Search Icon

What is Global Watchlist?

Databases used to run regular identity checks against suspected terrorists, money launderers, fraudsters, politically exposed persons (PEPs), adverse media or sanctions lists maintained by governments, regulatory agencies and commercial enterprises.

Global Watchlist Matching Techniques

Matching techniques are a critical component of Socure’s Global Watchlist solution. They involve the algorithms and methods used to compare the names or entities in the watchlist against a real-time database of names or entities. Several matching techniques can be used to identify potential matches, each with its strengths and weaknesses.

  • Exact matching: This technique involves comparing the name or entity in the watchlist against the names or entities in the database, character by character, to find exact matches. This technique is the most accurate, but it may miss potential matches due to minor differences in spelling or formatting.
  • Fuzzy matching: This is a more flexible matching technique that considers differences in spelling, formatting and other variations in names or entities. It assigns a score to each potential match based on the level of similarity between the name or entity in the watchlist and the names or entities in the database. The fuzzy matching technique is more inclusive than exact matching but may also generate false positives.
  • Phonetic matching: This technique compares the sound of names or entities rather than the spelling or formatting. This technique is useful when the names or entities in the watchlist may have needed to be corrected or transcribed correctly. It is less accurate than exact matching but can still identify potential matches that other techniques would miss.
  • Rule-based matching: This technique involves setting up rules that specify the conditions for a potential match. This technique is highly customizable and can be tailored to specific needs, but it may also miss possible matches that do not meet the specified rules. For example, the rules may specify that a match must occur if the name in the watchlist matches the name in the database and the birthdate is within a certain range.
  • Machine learning-based matching: This technique involves using machine learning algorithms, such as neural networks and Long Short-Term Memory (LSTM), to identify patterns and similarities between names or entities in the watchlist and names or entities in the database. This technique can be very accurate, especially when the extensive database and the names or entities in the watchlist are complex. It requires a large amount of training data to work effectively.

Socure’s unique approach to watchlist matching and entity resolution​

At Socure, we leverage artificial intelligence and machine learning to develop effective watchlist matching and entity resolution. Our approach uses state-of-the-art neural networks, specifically Long Short-Term Memory (LSTM) networks, to learn how to detect subtle patterns in names and identify relationships between variants.

LSTMs are neural networks well-suited to handling complex sequences and variants. Our models have learned to cluster names that likely correspond to the same individual, even across languages and writing systems. This means that when a name like Алехандро Фернандез vs. Alejandro Fernandez vs. AL3HANDRO FERN4NDETH is entered, our LSTMs can match it to the appropriate cluster and determine the likelihood it refers to the same person on the watchlist.

One of the key advantages of our approach is that it does not rely on rigid and limited rules-based systems, which are often unable to capture the complex variations and patterns in names and entities. Instead, our AI has learned these skills by analyzing billions of names and aliases and can recognize and cluster them accordingly.

Machine learning-based matching techniques, such as our LSTM models, are highly effective at identifying potential matches, even when the names or entities in the watchlist are highly complex and may have multiple variations. They require a large amount of training data to work effectively, but once they have been trained, they can detect subtle patterns and relationships between variants that would be difficult or impossible for humans to identify.

Our Global Watchlist solution is designed to provide unmatched levels of accuracy and scale, and we are committed to ensuring that our models are transparent and ethical. For this reason, we encourage you to ask for Watchlist Model Governance, which provides a comprehensive overview of our approach, methodology, and performance metrics.