Emphasis on mapping the invisible
Our system was originally conceived as an asset tracing tool, and in the course of its development particular emphasis was put on developing techniques to find undisclosed assets and business interests.
It is an obvious fact that in the majority of countries a certain number of economic actors constantly engage in behavior which is aimed at evasion of controlling and repressive functions of the State.
Examples of such evasive practices (subjectively legal or illegal) include:
— avoidance or evasion of domestic taxation
— confidential accumulation of personal wealth abroad
— using sham arrangements and entities, domestic and offshore, for business or criminal purposes
— using opaque foreign structures or domestic front persons to confidentially control assets, etc.
In some countries the persistent use of evasion techniques has lead to emergence of parallel asset control, business interaction, financial settlement and dispute resolution systems outside the country’s borders. Such business communities have developed dedicated ’hubs’ for offshore interaction on domestic matters. Examples are Riga and Cyprus hubs used for informal financial settlements and business dealings around the Russian, Ukrainian and Kazakh economies.
Practitioners who helped perfect our systems have extensive experience with countries where parallel systems for financial settlements, business agreements, corrupt interest partnerships and dispute resolution exist outside their borders.
Saturating the databank with knowledge and defining patterns
It is relatively easy to map relationships and links based on disclosed participations and ownership. Finding indirectly controlled assets is always a challenge to the investigator.
We have established that scenarios of informal or evasive economic activity follow a vast but nevertheless limited number of patterns.
Drawing heavily upon the expert knowledge available to us, we made considerable progress in defining those patterns and describing their manifestations algorithmically.
Originally aimed only at finding undisclosed assets and business interests for the use in debt collection, the system evolved into a universal model saturated with knowledge on specific ways economic actors structure their activity and asset control.
’Artificial Intellect’ is probably too fancy a word to describe the functionality of our system, but nevertheless it not only easily imports the patterns discovered by our experts, but allows to detect patterns which were not observed by us before.
Each time a pattern is put into the system or detected there in the course of a specific query, the whole databank is automatically analyzed for occurrences of that pattern, and positive matches get recorded there as an additional analytical layer. Since new data is added to the databank all the time, the system is programmed to test the databank for occurrences of consistent patterns and ’rules’ (if-then correlations) at regular short intervals. Thus the databank becomes smarter each day.
Since much of the data we use is ’soft data’, such as media reports of uncertain veracity or insiders’ guesses formulated and put into the system, we have developed a way of using such information in the dataset with due regard to its inherently uncertain properties.
Judicial, regulatory, registration and other formal data sources as well as media articles, web platform posts etc. may contain valuable bits of information on persons and assets of interest and externally observable (described in the source or following from it in this or that way) patterns of how those persons conduct their affairs. When a critical mass of such data is collected and structured, the dataset acquires a new quality because the algorithm starts linking pieces of data and finds correlations which were not observable to an analyst before.
Much similar to how sophisticated platforms such as Gmail and Facebook determine that visits through different browsers and from different IP addresses in fact come from one and the same computer (by analyzing the ’digital fingerprint’ comprised of the unique combination of various accessible parameters of the computer), it is possible to define the unique «economic fingerprint» of the given beneficiary’s activity.
One visual analogy being the crystal structure of a material, the economic matrix of the world in the formal plane is made of legal (formal) relationships and links between entities, natural persons and property. If you add an activity element to the analysis, then vectors of activity become observable, and almost all of them can be algorithmically described.
Social networks record social relationships and links between people. We work on describing algorithmically the economic relationships (such as being engaged in a joint business, legitimate or criminal, or deriving a gain from certain activity or asset) and legal (formal) links between economic subjects and property.
We are working on development of algorithmic descriptions of such activity vectors and links so as to include much deeper information than is usually possible, into the dataset.
Activity patterns observable through formal traces of such activity (records, reports, factual information) help us define an «economic fingerprint» (similar to the ’digital fingerprint’ collected by online applications) of a particular person or group or activity (legal or illegal).
