blank

Why network/graphs analytics in financial portfolio management?

Assets, markets, economic factors, and companies are full of relationships, influences, correlations and causation, which can change over time and are often hard to identify. There are complex underlying structures and patterns evolving and emerging.


Financial portfolios of investments or loans are exposed to the same complexity, of course. There is a microcosm within a portfolio generated by the influences and relations among the portfolio nodes. It is sometimes said ‘all the world is a graph,’ and consequently the most natural data model is not a table with rows and columns, but a network graph with nodes/dots and edges. It is all about ‘connecting the dots’. Graph analytics encompasses traditional methods of machine learning and deep data analysis.



Network Science is like mapping financial landscapes with financial cartography.

The data model IS the physical model.

Portfolio network relations can be algorithmically visualised and explored interactively (transparency!). The same applies to algorithms based on these network structures: they can be visually transported to create investment intuition and to amplify investment intelligence. They connect with the clients and engage them.

Exploring the network effect, curing the portfolio butterfly defect by de-clustering and de-networking.

Changing from traditional correlations to correlation networks means to switch from ordinary Statistics to Data (Network) Science.



We are involved in numerous industry projects with applied financial graphs

We started our research >10 years ago and our industry traction is >5 years. We are helping investment companies and banks to leverage the potential of financial graphs. We have built up our know-how in finance, regulation, analytics, machine learning and technology. There are projects where we help the financial industry to amplify their internal decision making and risk management of investment portfolios and loan books.


In the Asset&Wealth management industry we enable new unique products, client solutions and digitised services. Examples are:


  • Quantitative investment processes (single and multi asset)
  • Portfolio Health Checks with rebalancing suggestions (both for client servicing reasons as well for regulatory purposes)
  • Visualisation/animation of market dynamics, portfolio diversification effects and the mechanics of network-based investment strategies; these animations also support sales activities


Firamis is basically covering the following FinTech sub-categries:#InvestTech, #WealthTech, #TreasuryTech, #RiskTech, #RegTech, #SalesTech


One of our approaches is special as we can improve the quality of diversification – the only free lunch in finance. The formalism on diversification in Modern Portfolio Theory dates back to the 1950s and Markowitz. It is now time to extend this approach by the financial graph. We apply a concept known from biology here: asset diversity. Maintaining bio diversity in a population is an essential component of nature and it ensures the survival of species. The point is to introduce a graph-based asset diversity. In this way, portfolios can be de-networked and de-risked, making them more robust and even anti-fragile. Similar results have been published independently in prestigious finance journals. Overall portfolio diversity is increased and the contributions to diversity can be drilled-down to single position level. Also, different diversification regimes can be identified and forecast helping to adapt to market changes. This means that wherever diversification is an issue in Asset&Wealth Management, the network graph should be integrated if possible.


The financial graph is currently becoming a very successful alternative key concept in portfolio risk management. Network diagnostics may displace atomised metrics such as VaR. Therefore, we have developed a series of approaches to use the financial graph in many portfolio risk management disciplines: early warning, diversification regime shifts, forensic portfolio scenario analysis and diagnostics, risk concentration management, stress testing, etc.





In credit risk management we have done projects for large banks using the network approach, combined with risk and rating information of each portfolio node. The transactions network of huge corporate loan portfolios, for example, is represented as a graph which basically reflects the physical supply chain behind it. The network method entangles the complicated relations within firm conglomerates and large corporations with many subsidiaries. Also, other firms doing business with this firm are visible, revealing a supply chain with multiple degrees. It is obvious that this view detects fraud (rings) and contagion/spillover of risk across the portfolio.

On bank-wide portfolio level there is a total Credit Value-at-Risk which can be analysed and optimised with the financial graph. Most concentrated risk exposures and largest VaR contributions are today detected by traditional methods along dimensions like originating country or industry. The financial graphs finds hidden dimensions and the largest portfolio VaR jumps could be detected. These insights cannot be found with traditional approaches being characterised by silos and isolated analysis. Connecting the financial graph with the Credit VaR model gives access to a range of new network-based risk management approaches like early warning, forensic portfolio analysis and diagnostics, risk concentration management, stress testing, and scenario analysis.


