Corporate Risk Assessment
Incorporating Project Specific Risk Considerations into a Corporate Risk Assessment – the Probability of Performance Delivery
While many companies utilize advanced systems for quantifying and normalizing geologic risks, and some have standardized above ground risk characterization, relatively few use this information effectively when assessing risks at a corporate portfolio or aggregate level. This can be attributed to many factors, including the belief that data compilation or data management requirements will be too complex, or lack of perceived value in this information by corporate decision makers.
The methodology outlined in this paper will directly address the issue of complexity by outlining a process for systematically incorporating uncertainties into corporate forecasts as material to performance projections. Aligning corporate needs for stochastic forecast data with the project level assessment does not have to be overly complicated and detailed.
Regarding the issue of perceived value, executives often reject probabilistic forecasts provided as a range as being too vague. They want “a number” they can provide to analysts or to the board. This methodology provides that number, coupled with valuable insight into the probability of achieving it. In the simplest case, this could involve the probability of delivering a portfolio with a specific value. A more typical application would involve multiple simultaneous objectives. This more accurately reflects the way risks are actually balanced at many companies, as executives make trade-off decisions for various operating objectives, cash generation, capital investment, dividends, and debt.
Case Study – National Oil Company
We will start with a short case study to illustrate why companies incorporate these analytic techniques and the benefits. The final section will focus more on the details of how this analysis is conducted and how the required data is generated.
Missed Targets and Reduced Credibility
The executive team at a large, integrated, National Oil Company needed to gain greater insights into the key risks that the organization faced. Falling commodity prices, failures in a few key projects, and timing delays on some large developments had resulted in a need to adjust capital programs in prior years to balance required dividend payments and funding needs. The executive team was concerned that these adjustments damaged their credibility with the board. Going forward, they desired a higher level of confidence that forecasts accurately described the potential.
A New Approach to Corporate Risk Assessment
Although stochastic data was available at the project level, historically the planning process did not fully utilize it. All cases were submitted as expected value outcomes. After evaluating which risks were critical at the corporate level, the planning team compiled simple stochastic descriptions of an appropriate subset of the opportunity inventory. This targeted approach to using simple stochastic descriptions greatly simplified data compilation and management, making stochastic portfolio analysis a viable option.
After the development of a corporate risk assessment based on this data, the executive team could evaluate if the probability of meeting their goals was satisfactory, and refine of the capital allocation to optimize the trade-offs between dividend payments and new investment programs. As contrasted to former allocation plans, the plan that was developed provided a clear assessment of the probabilities (and risks) to the key metrics for the company. This provided insights into timing uncertainties and allowed the company to be more aggressive in their growth plans while maintaining a higher confidence in performance delivery.
Methodology – Best Practices for Translating Project Uncertainties into Corporate Risk Assessments
Since company data could not be used for a public article, an example portfolio model was constructed to illustrate the methodology. The simple analytic structure described in this paper may be scaled to represent much more complex uncertainty environments and more diverse assets as required. The value in properly capturing project specific uncertainties and maintaining these descriptions when assessing portfolio value and risks is quantified for the examples depicted.
The process for translating project uncertainties into a corporate risk assessment consists of 4 steps.
- Identify the key corporate risks and their related metrics
- Identify key project level uncertainties
- Develop integrated project level scenarios using simple stochastic descriptions
- Evaluate the probability of meeting goals at the portfolio level
Identify Key Corporate Risks
Identifying which uncertainties are critical to the corporation in advance of data compilation can assist project teams in proper assessment of critical uncertainties – without burdening the organization with superfluous data requests.
Key corporate risks are normally related to performance deliverables that have a direct bearing on the value or reputation of the company. For example, a company may determine that the key risks are related to not meeting objectives for free cash flow, EBITDA growth, and/or reserve replacement.
Once these critical corporate metrics are determined it is relatively easy to map these back to the required project source data. For instance, if free cash flow is deemed critical, then project level data needed include production forecasts and the related capital and operating costs, including country-specific fiscal terms.
The executive team should be able to define the tolerance ranges around the key deliverables, such that the analyst may then gauge the materiality of variance as portfolios are aggregated. As the company performance objectives may evolve, assessment of the tolerances and performance variances should take place periodically to ensure continued alignment.
Identify Key Project Uncertainties (Drivers to Value)
Assessment of the major drivers to value utilizing tornado diagrams can be an effective way to gain insight into the uncertainty information needed at the corporate level relating to a specific project.
The tornado diagram (Figure 1) is generated by running a Monte Carlo simulation on a project economic model, sampling across key variables. This provides a quick way for analysts to assess which metrics are most relevant to project value, allowing for focus on these primary drivers. In this example, Initial Production (IP) – Phase 1 is the most significant variable, with an impact approaching an order of magnitude greater than the next most significant variable of oil price. Note that in most cases price is correlated across the portfolio, so it is often better to assess the impact of price as a portfolio level scenario, and not as a variable when analyzing a project. Evaluating the impact of price changes at the portfolio level provides decision makers with a clear assessment of the optimal decisions in a given price environment, allowing for an understanding of the different decisions potentially required under various price scenario assumptions.
Figure 1. Project Uncertainty – Value Drivers. Understanding which metrics have the greatest impact on value allows the analyst to more effectively ‘prune’ the uncertainty tree.
Develop Integrated Project Level Scenarios Using Simple Stochastic Descriptions
Fortunately, robust uncertainty consideration and normalization processes of below ground risks are well established at many companies, as developed and refined through the application and well documented in the literature [Rose, 2001], [Lerche, MacKay, and MacKay, 1999].
Stochastic techniques have been widely applied in E&P project assessments. Monte Carlo simulation of project uncertainties can provide analysts and planners with a comprehensive view of potential project performance.