In the simplest example, a UK limited liability partnership with two Belize members registered to a post box or a mass incorporation address, revealed by the customs database as sending goods to Russia or Ukraine, with 90% probability would be an instrument in a ’grey imports’ business (goods clear customs at an undervalue, transferred for sale to a dummy entity, and the trade profit sent offshore through one of the illegal capital flight channels). With 50% probability it would have an account with one of the 10 banks in Latvia on our list known to cater to the ’grey imports’ businesses. With 30% probability, it would be banking in Cyprus, 15% — with other banks on our «Banks Favouring Russian Businesses» list, 5% — elsewhere.
Deep research of the offshore world
The offshore world is essential to functioning of ’parallel economies’ and maintaining confidential asset control arrangements in the developing and even in the developed world.
Offshore tools include entities in countries which shield the attributes of such entities (directors, owners etc.) from the public, various financial services (banking, brokerage, fiduciary, trust etc.) rendered in such a way as to disable easy outside detection and a plethora of other privacy enhancing and asset protecting instruments and arrangements.
When creating our system we placed special emphasis on meticulously collecting factual information on offshore entities and how they are linked to people and assets. Besides that we developed algorithmical descriptions of typologies of using offshore tools for various purposes.
We have collected data on offshore actors (corporate service providers, trustees etc.) to be able to link particular offshore nominee officers with specific offshore tax haven providers, and them in turn — with onshore service providers.
We have been able to collect a database of company names that had been made available for sale as ’shelf’ (ready-made) companies over the past 10 years — so that if this or that company resurfaces later in an asset tracing scenario, we know who to ask about the ordering party.
We have worked over hundreds of thousands of pages of ’soft’ sources including media reports, regulatory acts, litigation submissions and court records etc. to extract information on the use of offshore structures.
The result is our increased ability to detect economic behavior and assets of target actors which is shielded from public view and State control, and is not readily detectable by private interest.
This provides opportunities for interested parties in the course of investigations, pre-litigation discovery, post judgement recovery and other scenarios where information about the target’s assets is not readily available.
Modeling of protective arrangements and structures
In-depth knowledge of asset protection techniques by our contributors helped us single out and then algorithmically describe the real-world manifestations of various asset protection schemes and tricks. Thus we can deduce with varying certainty that this or that asset is being controlled by a structure specifically designed to maintain secrecy of its owner and/or thwart litigation against its economic beneficiary.
We have also used the knowledge of available techniques for penetration of protective structures, so that simultaneously to detecting an asset protection scheme the system would be looking for indicators of possibilities to disrupt the protection.
Breaking open asset protective structures is a complex legal area, and when multiple jurisdictions are involved the costs of even only assessing the potential legal options could be very high. Our close cooperation with lawyers having decades of asset protection and asset recovery experience helped us design tools to optimize this research stage.
It is important to note that PRIVATEPOL Data Analytics is not a law firm, and the tools and services we make available to our customers should be used to complement, not supplement legal advice.
Dealing with uncertain information
In an ideal scenario we would be able to access complete, true and up-to-date datasets and other information for the purpose of database construction. The reality is different.
Many corporate registers are closed for automated access or not available in one dataset at all (such as the BVI register of international business companies, for example), which means they have to be reconstructed, and inevitably some errors and omissions would occur. Personal information is by definition patchy and sometimes comes from unreliable sources. When working over media reports we have to always allow for certain bias and sometimes — outright misinformation. Our experts formulate database ’rules’ on the basis of their experience and insight, but they, too, may err.
Existing global corporate/personal databases mainly contain data derived from official or otherwise authoritative sources, and this position fits well with their claim of providing high quality information. We are, on the contrary, developing techniques allowing us to use incomplete and dodgy bits of information and still enrich the dataset. Our aim is not to guarantee the accuracy of every bit of data in the system, but to allow the system and the user to formulate leads, hypotheses and probable scenarios on the basis of the entirety of the data.
In an asset tracing or crime investigation any probable clue is better than no clue at all. A source with very low trustworthiness rating (Eg. internet post) may nevertheless provide a missing link enabling the analyst to put a workable hypothesis together.
One big challenge we had to overcome when developing the software for our system was allowing for various degrees of certainty in the data, including the ’if-then’ rules formulated by our experts. The issue was not only exponential complication of the database’s architecture (due to introduction of another dimension — degree of certainty for some data elements and ’rules’), but also the ability to deliver probability-weighted results.