Matching Frameworks within the Business Registry Verification

Overview

Persona offers multiple matching frameworks inside the Business Registry Verification. These frameworks can be leveraged in different ways or in different use-cases to help increase match rates, while preserving accuracy. The combination of these frameworks is known as Contextual Matching

Contextual Matching

The BRV offers two main matching systems:

1. Rules-Based Match Groups: A set of rules defined on a per-country basis for matching businesses, based on extensive research on businesses in those countries
2. AI-Powered Matching: A matching system that uses LLMs to evaluate business information and determine matches

The combination of both of these frameworks is known as our Contextual Matching system. By default, AI-powered matching is used as a fallback for rules-based matching.

In early testing, we saw Contextual Matching (Rules-based Match Group with fallback to AI-Powered matching) improve match rates for legitimate businesses by up to 55% compared to standard matching frameworks.

Availability of matching frameworks

The Business Registry Verification offers AI-powered matching for all countries the BRV covers and offers country-specific match groups for select countries.

Rules-based Match Groups

Our rules-based matching system is based on explicitly defined criteria for each attribute of a business (business name, registration number, street, city, etc.) to determine matches. We’ve defined these rules on a per-country basis by doing research into specific countries and the business registry information available in that country. For example, we’ve meticulously researched businesses in the EU, identified what legal elements can be found in business names, and created rules to account for the presence or absence of these legal elements.

This matching system is deterministic and explainable. For any inputted business information, Persona will always arrive at the same decision and clearly explain how that decision was made by looking at the rules that were encoded. Persona's rules-based matching system excels at matching businesses in countries where business registry information is well-organized and reliable.

AI-powered Matching

Persona' AI-powered matching system uses LLMs (large language models) for matching business information. While our rules-based match groups examine the attributes of each field via string similarity, our AI-powered matching system looks at each attribute holistically. This approach allows LLMs to apply their built-in understanding of natural language and context to interpret the business information and determine matches.

The benefit of using AI-powered matching is that it is able to find matches where our rules-based match groups might not. For example, we’ve seen our AI-powered matching system account for typos from users, determine a match when users abbreviate part of their business name, and even determine a match when users enter business information in their country’s native language, but our vendors store the business information in English.

The match requirements for AI-powered matching can be configured with your Post-Sales team but are not available for self-service editing. By default, Persona has defined a match that leverages AI-powered matching as:

1. Full business name match + partial registration number
2. Full business name match + partial business address
3. Partial business name + full registration number
4. Partial business name + full business address

Interpreting AI-powered Matching

BRVs that run with AI-powered matching will offer a view similar to match groups. Across each of the properties, you can see the matched record, match result, and a natural language analysis of the match result.