishing by the planned finish date. As he gained experience, he saw the need for integrating risk events into schedule risk modelling and led a group that developed a risk management database to identify, manage and map risk events into the SRA model. Once risk events were being incorporated, cost impacts of some risks were recognised as requiring modelling as well as schedule impacts. In late 2006 it became clear that the only way to deal effectively with both cost and time impacts was to analyse cost and time uncertainties and risk events in the one model. This followed an engagement to conduct separate SRA and CRA analyses for a $3 billion mining project. No satisfactory solution to the problem of integrating the SRA analysis with the CRA analysis was found and certain major sovereign risks were excluded from analysis.
By 2008, the author’s group had developed a workable methodology and supporting risk management database software to work with the then leading MC SRA tool Pertmaster Risk Expert to offer IRA consulting services to clients.
In the same period, Pertmaster was bought by Primavera Systems, Inc. (PSI) and was renamed as Primavera Risk Analysis (PRA). In 2009, Oracle Corporation bought PSI, but was not interested in further developing PRA. PRA has remained much as it was in 2008, still one of the best MC simulation tools with superior planning and graphics capabilities, but increasingly competing against some more modern products with better simulation speed.
Key Terms in this Chapter
Joint Confidence Limits: NASA methodology (since 2009) for assessing the risk-based cost and duration of a proposed program using IRA. Requires that funding approval applications are based on 70% probable cost and schedule forecasts, including risk events.
Probabilistic Weather Modelling: Method of applying seasonally variable weather risk to projects separate from the activities. Usually involves generating random patterns of non-work periods during defined time periods (usually calendar months) that conform to probability distributions based on long term weather records for each month using thresholds for stopping work due to weather advised by the stakeholders.
Risk Factors (Risk Drivers): Underlying causes of uncertainty in a quantitative risk analysis model that act on groups of tasks to change their durations and/or costs with probability that may be equal to or less than 100%. May cause performance better or worse than planned or estimated. Examples include Quantity Uncertainty, Market Conditions and Productivity factors. Their application to overlapping groups of tasks automatically creates correlation groupings.
Probabilistic Net Present Value: Based on the outputs of an IRRA and using P-value Probabilistic Cashflow curves to calculate the Net Present Value (NPV) at the selected P-value. Along with analogous Probabilistic Internal Rate of Return (IRR), used to determine the probabilistic balance between profitability and loss of a project/asset investment proposal.
IRA: Integrated Cost and Schedule Risk Analysis using Monte Carlo Simulation to determine project cost and duration uncertainty simultaneously in a schedule overlaid with the cost estimate and including time and cost impact risk events from the project risk register.
Probabilistic Cash Flow: Time based distribution of cashflow forecasts for a project and its revenue-producing assets produced from the outputs of an IRRA. Usually expressed as a range of P-value cashflow curves encompassing Optimistic (say P10), Likely (say P50) and Conservative (say P90).
Cost Hammocks: Activities in a schedule that do not drive critical path calculations but take up the duration of tasks that they span. Used to calculate time dependent costs applicable to the group of tasks spanned.
SRA2CRA: Serial Schedule Risk Analysis (SRA) to Cost Risk Analysis (CRA) widely used to simulate performing IRA by first running a SRA, then passing a measure of the schedule uncertainty into a subsequent CRA, assuming an average rate of conversion of time to cost. May involve more sophisticated variants including multiple schedule uncertainty allowances for different aspects of the project and transfer of the discrete distributions generated by the SRA into the CRA. Easier than IRA but provides less information than IRA.
IRRA: Integrated Capital and Operating Costs, Schedule and Revenue Risk Analysis. Extension of IRA to include the operating life of the asset produced by the project. Enables modelling of business risks to the project and asset, including revenue uncertainties and risks.
Quantitative Exclusion Analysis: Method of measuring the contribution individual contributors of uncertainty make to an IRA model. Involves removing the individual contributor from the model, re-running the simulation and expressing the uncertainty contribution as the difference between the fully-risked model and the model excluding the contributor, at one or more agreed P-levels. The individual contributor could be a single task, a group of tasks, or complete classes of uncertainty. Could be expressed in probabilistic delay or probabilistic cost. Overcomes the problem of ranking schedule or cost drivers accurately in correlated models.