Interpreting Identity Match Scores

Understand the scores returned by /identity/match and how to interpret each field

This page explains the data returned by Identity Match and how to interpret each field's match score.

Data returned by /identity/match includes scores from matching a user's name, address, phone number, and email with the account owner's data that was present on the connected account. A score ranging from 0 to 100 is provided for each identity field. The score for a given field will be missing if the identity information for that field was not provided in the request or unavailable at the connected account. In the case of an Item containing accounts with multiple owners, the highest matched scores are returned.

You should typically not set the match score requirement for a field to 100. For example, if a phone number match score of 100 is required, formatting differences in a number that is not a valid phone number, such as a missing country code, will cause a mismatch. 70 is the default recommended match score threshold for all fields.

Name match score

Plaid normalizes names and accounts for many common discrepancies, such as nicknames, prefixes and suffixes, reverse name order, omitting or initializing middle names, multiple surnames, diacriticals, case inconsistencies, and more. After accounting for these common issues, names are compared for similarity using a fuzzy string matching algorithm.

Name match scores of 70 or higher are likely to be matches. Scores from 50 to 69 may be matches and you may wish to configure these for human review. Scores below 50 are unlikely to be matches.

The legal_name object also returns booleans that help explain the score. is_nickname_match and is_first_name_or_last_name_match indicate why a partial score was assigned. is_business_name_detected is true when Plaid determined that one of the names being compared is a business name.

Range	Meaning	Example
100	Exact match	Jennifer Smith, Jennifer Smith
99	Exact match after normalization	Jennifer Smith, Ms. Jennifer Smith, Jennifer W. Smith; Li Chen, Chen Li
95-98	Strong match	Andrew Smith, Andrew Smith Jr.
90-94	Strong match, likely single typo or nickname, or missing portion of Spanish surnames	Jennifer Smith, Jen Smith, Jennfer Smith; Alejandro García Rodriguez Santos, Alejandro Garcia Rodriguez
70-89	Likely match, likely abbreviated / missing portion of name, or multiple typos	Jennifer Smith, J Smith, Jennifer Williams Smith, Jennfer Smth
50-69	Possible match, likely relative or missing portion of non-Spanish surnames	Mohammed Naseer Daoud Mohammed, Mohammed Nasser Daoud; Jennifer Smith, Betty Smith
0-49	Unlikely match	Jennifer Smith, Ray Charles

Phone number match score

For phone number and email match scores, Plaid returns discrete values rather than a range, as fuzzy matching is not used for these comparisons.

Range	Meaning	Example
100	Exact match, or the same valid phone number formatted differently	+1-212-867-5309, (212) 867-5309
80	Same digits with different formatting and/or missing country code, for a number that is not a valid phone number	+1-555-867-5309, 5558675309
0	Not a match	+1-555-867-5309, +1-555-867-5308

Email match score

Two strings are considered to be an email match if an email sent to both email addresses will end up in the same inbox. For phone number and email match scores, Plaid returns discrete values rather than a range, as fuzzy matching is not used for these comparisons.

Range	Meaning	Example
100	Exact match	lknope@pawnee.gov, lknope@pawnee.gov
85	Exact match after normalization	lknope@pawnee.gov, l.knope+plaid@pawnee.gov
0	Not a match	lknope@pawnee.gov, thaverford@pawnee.gov
`null`	Not a properly formatted email address	lknope@pawnee.gov, lknope@pawneegov

Address match score

Postal code matching is prioritized above city and state matching. A postal code mismatch, even by a single digit, will automatically result in a score of zero. An exact five-digit postal code match with a matching street address and an incorrect city or state will still result in a match.
When matching US postal codes in a nine-digit format (e.g. 91101-1002), Plaid will ignore the four-digit suffix and compare only on the five-digit code.
Street addresses are normalized, and the order of the street address components is ignored. For example, "100 First Avenue N, Apt. 5B" and "Apartment B5, 100 1st Ave. North" would be an exact or close-to-exact match.
After these rules are applied, street addresses are compared for similarity using a fuzzy matching algorithm.

Range	Meaning
100	Exact match for normalized street address and postal code
99	Exact match for normalized street address and postal code; likely city name or state mismatch
90-98	Likely a single typo, e.g. 203 Main St. vs. 204 Main St.
80-89	Likely two typos, e.g. 295 Laffayete vs. 295 Lafayette
70-79	Likely three typos, e.g. 295 Laffayete vs. 294 Lafayette
1-69	Postal code match, likely street address mismatch
0	Postal code mismatch

Sample Identity Match data

{
  "accounts": [
    {
      ..
      "legal_name": {
        "score": 90,
        "is_nickname_match": true,
        "is_first_name_or_last_name_match": true,
        "is_business_name_detected": false
      },
      "phone_number": {
        "score": 100
      },
      "email_address": {
        "score": 100
      },
      "address": {
        "score": 100,
        "is_postal_code_match": true
      }
      ..
    }
  ]
}