Strengthening Contingency Estimation: Key Takeaways from AACE’s Risk-Analysis RPs

Successful project delivery necessitates effective cost contingencies and risk assessments. AACE International's Recommended Practices (RPs) provide structured guidance in this area. This article focuses on RPs 40R-08 through 44R-08, which address general principles, range estimation, parametric approaches, examples from the process industry, and expected-value analysis. Additionally, it introduces a hybrid methodology that combines parametric and expected-value approaches. 💡

Throughout the article, we offer expert insights based on North American project experience, highlighting common issues, such as the misapplication of Monte Carlo models, evidence-based techniques, and the distinction between systemic and project-specific risks. ⚠️ We also reference AACE guidelines, ISO-31000, and studies by Hackney, RAND, and Galway to support these points. 📚 Finally, we advocate for more empirical research and encourage readers to share their thoughts on the topic. 💬

• RP 40R-08: Contingency Estimating – General Principles

AACE RP 40R-08 outlines the fundamental principles for any contingency approach. 📊 It also provides an objective standard to evaluate the appropriateness of various contingency estimating methods. 🔍 Essentially, this means that a method must be fit for its intended purpose, integrated into the risk management process, and grounded in both expertise and data. 💡 The key principles include:

• **Empiricism and Expertise:** Utilize historical data 📊 and lessons learned whenever possible. Methods should be rooted in empiricism and informed by experience and competency.

• **Avoiding “Iatrogenic” Risks:** Refrain from creating self-inflicted uncertainty. For instance, applying broad ranges to every line item can often inappropriately narrow the output distribution 🚫.

• **Probabilistic Output:** Provide probabilistic results that aid in decision-making 🤔, such as contingency plans at a selected confidence level.

Methods that violate these principles are not recommended. AACE’s Total Cost Management framework emphasizes risk management that aligns with ISO 31000, which itself does not specify the details of contingency. ISO 31000 simply states that “risk estimation needs to be done,” while RP 40R-08 provides specific guidelines on how to achieve this. A review of AACE’s process found it to be “consistent with ISO 31000” and noted that it is the only framework that explicitly addresses contingency estimation.

• RP 41R-08: Range (Monte Carlo) Estimating

RP 41R-08 outlines a Monte Carlo-based method for estimating project range. Its purpose is to assess the probability of a cost overrun for any given estimate and to determine the necessary contingency to achieve a desired level of confidence. Essentially, the project team identifies a few key elements that contribute to the most uncertainty, assigns realistic low, most-likely, and high-range values to these elements, and then runs a simulation to determine the cost distribution. Here are the crucial points:

1. Prioritize significant risks by ranking them based on their impact on cost 💰 or schedule thresholds ⏰, while assessing non-critical items with a simple estimation. This approach helps avoid the unintended narrowing of outcomes. Research by Michael Curran indicates that project uncertainty is often concentrated in a limited number of factors.

2. Identify all risks and determine which items are "critical." ⚠️ Assess the extreme ranges (worst and best outcomes) of these critical items and select appropriate probability distributions, such as triangular 🔺 or double-triangular, to accurately reflect skewness. Conduct a Monte Carlo simulation 🎲 and determine the contingency needed for the desired confidence level. The risk management process provides guidelines for identifying key items and highlights the importance of obtaining accurate feedback on these extremes.

3. Recommended distributions include triangular or double-triangular probability density functions (PDFs), while lognormal 📊 or beta distributions may also be used for better skew representation. The main objective is to model the asymmetry of severe hazards with minimal complexity.

This method requires a high level of rigour in practice. Common mistakes include not properly identifying critical items or failing to challenge experts regarding realistic worst-case scenarios. It’s important to note that many implementations of “range estimating” in the industry skip these essential steps. As Hollmann warns, many so-called range analyses do not adhere to the RP 41R-08 guidelines and can lead to dangerously narrow contingencies. Always ensure that the technique you are using aligns with the intent of the RP.

• RP 42R-08: Parametric Estimating (Systemic Risks)

RP 42R-08 shifts its focus to parametric estimation for risk assessment. Instead of modelling risks on an event-by-event basis, parametric models estimate contingencies based on project-level parameters, such as scope maturity, technology level, and team competency. This RP is unique because it is rooted in empirical research: John Hackney’s pioneering work has shown that poor scope definition is often the primary driver of cost and schedule risks, with predictable impacts. In AACE’s framework, risks are categorized into two taxonomies:

• Systemic risks are general factors related to a project's overall system, such as new technology, regulations, project complexity, and how an organization operates. These risks can lead to measurable increases in costs. RP 42R-08 provides various examples, such as how clearly processes are defined (including design completeness, complexity, and technology maturity) and factors that define the project (like site conditions, environmental issues, and the robustness of the schedule/process). It also considers the quality of the estimation process, which includes data availability, estimator expertise, and bias. Typically, these systemic factors are identified early in the project (during Class 5–4 estimates) and significantly impact cost uncertainty. Parametric models use regression-based methods (such as Hackney’s or more recent RAND models) to change assessments of these factors into a range of possible cost growth.

• Project-specific risks: Discrete events like rainstorms, subcontractor disputes, or material shortages. Their impacts vary from project to project and can be estimated more directly by the team (see RP 44R-08 below).

There are two steps to using RP 42R-08 🚀: development and application. First, you need to build a model by gathering historical data 📊 and fitting regressions or risk correlations. Then, to apply it, assess your project's systemic elements according to the model's guidelines, coordinate with any specific provisions for project-related risks ⚠️, and calculate an output distribution along with contingency measures.

