Unleashing the Power of Predictive Analytics: Transforming Risk-based Decision Making with Data

In our work as expert witnesses and investigators we have noticed that rigorous, quantitative evaluation of security threats and other related risks is almost always absent in feasibility studies and other formal evaluations of the viability of large projects and investments. Insofar as these risks are considered, the risk assessment methodology tends to lack depth and reliability, in comparison to other technical and operational disciplines. This perhaps surprising omission is especially relevant when the threats being assessed have the potential to threaten strategic objectives – after all, serious failure and / or legal dispute is the very reason we are considering these things in the first place.

Flaws in risk assessment are often the pivotal issue in determining liability in legal disputes, and are frequently the root cause in investigations into catastrophic incidents.

In our experience, security risk assessments generally are over-reliant on experience and heuristics. While this approach can be valid in some cases, what tends to happen is that optimism bias (‘the project must proceed, because the opportunity is too good to be missed’), combines with visibility bias (‘this is the information we can see, so it must be the most important’), to produce unreliable answers about risk. In turn, these bad answers get hard baked into the risk culture as the project proceeds. In simple terms, this means that when warning signs start to appear later, they get ignored because they don’t fit the narrative. In Israel this phenomenon has come to be called the “Conception”: whilst ample warning signs preceded the Hamas attack of 7th October 2023, neither the Israeli intelligence services, the Army nor the Government had imagined an attack on that scale, leading to the warnings signs being missed or ignored.

One powerful tool that has emerged as a game changer is predictive analytics. Combining the expertise of risk practitioners with novel approaches to mathematical modelling of risk scenarios resolves much of the wooliness of traditional approaches. A well-designed model using good data can provide explicit focus on the probability of serious incidents over time, and more accurate quantification of their likely consequences (whether these be physical, financial, reputational, or others). 

By leveraging historical data, advanced algorithms, and machine learning techniques, predictive analytics can uncover patterns, make accurate predictions, and optimize decision-making processes. In this article, we discuss the most important practical benefits of this approach, along with a sketch of the technical background.

A real-world example:

A few years ago we acted as the expert witness for the host country, which was in dispute at international arbitration with the operating company of a multi-billion-dollar mining project.

A JV metals mining company planned to build and operate a very large mine in an Asian country where there was a longstanding separatist insurgency. The resource was located in a very remote, sparsely populated and undeveloped area with historically low areas of insurgent activity – however, the insurgency consistently and successfully attacked existing infrastructure in other adjacent areas. Over the last 30 years, the insurgency had gone through periods of high, medium and low intensity. At the time the project’s feasibility studies were written, the insurgency was in a high-intensity phase.

The feasibility studies’ Security chapters did not analyze the background of the insurgency or its likely future trajectory (-ies) in any depth, but instead focused on very basic operational level considerations. 

The operating concept was to export the unrefined mineral product via approximately 700 km of pipeline across desert terrain where the insurgency had relative freedom of movement. The pipeline was not mentioned in the formal security risk assessments – therefore, the project would have proceeded to construction with no assessment of the risk to the pipeline. 

The tribunal’s judges accepted our evidence that the omission of the pipeline from the risk assessment introduced significant risk, and discounted the value of the project by several billion dollars, in favour of our client.

This is an instructive example of how rigorous modelling of the threats could have delivered a range of explicit probabilities and quantified consequences for a range of future risk scenarios. There was ample data from the country itself, and from other countries where threats existed (for example, the Arab Gas Pipeline), to model accurate scenarios for periods of low, medium, and high insurgent activity. Doing this would have delivered a much more reliable insight into risk during the project’s lifecycle, than the purely qualitative assessment that was conducted.

This example is one of many, where we have seen similar flaws in risk assessment methodology directly contributing to serious negative outcomes. As a result of this long-term observation, we believe that predictive analytics can offer a far better alternative.

So, what is predictive analytics?

Predictive analytics is a set of business intelligence (BI) technologies that uncovers relationships and patterns within large volumes of historic data that can be used to predict future behaviour and events. Unlike other BI technologies, predictive analytics is forward-looking, using past events to anticipate the future. Predictive analytics statistical techniques include data modelling, machine learning, AI, deep learning algorithms and data mining. 

Our own approach to using predictive analytics in modelling security risks combines ‘training’ of the model by a human expert in security risk, with the technical expertise of OLSPS, which is an international market-leading data science company.

The Foundation of Predictive Analytics: Data Quality and Diversity:

At the core of predictive analytics lies high-quality and diverse data. To generate reliable insights, organisations must collect and aggregate data from various sources, including open-source databases, your organization’s own databases, local intelligence (if your organization has the capability to source this), and third-party intelligence reports. The availability of diverse data allows for a comprehensive analysis, capturing a wide range of variables that influence desired outcomes. However, the raw data alone is insufficient. Organisations must employ rigorous preprocessing and cleansing techniques to ensure data relevance, accuracy and reliability. By addressing outliers, handling missing values, and rectifying inconsistencies, organisations establish a solid foundation for accurate predictions.

Predictive Power through Advanced Algorithms:

Machine learning algorithms play a vital role in predictive analytics, enabling organisations to derive meaningful insights from data. The choice of model depends on the type of questions that our clients need answering: regression analysis, decision trees, random forests, and neural networks are among the powerful techniques used to build predictive models tailored to specific needs. These models learn from historical data and continuously refine their predictions as new data becomes available. 

Optimizing Decision-Making: Insights for Strategic Advantage:

Predictive analytics unlocks valuable insights that drive informed decision-making. By utilizing the power of accurate predictions, organisations can optimize operations, enhance resource allocation, and identify risks and opportunities. For example, in the finance sector, predictive analytics enables early fraud detection and precise creditworthiness assessments. In healthcare, it aids in early disease detection and the development of personalized treatment plans. In strategic security risk, it identifies trends and aids accurate quantification of the probability and consequences of specific scenarios. By aligning strategies with accurate predictions, organisations gain a competitive advantage and proactively respond to market changes.

Continuous Improvement and Adaptation:

Predictive analytics is a dynamic and iterative process. Models need regular updates to ensure accuracy and relevance in ever-changing business environments. Organisations must adapt their models to account for emerging trends and evolving risks. 

We recognize that most organizations lack the expertise and resources to implement this type of approach, and we will be delighted to discuss how we and OLSPS can support your own requirements.

Back to Insights

Share this insight