The ValidRisk tool quantifies systemic risk using a parametric model. Parametric models are typically developed using multi-variable regression analysis. The “root” of the ValidRisk model is research by the Rand Corporation in the 1980s (see the AACE version of the Rand model in RP 43R-08). They studies cost growth, schedule slip and project practices data collected from large process industry companies (being sponsored by the US DoE, the study and results are in the public domain). Later published research on team development, project controls and other systemic risk drivers was melded into the root model while keeping the result consistent with industry outcomes.
Because the model addresses the level of technology and complexity, study has demonstrated that the model is “industry generic”. It applies to any engineering and construction project from buildings, to transit systems, to nuclear power plants. Pre-built definition matrices are provided for most asset types.
Because parametric modeling is new to many, a simplified demonstration model is provided free with the PRQ book. The book documents the demo model and provides instructions for its use. Several companies have trialed the demo model alongside their legacy risk quantification method until confidence has been gained in its efficacy.
Based on the research, the ValidRisk parametric model is based on lognormal distributions (for cost growth and for schedule slip) that are “molded” if you will by the systemic risk inputs (the parameters) to replicate the updated Rand results. Lognormal is used because it is a best (or near best) fit for cost growth and schedule slip data distribution.
The model can be easily calibrated if it is found that the user’s results are different than industry. It can be calibrated for bias (modified mean) or uncertainty or spread (modified standard deviation). The ValidRisk partners can assist with calibration studies of the client’s cost growth and schedule slip data. However, experience has demonstrated the generic nature of the model; one is unlikely to need to apply the tool’s calibration capability.
To apply the model, one simply rates the pre-defined systemic risks. For example, the tool asks for the state or level of development of the project schedule; the user simply selects the state from the choices offered (e.g., schedule is fully integrated, resource loaded CPM, etc.). The ratings are then used to adjust the lognormal distribution in its “engine”.