Fintechs: Using Machine Learning To Fight Money Laundering

FinTechs are at the forefront of the fight against money laundering. It is the responsibility of the financial institutions' compliance departments to comply with the European Anti-Money Laundering guidelines. The money houses try to track down the perpetrators by unmasking, and removing from their customer bases the most widespread money laundering vehicles, so-called letterbox companies. With conventional compliance tools, however, the institutes quickly reach their limits.



Shell Companies are the classics to launder black money. These are companies that mostly only exist on paper and are located in tax havens such as the Cayman Islands or the Virgin Islands. Criminal organizations are a sure-fire way to remain anonymous and disguise criminal transactions by faking legal purposes for illegally acquired funds.



According to estimates by the United Nations Office on Drugs and Crime (UNODC), between 800 billion and 2 trillion are produced worldwide each year and washed us dollars. This corresponds to a share of 2 to 5 percent of the gross domestic product generated worldwide.


Rough Conclusions About the Legality


FinTechs are at the forefront when it comes to draining the structures behind letterbox companies. For a long time, it was possible to draw rough conclusions about the legality based on indications such as the amount and frequency of incoming and outgoing amounts of money, a strikingly small number of employees, and the company's location. Today, it has become much more difficult for fintech to identify mailbox companies and close loopholes that criminal organizations use to launder money.



Digitization has increased the complexity of money laundering cases. FinTech must be able to separate the wheat from the chaff here: which company is still operating legally, and who is actually making itself punishable? Because not every tax optimization is illegal, and not every subsidiary is prohibited in a tax haven. To make matters worse, money laundering-related company data is mainly collected at the individual country level. Intergovernmental cooperation solutions are not yet sufficiently developed to recognize potential mailbox companies across the board.


Collecting, categorizing, and analyzing data is, therefore, the best opportunity for fintech. However, the right processes are still missing in many institutes. The use of supportive technology is becoming increasingly important so that fintech can comply with all regulatory guidelines and meet compliance requirements.


Data Analysis as a Clarification Accelerator


The solution is Data Analytics and Machine Learning (ML). Practice shows that a significant number of suspected cases can be identified using these methods. Various solutions are conceivable, from guided to self-learning machine learning models, including self-building decision trees, neural networks, and integrated analytics platforms. Which of these makes the most sense depends on the data available. If the right technology is selected, it detects abnormalities from large amounts of data that would otherwise only have been reported externally via police authorities, correspondent fintech, or from the industry network.


Through constant adaptation, technology is able to react independently to changes in transactions and customers. If it issues a hit, an employee checks it in the process and decides on a reaction, for example, whether the customer has to be reported to the authorities or not. The machine learning model aims to keep the rate of "wrong hits" as low as possible and to raise the alarm only in justified cases. If the learning system reduces the false-positive rate, additional efforts can be cushioned directly. This saves project budgets and enables the solutions to be scaled faster. Ten-year plans are a thing of the past.


The key to success is "know-your-data" and must be part of the overall concept for preventing money laundering by fintech. Data and its analysis are an effective way to unmask mailbox companies. But first, the data has to be put in order, and its quality checked. FinTech has numerous data points for its business and private customers: from payment behavior, the length of stay, and the amount of the amounts in the account to the transaction parties. However, if these data sets are distributed over many data silos, comprehensive data analysis becomes difficult. FinTech can control their amounts of data and their modeling themselves with a five-point plan:

  • Strategic vision
  • Data collection
  • Data analysis
  • Understanding of data
  • Machine learning 

Strategic Vision

The strategy comes first. Fintechs must be aware of the requirements that a machine learning model meets and what the result should look like. The model must also fit into a predefined target image so that it is clear how the results delivered are interpreted and which answers are sought.


Data Collection

Based on the strategic concept, the next step requires the exact description and procurement of required transaction and customer data as well as the establishment of technical databases and software. Financial institutions should also include external data such as company lists, negative reports, and sanction lists in the calculation.


Data Analysis

Data Analytics helps to categorize company constructs, put them in relation, and make frequencies and distributions visible and unify them. Numerous analysis programs can be made available quickly and efficiently and easy to use.


Understanding of Data

Analyzed data must be understandable. The more detailed the data, the more precise analyzes machine learning programs can perform and recognize patterns or cross-connections. Using data from freely accessible sources can complete data sets and thus enable better results. Based on this logic, the first prototype of a machine learning model can be created.


Machine Learning

A machine learning model lives from its data and learns more the more different data samples are repeated. This increases statistical accuracy, explains the dynamics of the model, and leaves room for further design. The model can be adapted and improved by subsequently making the results plausible.


Conclusion

The advancing digitalization - also on the part of criminal organizations - is forcing FinTech to keep up with technological advances in the compliance area and to review their anti-money laundering framework constantly. With data analytics and machine learning, it is possible to react quickly to the increasingly complex money laundering processes, reduce the risk of non-reporting, and increase cost-efficiency. If FinTech succeeds in this technological leap, they not only have more capacity for their core business but can also increase their hit rate in suspicious transaction reports.


Let's Fight Financial Crime Together

Previous Post
Negative Effects of Money Laundering on The Economy
Next Post
Anti-Money Laundering and Cybersecurity