The fight against fraud and AML often hits a ‘brick wall’ when encountering labyrinthian international, corporate structures. Even well-known brands such as Google and Starbucks, use controversial, though currently accepted, international structures to optimize capital flows and to reduce tax liabilities, for instance the so-called Double Irish or Dutch Sandwich.
Due to this complexity, increasingly it is a challenge for regulators to determine whether a particular construction, and associated transactions, are acceptable or questionable. A part of the problem is that the tools for tracking capital flows between structures are typically linear, following paths on a static process diagram. When transactions are circular, or involve more than several ‘hops’, the linear method breaks down.
The reality of global conglomerates is that they are typically less a top-down organization, and more a complex, interconnected network of cross-ownership, control, and co-investment. Auditing such labyrinthian beasts is enough to bring a grown auditor to tears. This is where network ‘graph analytics’ can be invaluable. By tracking complex structures as native networks, fraud and AML investigators can simplify the process of understanding and tracing complex transactions.
As an example, Samsung Group, the South Korean multinational conglomerate, is famously known for the complexity of its interconnected cross-ownership structure. Samsung has run into antitrust, bribery, tax evasion, and embezzlement accusations and scandals. A difficulty facing regulators is the complexity of the partial ownership structures involved. As per a recent Economist article attempting to explain the situation, “for example, the group’s holding company, which has just changed its name from Samsung Everland to Cheil Industries, owns 19.3% of Samsung Life, which owns 34.4% of Samsung Card, which owns 5% of Cheil.… This corporate hairball has let the Lees exert control over the group with a stake of less than 2%.” 
Samsung Group cross-ownership structure (Credit Suisse)
Indeed Samsung Group is much more a network of interconnected interests and ownership. As such, representing it natively as a network graph allows for a greater degree or insight and control. As an example to the use of graph analytics, we translated the complex Samsung Group ownership structure map into a network graph. The below visualization is via the open source network visualization tool Gephi.
Samsung Group cross-ownership structure in Gephi network visualization
More than a pretty picture, by storing the structure as a native network graph, we can go much further. By storing the network structure in a graph database, such as Neo4J, we set up a structure which allows for for deep quantitative analytics.
- Want to see the code and live visualization? Link here to Neo4J Graph Gist
- A second example on graph gist for money laundering
Once in a network format, using graph mathematics and network analytics, we can examine the network itself for classical network properties such as classical measures of centrality, such as degree and betweeness. Across many case of international constructions, it may be possible to correlate such measures with a propensity of fraud. This would involve examining known cases of fraud in conglomerates and examining the correlation with particular network measures.
Beyond this, forensic accounting is often flummoxed when encountering complex circular money flows between subsidiaries and cross-ownership structures. The native graph structure can also be used to input, track and calculate transactions between the businesses, and to summarize and aggregate the impact of these transactions across complex chains.
For instance, if profit distributions flow from one company to another via many different paths, these chains can quickly be queried and calculated using a graph database. Graph databases are a promising, powerful tool for tracking and mitigating complex cases of international fraud involving international business structures. The tools and expertise for powerful AML are available and we are ready to help you deploy!
Want to learn more?
- Network analytics for fraud detection & mitigation (Deloitte video demo)
- Deloitte blog posting on SNA for fraud: https://www2.deloitte.com/nl/nl/pages/risk/articles/cutting-edge-network-analytics-detect-fraud-financial-crime.html
- RSM analytics lecture Demonstration of semantic analytics
- Blog post Excuse me, do you speak fraud?
- Blog post Network analytics for fraud detection
- Blog post Network analytics: more than pretty pictures
- Blog post What information is gained from social network analysis?
Samsung Group cross-ownership structure in Neo4J graph database