This paper does not seek to dispute the value in full stochastic assessments at a project level but does call into question the value of this level of detail when corporate or aggregate portfolio allocations are considered.
The data management overhead and computational requirements of maintaining fully stochastic descriptions at the portfolio level have led to many organizations to seek out more simplified approaches to corporate level portfolio uncertainty review.
Einstein’s famous quote that “Everything should be made as simple as possible, but not simpler” directly applies to the development of project level integrated scenarios. Building out integrated descriptions of a project as based on a few key variables, can be an effective way to capture the range of potential for an asset. In the example depicted in Figure 1, the full uncertainty tree can effectively be distilled into the three-node tree focused on the primary variable of initial production (shown in Figure 2). This can be an effective way to consider all the ancillary variables such as operating cost and capital, as each scenario is considered in its entirety.
Figure 2. Integrated Scenarios. Uncertainty tree as developed around the metric predominately driving value.
Teams are usually very effective at developing a consistent case around the primary variable, thus considering all the implications were this condition to occur. Rather than having to independently consider variables such as facility scale and cost, pipeline sizing and cost, and decline profiles, the analyst can develop a realistic case that directly applies to the primary variable (initial production rate). Figure 3 depicts a simple example of the type of time series data used to describe the integrated scenario.
Figure 3. Integrated Scenarios – Time Series Data. An example depicting integrated time series variables for each outcome as considered.
Full Stochastics vs. Simple Stochastics
A common concern when compiling datasets from project level detail into simplified corporate level aggregations is that data is no longer valid. While expected value descriptions are in many cases inadequate, full stochastics can overload data management systems and provide relatively little additional insight. Figure 4 below illustrates the insight gained from using stochastic data vs. expected value. Figure 5 illustrates that moving from full stochastics to simple stochastics changes the numbers slightly, but does not change the ability to gain valuable insights. We have found this to be the case in numerous client engagements. In situations where there is a concern, it is relatively easy to test the sensitivity by using a limited data set to compare the results using full stochastics vs. simple stochastics against corporate level measures. This can provide confidence that simplifying assumptions will not alter the decision space.
Figure 4 provides an example of the insight that can be gained from including full stochastics in project descriptions, focusing on a single metric (oil production) as derived from multiple input variables (initial rate, decline rate, hyperbolic exponent). The value of a stochastic perspective may clearly be seen in this comparison, as the expected value assessment could be seen to imply a 100% chance of meeting the objectives in 2017 through 2020. When 1000 trials are applied for each case in the portfolio model, the probability of meeting the objective is indicated to be 52% in 2017, 74% in 2018, 80% in 2019, and 66% in 2020. If this is a critical corporate metric to be used in publicly announced earnings projection or production target, most CEO’s would not be overly comfortable reporting what is essentially a 50/50 chance of occurrence as guaranteed performance delivery.
Figure 4. Expected Value vs Stochastic Descriptions. The target (production growth) is depicted by the shaded area. The expected value of the portfolio is represented by the solid line. The diamonds (and right axis) reflect the probability that the portfolio delivers the target. The chart on the left utilizes expected value data only. The chart on the right was developed using a Monte Carlo simulation with 1000 trials.
Figure 5. Simple Stochastic Descriptions. The impact of simplifying project descriptions on portfolio performance delivery (as represented by the probability of achieving a production growth objective).
Running full stochastics and maintaining datasets with each project represented with 1000 Monte Carlo samples while potentially more accurate, would be both impractical and infeasible at most companies. Figure 5 shows the impacts of reducing the project description detail from the fully stochastic case in Figure 4 to simple stochastics using a 3-node tree focusing on an integrated scenario using a single variable. While the probabilities vary slightly, all are within a range that would yield adequate insight and relatively consistent confidence levels for each of these cases. The underlying uncertainty in the source forecasts (as measured by historical trends in forecast accuracy) would indicate a much larger difference than any of the probability differences indicated between the full stochastic and simple stochastic 3-node tree. Testing of the 3-node tree by biasing towards the high and low outcomes might further prove that this representation is totally adequate for an effective portfolio assessment.
Evaluate the Probability of Performance Delivery at the Portfolio Level
Once each project is described in terms of the relevant metrics (as shown in Figure 6) these may then be summed into the aggregate portfolio view. A Monte Carlo simulation may then be run on the entire portfolio selection. In this case, each trial would sample from either the High, Most Likely, or Low outcome metric as biased by the weight factor applied. The resulting statistical distribution data may then be used to assess the probability of meeting specific portfolio performance objectives. The expected value of the portfolio performance is the sum of the risk-weighted individual metrics (in the case of Figure 6, 30% of the High metrics + 40% of the Most Likely metrics + 30% of the Low metrics).
Figure 6. Probability of Performance Delivery. Depicts chance of achieving specific performance measures. This allows multiple metrics to be assessed at the same time in terms of the probability that this portfolio will deliver the desired performance. The dots and right axis represent the probability that the portfolio expected value (red line) delivers the performance objective (shaded area).
Assessment of multiple metrics for different portfolio allocations can provide decision makers with clear guidance as to the performance trade-offs with given decisions.
Detailed project assessments are often distilled into expected value representations when combined at a corporate level, losing essential information. In this case study, we have demonstrated an approach to assessing value and risks at a portfolio level that incorporates the specific and potentially unique risk characterizations of the underlying components through the use of simple stochastic project level descriptions.
This approach provides decision makers with tangible and practical guidance as to the real impact of specific decisions on objectives. It has proven to be a pragmatic and effective method of assessing corporate risk via a direct assessment of the chance of achieving various performance measures.
This basic methodology may be employed and expanded to a wide range of asset types and diverse portfolio sets, independent of company size or complexity.