When advising on North American projects 🇺🇸, one common challenge is objective rating. For instance, it can be sensitive to acknowledge that a team's estimation approach is flawed or that the quality of data is inadequate. However, the advantages of addressing these issues are clear: parametric techniques effectively capture significant uncertainties in the early phases and work well within stage-gate scope-development frameworks 📈. For early Class 5 estimates (when the scope is poorly defined), RP 42R-08 recommends relying on parametric analysis alone.

• RP 43R-08: Parametric Models for Process Industries

RP 43R-08 delivers compelling examples of parametric contingency models specifically tailored for the process industries. It features two powerful Excel tools: the “Hackney” model, grounded in Hackney’s research, and the “RAND” model, created by Merrow et al. from the RAND Corporation. These models provide clear insights into how systemic drivers, such as the completeness of scope and technology level, directly influence cost growth. Users will benefit significantly from reviewing the original documentation and calibrating the tools with their industry data for optimal results. Overall, RP 43R-08 convincingly illustrates the practicality of empirical, regression-based risk models in enhancing contingency estimating practices.

• RP 44R-08: Expected-Value (Project-Specific Risks)

 RP 44R-08 covers the classic expected-value approach, often called the “standard risk model.” In its simplest form, Expected Value = (Probability of Occurrence) × (Impact if it Occurs). This formula underpins decision-tree analysis and risk ranking.

Key points: 

• **Risk Linkage and Iteration:** The expected-value method establishes a direct connection between individual risk events and their corresponding cost outcomes. Due to its straightforward nature and consistent application, this method is effective for both risk screening and contingency analysis, enabling teams to discern which risks substantially influence overall costs.

• **Project-Specific Risks Only:** It is advisable to utilize this method exclusively for the project-specific risk category as defined in RP 42R-08. Examples of such risks encompass weather-related delays, subsurface surprises, challenges related to material delivery, constructability issues, resource shortages, and the need for rework. These events can be accurately estimated by the project team regarding their specific impacts on designated Work Breakdown Structure (WBS) items.

• **Timing:** The application of the expected-value method is most appropriate when detailed risk information is available, typically at Class 4 estimates or higher. In the case of very early estimates (Class 5), systemic uncertainties are likely to prevail, thus making the expected-value method unsuitable as a standalone approach. Nonetheless, at later project stages, RP 42R-08 explicitly endorses the integration of the expected-value method with parametric analysis (refer to the hybrid approach below).

• **Monte Carlo Usage:** The expected-value analysis employs Monte Carlo simulation to aggregate the outcomes of various risks. The project team assigns probability distributions to both the likelihood and impact of each risk, considers relevant correlations, and subsequently conducts the simulation. Similar to range estimating, the challenges lie in defining realistic extremes and selecting appropriate probability density functions (PDFs). Notably, the expected-value method prioritizes risks based on their contribution to contingency funds rather than merely identifying "critical items." This focus facilitates targeted mitigation efforts on high-impact risk factors.

Other Methods:

• **Hybrid Parametric and Expected-Value Approach**: In practice, many firms (including those in our consultancy experience) combine parametric and expected-value methods. This hybrid approach is effective because both methods provide a cost-impact distribution for their risks ⚖️. The process involves including the entire output of the parametric model as a single risk in the expected-value (EV) analysis, covering all systemic factors. Each project-specific risk is then quantified in the EV model as usual. Since systemic and project-specific risks are generally independent, a Monte Carlo simulation can be run on this combined model 🎲. The result is a single probability distribution that encompasses both types of risks. As noted by RP 42R-08, while parametric models are adequate for very early estimates, later-phase Class 4 estimates require the use of additional methods alongside parametric analysis 📊. The hybrid approach prevents double-counting; each risk is categorized as either systemic or project-specific and managed in the appropriate context.

• **Empirical Models and Risk Taxonomy**: The RP series emphasizes that cost risk models should be based on evidence 🔍. Parametric risk parameters (RPs) (42–43) are grounded in decades of research (Hackney, RAND, Merrow), demonstrating that systemic risks lead to statistically predictable cost growth 📈. For instance, projects with only concept-level scope (Class 5) often experience significant cost increases, which Hackney quantified using regression equations. In contrast, project-specific events are inherently random and are best addressed through event-by-event analysis (EV) 🎯. Distinguishing between systemic and project-specific risks is essential, as it informs when to apply empirical parametric models and when to utilize expected-value simulations.

References:

1. AACE International's Recommended Practices (RPs)—RPs 40R-08 through 44R-08:

2. AACE RP 40R-08 – Contingency Estimating – General Principles

3. AACE RP 41R-08 – Risk Analysis and Contingency Determination Using Range Estimating

4. AACE RP 42R-08 – Risk Analysis and Contingency Determination Using Expected Value

5. AACE RP 43R-08 – Risk Analysis and Contingency Determination Using Monte Carlo Simulation

6. RP 44R-08 – Integrated Cost and Schedule Risk Analysis Using Monte Carlo Simulation

7. Hackney (1997) Emphasized the importance of distinguishing between systemic and project-specific risks and recommended practical tools like range estimating to capture realistic outcomes in project planning.

8. Merrow et al. (RAND, 1981): In their foundational research on capital projects, they found that project cost and schedule overruns were primarily due to a small number of underestimated critical risks, supporting the prioritization approach used in the Monte Carlo simulation.

9. Galway (RAND, 2004): Advocated for better modelling of skewed distributions and focused simulations to reflect realistic risk environments, corroborating the value of using triangular, lognormal, or beta distributions for risk quantification.

10. ISO 31000 – Risk Management: Provides internationally recognized guidelines on risk assessment principles, including identification, analysis, and evaluation of uncertainties. It supports the integration of structured techniques like Monte Carlo simulation into broader decision-making frameworks.