Phonetic matching finds values that sound similar.

It is useful for names because the same name can be spelled in several ways.

Examples:

  • Steven and Stephen
  • Smith and Smyth
  • Jon and John

How phonetic matching works

Phonetic algorithms convert words into a code based on pronunciation. If two words produce the same or similar code, they can be treated as a match.

Common phonetic algorithms include Soundex and Double Metaphone.

🔍 Good to know

Phonetic matching is useful for names. It is not a good default for emails, domains, IDs, or product SKUs.

Phonetic matching vs fuzzy matching

Phonetic matching focuses on sound.

Fuzzy matching focuses on string similarity.

Use phonetic matching when names may be spelled differently but pronounced similarly. Use distance matching when you want a similarity threshold across text values.

Datablist workflow

Datablist exposes Double Metaphone through the Metaphone algorithm in the Duplicates Finder. Metaphone is available on paid plans.

Use it for name properties when Exact or Smart matching misses likely duplicates. The tested 1,000-row sample produced many unrelated name-only groups, so treat phonetic output as candidates rather than confirmed duplicates.

Read the measured Double Metaphone example and compare it with the other algorithms in the data matching guide.

A phonetic score is not a probability

Phonetic matching compares codes derived from text; it does not listen to audio or verify a person's identity. Different names can produce the same code, and pronunciation rules do not work equally well for every language.

Datablist first checks normalized spelling comparisons, then compares the combined Double Metaphone codes when those earlier checks do not match. A match at the phonetic stage receives an 80% field score. That does not mean there is an 80% chance that the records describe the same person.

For a contact list, use phonetic names to discover candidates, then check a second field such as email, phone, company, or location. Keep records separate when the identifiers conflict. For multilingual names, review a sample of both matched and unmatched pairs before using the same rule across the full list.