Leading Aluminum Manufacturer Strengthens Ability to Predict Losses for Risk Scenarios through Monte Carlo Simulations

The Client: Leading Aluminum Manufacturer

Overview

The client considers successful Enterprise Risk Management (ERM) a source of competitive advantage. Under the ISO 31000 standard, the company has implemented a five-step ERM program that involves identifying, evaluating, and rating multiple business unit risks; quantitatively assessing the risks based on their financial impact -- minimum, maximum, and most likely; running a Monte Carlo simulation to determine the probability of occurrence and value of each risk; conducting both qualitative and quantitative risk assessments; performing another round of Monte Carlo simulations to thoroughly understand the risks.

A small ERM team manages these efforts across thousands of employees. To enable and support the team, the client implemented MetricStream ERM Solution. The solution, which is widely used by some of the largest energy companies, has helped the team standardize enterprise-wide risk data, automate risk workflows, improve risk reporting, and facilitate collaboration and information-sharing.

Integrated with the MetricStream solution is the MetricStream analytics engine which enables the client to improve the quality and accuracy of quantitative risk assessments. The demand for these assessments grew out of the need to predict and mitigate the risk of losses arising from events such as power outages, disruptions in the supply of strategic materials, market price volatilities, property damages, and business interruptions. These are only a handful of the risks and loss events that are likely to occur, and adversely impact profits and business performance. The client knew that if they were able to accurately forecast these events in terms of Value at Risk (VaR), they would be able to proactively take measures to reduce the associated losses.

As part of these quantitative risk assessments, the client performs Monte Carlo simulations to quantify the risks in a common framework so that risk managers, as well as other business managers and, ultimately, the board of directors can study this data, and act accordingly. These Monte Carlo simulations are enabled by the MetricStream analytics engine. Built on the powerful and open-source R software environment, the analytics engine is a leading statistical modeling tool that enables organizations to model and visualize risk measures such as Monte Carlo VaR, Conditional VaR, Capital Adequacy Ratios (CARs), and loss distributions.

Through the system, the client can automatically run data-based or scenario-based Monte Carlo simulations to calculate the VaR at various confidence levels. The system provides the flexibility to either integrate existing empirical loss data, or feed in ad hoc risk scenarios based on which the client can plot a range of probability distributions, including Normal, Lognormal, Triangular, Pert, Discrete, or Uniform. These graphs provide a real-time quantitative model of the company’s risk portfolio, enabling stakeholders and decision-makers to better predict loss events.

The Solution

Although the client already had an existing system to aggregate and manage loss data, they wanted an end-to-end, integrated solution for assessing enterprise risks as well as calculating VaR. MetricStream was best suited to meet this need as its ERM Solution was seamlessly integrated with its analytics engine. The end result was a single, cohesive system for managing the full range of risks as well as loss event calculations.

Through the MetricStream analytics engine, the client can perform two broad types of Monte Carlo simulations:

• Data-Based Simulations
  This method is used when the client wants to leverage empirical data stored in existing loss management systems to predict future losses. Data-based simulations are usually run once every quarter. Users upload the empirical data via a spreadsheet, along a specific risk from the risk library in the solution. The system then runs a Monte Carlo simulation to provide the graphical output and VaR at various confidence levels (e.g. 99%). The data can be viewed in either of the following probability distributions:

  Normal / Bell Curve: In this graph, the user simply defines the mean/ expected value and a standard deviation to describe the variation about the mean. Values in the middle near the mean are most likely to occur.

  Lognormal: In this graph, the values are positively skewed, and not symmetric as in a normal distribution. The graph is used to represent values that don’t fall below zero but have unlimited positive potential.

• Scenario-Based Simulations
  This method is used when the client doesn’t have empirical data for simulation but still wants to, on an ad hoc basis, determine the VaR for specific risk scenarios. In these simulations, users select a risk from the available risk library, and identify the scenario they need analyzed, along with the desired confidence level (e.g. 99%). The only other input they need to provide for most of the probability distributions are the minimum, maximum, and most likely values for each scenario. 
  
  The system then runs the Monte Carlo simulation. The data can be viewed in one of the following probability distributions:

  Triangular: In this graph, the user defines the minimum, most likely, and maximum values. Values around the most likely area are more likely to occur.

  PERT: In this graph, the user defines the minimum, most likely, and maximum values, just as in the triangular distribution. Values around the most likely area are more likely to occur. However values between the most likely and extremes are more likely to occur than in the triangular distribution i.e. the extremes are not as emphasized.

  Discrete: In this graph, the user defines specific values that may occur, along with the likelihood of occurrence for each.

  Uniform: In this graph, all values have an equal chance of occurring. The user simply defines the minimum and maximum values.

  The analytics engine enables the client to view the VaR at 95% or 99% confidence levels. Therefore, if users run a simulation to determine the likelihood of a loss scenario in the next month, the system allows them to determine at a 99% or 95% confidence level that the loss would occur, and would not exceed a specific amount for a particular risk.

Challenges

The client's need for the MetricStream analytics engine grew out of the following challenges:

• Complex risk environment marked by uncertainties and market volatilities
• Need for robust systems to predict and control future losses
• Insufficient top-level visibility into VaR data needed for strategic decision-making
• Growing need to efficiently integrate loss data within the larger context of enterprise risk data

Why MetricStream was Selected?

The client chose MetricStream for the following reasons:

The MetricStream GRC Platform analytics engine provides a range of cutting-edge tools to predict VaR at desired confidence level. 

The system has the flexibility to be configured to meet the client’s specific analytics needs. 

It supports a range of probability distribution models to help stakeholders understand the big picture of loss events. 

The analytics engine is seamlessly integrated into MetricStream GRC Platform which means that it can be used in conjunction with MetricStream Risk Management Solutions to provide a complete context to enterprise risks.

Benefits

• Increased ability to predict losses for various risk scenarios
  By conducting Monte Carlo simulations through the MetricStream analytics engine, the client is able to proactively anticipate and reduce losses arising from various scenarios such as a disruption in the supply of strategic materials, aluminum price volatilities, and casthouse property damages.
• Data-driven strategic decision-making
  The loss data generated by the MetricStream analytics engine allows decision-makers in the company to assess the impact of risk, and examine the complete range of possible outcomes. This, in turn, allows them to make informed and balanced decisions about risk.
• Consistent risk monitoring
  The quarterly data-based risk assessments and scenario-based assessments run through the MetricStream analytics engine enable business and risk managers to stay in touch with the ground reality and progress on risk management programs. These assessments also help them proactively identify areas of concern, and implement controls and other measures to improve protection against losses and risks.
• A single, centralized solution to manage all risk related activities
  The client has a common system to conduct the full range of risk management processes and quantitative risk assessments on MetricStream GRC Platform. This allows the Monte Carlo simulation data to be easily viewed and leveraged within the context of enterprise risk data. The client can also use the same system to plan and monitor loss management and risk mitigation measures.