Moov Watchman’s Advanced Similarity Analysis

Introduction

Moov Watchman implements a sophisticated multi-layered similarity matching system that significantly outperforms basic search tools like OFAC’s SDN search portal. The system features a well-architected matching framework designed for high accuracy in sanctions screening and compliance workflows. This document provides a detailed overview of how Watchman’s scoring algorithms work.

Core Matching Architecture

Watchman’s matching engine uses a hierarchical approach to entity matching that considers multiple dimensions of similarity with appropriate weighting for each component:

• Critical Identifiers (highest priority): Exact matches of government IDs, crypto addresses, and contact information

• Name Comparison (high priority): Sophisticated fuzzy matching of entity names and titles

• Supporting Information (medium priority): Contextual details like addresses, dates, and other entity metadata

This layered design allows Watchman to make nuanced decisions about match quality, prioritizing the most definitive identifiers while still considering supporting evidence.

Advanced Name Matching

The core of Watchman’s matching capability lies in its sophisticated name matching algorithms:

Jaro-Winkler with Customizations

While OFAC or other typical search uses basic keyword matching, Watchman employs enhanced Jaro-Winkler string similarity with custom adjustments:

Token-based Comparisons: Names are broken into tokens/words, and each token from the search query is matched against the indexed terms using a pairwise approach

Length Difference Penalties: The algorithm penalizes matches between terms with significant length differences

Different First Letter Penalties: Additional penalties are applied when words begin with different letters

Adjacent Term Similarity: The algorithm checks for matches in adjacent positions, accommodating name inversions and word order differences

Phonetic Matching

Watchman implements phonetic matching to catch similar-sounding names regardless of spelling variations:

Soundex: The system uses a modified Soundex algorithm that groups phonetically similar letters

First Character Analysis: Names are compared based on their first character’s phonetic class, reducing unnecessary comparisons for clearly dissimilar names

Efficiency Optimization: Phonetic filtering is applied early in the comparison process to eliminate obviously non-matching pairs

Entity-Type Specific Matching

Watchman differentiates between various entity types (Person, Business, Organization, Vessel, Aircraft) and applies specialized matching rules for each:

Person Matching

Government ID Comparison: High priority for exact matches of ID numbers

Title Comparison: Fuzzy matching of professional titles with abbreviation expansion

Birth/Death Date Analysis: Date comparison with consistency checking between birth and death dates

Logical Validation: The system checks that dates make temporal sense (birth precedes death)

Business/Organization Matching

Registration ID Comparison: Primary focus on government/tax identification numbers

Name Variations: Consideration of alternate names and historical names

Creation/Dissolution Dates: Temporal validation for business lifecycle

Vessel/Aircraft Matching

Industry-specific Identifiers: Primary matching using IMO numbers, call signs, MMSI (for vessels) or serial numbers and ICAO codes (for aircraft)

Registration Details: Flag country, model, and ownership comparison

Address and Contact Information Matching

Watchman applies weighted matching to different address components:

Primary Address Line: Highest priority (most important component)

City and Country: High priority (critical for location)

Postal Code: Strong verification priority

Secondary Address Info and State: Supporting information priority

This weighted approach prioritizes the most important address components while still considering all available information.

Comprehensive Scoring System

The final match score is calculated through a sophisticated process:

Weighted Component Aggregation: Each component score is multiplied by its importance weight

Critical Field Multiplier: Required fields are given additional weight

Coverage Analysis: Penalties are applied if the query doesn’t cover enough of the index entity’s available fields

Perfect Match Boosting: High-quality matches that meet specific criteria receive a boost

Type Mismatch Handling: Entity type mismatches result in a zero score

Score Adjustment Based on Match Quality

Watchman applies various score adjustments based on match quality:

Minimum Field Requirements: Penalties for queries with insufficient field coverage

Name-Only Match Penalties: Reduced confidence in matches based solely on name

Historical Name Handling: Names identified as “Former Name” receive a penalty

Exact Match Override: Certain exact matches of critical identifiers automatically receive a perfect score

Implementation Optimizations

Watchman includes several optimizations that improve both accuracy and performance:

Early Termination: The algorithm stops comparing once a high-confidence match is found

Phonetic Filtering: Preliminary filtering reduces unnecessary detailed comparisons

Selective Deep Analysis: Detailed analysis is applied only to promising candidates

Configuration Flexibility: The system allows tuning of algorithm parameters

Benefits for Compliance Teams

For risk and compliance professionals, Watchman’s sophisticated matching offers several benefits:

Reduced False Positives: The multi-dimensional scoring and contextual analysis help differentiate between true and false matches

Improved Match Confidence: Detailed scoring provides better justification for match decisions

Audit Trail: The system’s detailed scoring components provide clear rationale for matches

Configurable Thresholds: Teams can adjust sensitivity based on risk tolerance

Enhanced Detection: The sophisticated algorithms can identify potential sanctions matches that would be missed by conventional systems

Conclusion

Moov Watchman’s advanced match scoring system represents a significant improvement over basic search tools like OFAC’s search portal. By combining probabilistic string matching, phonetic analysis, entity-specific logic, and weighted multi-dimensional scoring, Watchman provides compliance teams with more accurate and explainable match results.

The system’s attention to detail—from handling name variations to applying logical consistency checks—demonstrates a deep understanding of the challenges in entity resolution for sanctions screening. For auditors and compliance professionals, this means greater confidence in screening results, better justification for match decisions, and a more robust compliance program overall.