Proximity Matching

This page describes TargetSmart's proximity matching technology. Proximity matching is utilized when insufficient address information is available on an input record or list. While this isn't an optimal approach to matching, it can be highly effective when its necessary due to data constraints. TargetSmart's system has been designed as a matching framework that can be customized by our clients based on their application, business processes, and/or their comfort level.

Components

• Similarity - This is an evaluation of how closely the name elements match the prospect names in a given geographic search area. We use extensive nickname tables, gender lookup tables, and various difference algorithms to evaluate the similarity between the input record and the prospect list.
• Uniqueness - The number of potential matches available for a given geography, given the similarity of the names.
• Proximity - The distance (in miles) a potential match is from the input address or geographic centroid. Each input records and candidate match is geo-coded, and then distance between that pair is calculated. The centroid is calculated at the most granular level possible given the input data available.

Scoring

TargetSmart provides a series of scores ranging from 0-100 for each of the proximity matching components.

• Similarity - 0-100 Score - 0 is a "no similarity" and 100 is an exact, byte-for-byte match on all name components.
• Uniqueness - 0-100 Score - 0 is an extremely common name with more than 25 potential matches in the geographical area, and 100 is a single name in the area.
• Proximity - 0-100 Score - 0 is a distance greater than 100 miles from the centroid of the input record and 100 is less than a mile.

Composite Scoring & Confidence Coding

The composite score is a weighted combination of the three underlying proximity scores and can be customized by our clients.  Our default composite score is calculated by the following formula:(Similarity + Proximity + Uniqueness) / 3. Additionally, we translate the composite score into a confidence code, which can also be customized by our clients based on their particular application and business needs. Our default confidence codes are:

• Composite Score of 90-100 Score - Excellent Match - Match Candidates Returned
• Composite Score of 80-89 Score - Highly Likely Match - Match Candidates Returned
• Composite Score of 70-79 Score - Likely Match - Match Candidates Returned
• Composite Score of 50-69 Score - Potential Match - Match Candidates Returned
• Composite Score of Less than 50 - No Match - No Match Candidates Returned

Examples

The following examples illustrate the various types of matches that can be returned from our proximity match system:

Input Record - John Q Public | No Address Information | ZIP Code 12345

Candidate Record	Scoring	Confidence
John Quincy Public 10 Miles from ZIP Centroid 4 Potential Matches	Similarity Score = 95 Proximity Score = 90 Uniqueness Score = 75 Composite Score = 87	Probable Match
Jack Public 18 Miles from ZIP Centroid 4 Potential Matches	Similarity Score = 85 Proximity Score = 82 Uniqueness Score = 75 Composite Score = 81	Probable Match
J Q Public 27 Miles from ZIP Centroid 4 Potential Matches	Similarity Score = 80 Proximity Score = 73 Uniqueness Score = 75 Composite Score = 76	Potential Match
J Publick 47 Miles from ZIP Centroid 4 Potential Matches	Similarity Score = 70 Proximity Score = 53 Uniqueness Score = 75 Composite Score = 66	No Match