"Advanced analytics” along with words like machine learning, artificial intelligence (AI), supervised and unsupervised techniques, and anomaly detection have quickly become buzz phrases in the anti-money laundering (AML) compliance sphere. But what, exactly, do these terms mean, and how do we use these technologies to improve our counter-fraud and AML strategies?
At a high level, advanced analytics can drastically impact business by predicting and optimizing outcomes, contributing to growth, producing economies of scale, and improving day-to-day efficiencies of teams. It also provides a competitive advantage to combat financial crimes.
Advanced analytics defined
According to Gartner: “Advanced Analytics is the autonomous or semi-autonomous examination of data or content using sophisticated techniques and tools, typically beyond those of traditional business intelligence (BI), to discover deeper insights, make predictions, or generate recommendations.” It can be used to answer questions like “What is happening?”, “Why is this happening?”, “What happens if the trends continue?”, “What will happen next if…?” and “How can we achieve the best outcomes?”
Advanced analytics includes the techniques, approaches and tools that produce actionable insights from your data and that can be leveraged by AML specialists to examine a suspicious transaction or detect a new pattern of fraud. Anomaly detection, link analysis, textual analytics and machine learning are examples of advanced analytics.
Anomaly detection identifies items, events or behaviors that do not conform to an expected pattern or historical trend. It can be implemented using advanced tools like machine learning to differentiate legitimate transactions from fraudulent ones. When enhanced with context, such as segmentation, the ability to identify fraudulent transactions increases and in many cases, reduces the number of false-positives.
For example, a local retail store will have a different behaviour profile than a corporation that runs an import/export business. Flagged anomalies may include:
Deviation in number, aggregate or frequency of transactions: The local retail store that normally deposits funds into employees’ accounts every month for payroll suddenly sends wire transfers of $300,000 to a high-risk country.
Deviation by customer profile: The local retail store used to send wire transfers using banking services, but starts using small money services businesses (MSBs) to funnel money out of its enterprise accounts.
The world is an increasingly smaller global village with fraudsters leveraging new technologies to develop more advanced fraud schemes. Link analysis produces insights about how different individuals and businesses are connected, the nature of the relationship (provider-retailer; employee-employer; friendship) and the financial transactions that occur between these parties. It is key to unraveling the complex layers used by criminals to launder their proceeds, and to detect and prevent complex money laundering layering topologies, like smurfing and structuring.
Examples of where link analysis connect people include:
Relationships through addresses and geographical proximity: Through screening, a person is identified as a drug trafficker and has a high risk score. People with the same address as the drug trafficker may have an elevated risk score.
Relationships through transactions: If Customer A always sends or receives funds from Customer B and Customer C, link analysis will allow you to identify this pattern and score the strength and risk of the relationships.
Textual analytics is a natural language processing (NLP) technique that provides the capability to extract information from the text contained in unstructured data. The extracted information includes named entities together with their roles and functions such as beneficiaries, dependents, certifying parties, payees and payers; concepts such as topics, relationships between entities, locations of entities, events, actions, intentions, times and dates; and the discourse in which the concept occurs such as a question, a statement of a fact or an expression of sentiment.
The extracted information provides intelligent document summarization as well as input to detect possible money laundering and fraud schemes. Examples of areas where textual analytics is used include the following:
Fraud detection: Using information from previous fraud investigations, textual analytics can identify patterns that may indicate fraud.
Onboarding new customers: When onboarding a new customer, textual analytics can scour public data sources for negative news, allowing the institution to make more informed decisions about whether to do business with this entity.
Machine learning includes supervised, unsupervised, semi-supervised and reinforcement learning. In the case of supervised machine learning, the machine is first trained by providing labelled historical data that is and is not correlated with fraud cases. After the training phase is complete, it is tested with data that was not seen during the training phase to determine whether the model is capable of genericizing and predicting if a newly exposed case is fraudulent or not.
Some examples of how machine learning can benefit financial institutions include:
Advanced transaction monitoring: By applying unsupervised learning on years of historical data, glean patterns or identify potentially suspicious transactions
Learning from similar customers: Using transaction behavior for similar customers, identify unusual and fraudulent behavior in payment cards, loans, wires, transfers
Identifying money laundering topologies: Identify specific risks from both temporal and geographic standpoints
Improving SARs filing processes: Learning correlations between alerts and suspicious activity reports (SARs) to determine when regulatory reports will need to be filed
The future of AML compliance
With intensifying regulatory enforcement and operational losses applying significant pressure on profitability, fraud is no longer acceptable as a “cost of doing business”. One of the keys to address these challenges is to leverage advanced analytics, which are quickly being adopted and will soon be the industry standard. Fused with more traditional rules-based analytics, advanced analytics is the approach that every compliance professional must take to stop fraud and maintain a strong AML compliance program.