Project Risk Management System as Part of Digital Infrastructure

The integration of risk management systems into the digital infrastructure of an enterprise is becoming a critical factor for the successful implementation of large industrial projects. In the context of global uncertainty and sanction restrictions, traditional approaches to threat assessment demonstrate their inefficiency. During the webinar, Yana Krukhmaleva, Director of Digital Infrastructure Development at Gazprom-Media Holding, analyzes the evolution of risk management in Russia using the example of major infrastructure projects and explains why the industry is inevitably moving towards working with big data.

Cognitive Biases and the Subjectivity of Expert Assessments

Classic qualitative risk analysis often relies on expert opinions, which are inevitably subject to cognitive biases. Drawing on research in behavioral economics, the speaker shows that people tend to think irrationally in conditions of uncertainty. In practice, this leads to a formal approach: experts either fail to dive deep into threat analysis or fall into mental traps, such as the Dunning-Kruger effect or apophenia (finding patterns where none objectively exist).

Furthermore, collective risk assessment gives rise to the Condorcet paradox — a situation where experts' opinions diverge so much that deriving an objective average score is mathematically impossible. To identify the degree of opinion consistency, it is proposed to use Kendall's coefficient of concordance. If consistency is low, the project manager must take responsibility for the final decision, relying on their own competencies rather than averaged scores.

Limitations of the Monte Carlo Method

In an attempt to move away from the subjectivity of qualitative analysis, many companies are shifting to quantitative methods, particularly Monte Carlo simulation. However, the presentation details a key flaw in this approach: the input data for the mathematical model still consists of subjective expert assessments.

• Lack of integrated schedules: For the method to work correctly, a detailed, multi-level network schedule with strict resource and financial linking is required. On large projects, such schedules are rarely maintained in a single system without broken links.
• Ignoring risk correlation: The Monte Carlo method struggles with situations where multiple risks are interconnected. The system cannot automatically account for complex laws of probability theory when summing dependent events.

As a result, complex quantitative models often turn into a tool for reassuring management, painting an ideal picture of meeting deadlines and budgets that has nothing to do with the actual situation on the construction site.

Big Data and Predictive Analytics as the Future of Risk Management

The only way to radically increase the objectivity of event probability assessment is by using Big Data. Creating corporate Data Lakes allows for the accumulation of large-scale statistics on incidents, equipment failures, supply chain disruptions, and external threats.

The key task at this stage is to ensure data relevance and cleanliness, eliminating the human factor as much as possible during data entry. Information collection should occur automatically directly from production facilities and adjacent information systems. Only on the basis of reliable data sets is the effective application of predictive analytics and machine learning algorithms possible for real risk forecasting.

Evaluating Expert Efficiency and Choosing IT Tools

A significant problem in corporate governance remains proving the economic efficiency of risk management departments themselves. The speaker suggests using Pearson's biserial correlation coefficient to compare initial expert forecasts with the actual occurrence of events. If the coefficient is close to zero, the expert group's forecasts have no practical value, and the process requires revision.

The issue of software selection is also raised. Experience with heavy Western systems has shown their vulnerability to sanctions and inflexibility towards user requests. Today, the focus is shifting to Russian IT solutions (for example, the RisGap platform), which offer a more intuitive interface, ease of implementation, and independence from external restrictions.

What you will learn from this webinar:

• Why do classic risk assessment methods fail on large industrial projects?
• How to identify inconsistency in expert assessments using mathematical coefficients?
• What are the hidden limitations of the Monte Carlo method when analyzing project risks?
• How to eliminate the human factor when collecting data for predictive analytics?
• How can the performance of in-house risk managers be evaluated mathematically?