Network/graph analytics of financial portfolios is the ideal decision making and portfolio management tool

  1. Impact: enables decision makers to focus on those (risk) exposures that are likely to have the greatest potential impact on the performance and risk of portfolios. Already today, network/graph analysis is increasingly used in the financial industry not only for advanced analysis but even for creating new products, solutions and services.
  2. Information quality: in network approaches, the logical model IS the physical model. It is very natural to access the portfolio real world connections with a network model. This implies low assumptions on the model making it robust. Natural complexity is deciphered and rationalised as the driving non-linear relations are accessed.
  3. Human management intelligence: humans are used to think in social relations and networks. Network/graph models are predestined to be visualised by computers and humans are able to interactively browse, navigate and explore the graph structure. Information and insights can be communicated in a compact and engaging way. Machines actually augment the human decision making intelligence in a truly bionic/hybrid approach using the man-machine interface that graph models offer. Interactive network analysis marries judgment developed over time with data-driven insights. This is easily extended to group/community intelligence: several people analysing the same graph structure on their screens can come up with a group-based interpretation and decision.
  4. Digitalisation/Regulation: financial firms worldwide are faced with the challenges of regulation and digitalisation. Since network-/graph analytics is machine based it can be used to systemise decision making processes, documentation and reporting. Technology is about automating risk monitoring and notifying a human so that humans can use their intuition and insight to explore, analyse, and mitigate risks. Processes can be further automated and digitised. Regulatory and compliance issues can thus be addressed with technology (#RegTech).
  5. Speed: Real-time updates of the network enable decision makers to follow the dynamics, hidden shock transmissions and mechanisms, adaption processes and emerging pattern in the financial graph. Models are not created once, deployed and left in place as is; instead, they are continually updated to evolving conditions in real-time. Like in a video stream, variations in pictures are processed.
  6. Establishment: institutions, organisations, regulators, infrastructures, governments and – of-course – Natural and Social Sciences - worldwide use graph/network analysis today. Internet search engines, for example, also use graph analysis to find the best matches. Network analysis has disrupted several industries and finance is next. The development in financial graph analytics over the last few years has been fascinating.
  7. Scientific foundation: Financial Data Science and especially Financial Network Science are on the rise. It looks to create models that capture the underlying patterns of complex systems, and codify those models into working applications. It is data-driven, structure-extracting, explorative, and translates patterns and structures into actionable rules.
    Network Science is systematically mapping portfolio landscapes and therefore classifies as ‘financial cartography’. Also, one of the first model choices in applied complexity science is graph theory and network science.  It can be reasonably stated that financial markets and even portfolios are complex systems: they cannot be further simplified, have spontaneous order, strong dynamical non-linearity involving positive and negative feedback, and emerging large patterns that can originate from the interaction of smaller entities. Portfolios are influenced by the classical butterfly defect and the small-world phenomenon.
    As this had already been recognised by scientists there is a vibrant research in Financial Network Science and its applications in finance today. Natural Science journals are publishing research in financial network analytics and some of the most prestigious finance journals are already doing the same. For example, it was found that network analytics in portfolio management robustifies performance, lowers risk and increases returns. The portfolio choices are visualised directly over the graphic layout.

    Network analytics is revolutionizing how businesses make their most critical decisions. In the past five years we have seen an incredible convergence of algorithms, open source tools, and big data technologies which has empowered virtually anyone in the organization to attempt to solve the most difficult data problems. Network science has fundamentally changed the scale and complexity of the problems we can solve, and the speed at which we can develop and deploy analytics.
    Interconnectedness in financial markets and portfolios will increase due to globalisation like integrated supply chains and information speed.  Traditional formalisms won’t help us here. Network Science is a viable approach. We believe that in a few years  the use of graph analytics to model financial markets and portfolios will be commonplace. The future has already begun as we elaborate on these pages.

Want to read and learn more?

blank