Double Metaphone Algorithm: How It Works, Examples & Calculator

What Is Double Metaphone?

Double Metaphone is a phonetic algorithm that converts a word or name into a code based on how it is likely to sound. It helps identify names with different spellings that may have similar pronunciations. Developed by Lawrence Philips, it extends the original Metaphone algorithm with rules for a wider range of pronunciations. It can produce two codes for a name: a primary code and, where relevant, an alternate code representing another possible pronunciation. In data matching, Double Metaphone can help find potential matches that an exact spelling comparison would miss. For example, it may flag differently spelled surnames for further comparison. A shared phonetic code is only a clue, though: different people can have names that sound alike. For a reliable matching decision, the name should be considered alongside other fields such as address, email or date of birth.

How Does Double Metaphone Work?

Double Metaphone reads a name from left to right and applies pronunciation rules to its letters and letter combinations. It then produces a short phonetic code that represents how the name is likely to sound. The rules consider context. For example, the sound of a letter can depend on the letters beside it or where it appears in the name. This makes Double Metaphone more flexible than simply replacing each letter with a fixed number or symbol. The algorithm can produce a primary code and, when there is another plausible pronunciation, an alternate code. Matching systems can compare these codes to identify names worth investigating, even when their spellings differ. A matching code indicates phonetic similarity; it does not prove that two records refer to the same person. Other fields should contribute to the final matching decision.

Primary and Alternate Phonetic Codes

Double Metaphone can return up to two codes for a name. The primary code represents its most likely pronunciation under the algorithm’s rules. The alternate code represents another plausible pronunciation when the spelling allows for one. Some names produce the same code for both. These codes give a matching system more ways to find names that may sound alike. It could compare primary codes only, or also consider alternate codes. Including alternate codes can find additional candidates, but it may also bring in unrelated names with similar sounds. That trade-off matters when configuring a matching rule. A phonetic code is best treated as evidence to weigh alongside other fields, rather than as proof that two records belong to the same person.

Double Metaphone gives “Smith” a primary code of SM0 and an alternate code of XMT. Which codes are compared depends on the matching configuration.

Double Metaphone Examples

Consider three surnames with similar pronunciations but different spellings:

Smith and Smyth share a primary code. Smith and Schmidt share a code only when alternate codes are compared.

Smith and Smyth share the same primary code, so a primary-code comparison would identify them as a potential phonetic match. Smith and Schmidt illustrate why alternate codes exist. Their primary codes differ, but Smith’s alternate code matches Schmidt’s primary code: XMT. A system that compares alternate codes could therefore flag the pair for review. A system configured to compare primary codes only would not find a phonetic match between them.  Neither result establishes that the records describe the same person. A matching rule should also consider evidence from fields such as address, email or date of birth.

Double Metaphone vs Soundex and Metaphone

Soundex, Metaphone and Double Metaphone all turn names into phonetic codes, but they use different rules. This means the same pair of names may be identified as a potential match by one algorithm and missed by another.

For example, Smith and Smyth share a Double Metaphone primary code. Smith and Schmidt share a code only when an alternate code is considered. This illustrates why the configuration matters as much as the algorithm’s name: comparing alternate codes may find more candidates, while also requiring closer review of potential false matches.

There is no universally best phonetic algorithm for every dataset. The choice depends on the names being matched, the acceptable number of missed matches and false matches, and the other evidence available in each record. Test the algorithms against representative data before using a shared code as part of a matching rule.

When Is Double Metaphone Useful?

Double Metaphone is useful when names may be recorded with different spellings but similar pronunciations. This often happens when customer records are entered by different people, names are heard over the phone, or data is combined from multiple systems.

Common applications include:

• Customer deduplication: Finding potential duplicate records when a surname is spelled differently.
• Record linkage: Identifying possible matches between customer, patient or supplier datasets.
• Search: Returning relevant names even when a user enters a different spelling.
• Data migration: Finding records that may refer to the same person before systems are merged.

It is most valuable as one part of a wider matching rule. A shared phonetic code can bring two records into consideration; agreement on fields such as postcode, address or date of birth can then provide stronger evidence. The appropriate configuration should be tested against real examples from the dataset, since pronunciation patterns and the cost of false matches vary by use case.

Limitations and False Matches

Double Metaphone compares likely pronunciation, not identity. Two unrelated people can have names that produce the same code, while two spellings of the same person’s name may produce different codes.

Its main limitations include:

• Phonetic collisions: Different names can share a code, creating potential false matches.
• Missed variations: Nicknames, initials, name changes and substantial spelling errors may not produce matching codes.
• Pronunciation differences: The same name may be pronounced differently across languages or regions, and the algorithm cannot capture every variation.
• Alternate-code trade-off: Comparing alternate as well as primary codes can find more candidates, but may also flag more unrelated names for review.
• No measure of identity: A matching code does not tell you whether the address, date of birth or other identifying details agree.

For example, matching two records solely because their surnames share a Double Metaphone code could incorrectly combine different customers. Use the code to identify a possible match, then check other fields before making a final decision. Test the rule on representative data and inspect both missed matches and false matches when deciding whether to include alternate codes.

Why a Phonetic Code Shouldn’t Decide a Match Alone

A Double Metaphone code answers a narrow question: could these names sound alike? It cannot determine whether two records describe the same person. Different customers may have surnames with identical codes, and a name may be entered correctly in both records even when the remaining details conflict. Consider two records whose surnames share a phonetic code. If their postcodes and dates of birth differ, merging them on the surname alone would be risky. If their addresses and other identifying details also agree, the shared code becomes more useful evidence.

The phonetic surname match identifies a possible match; agreement on the address, postcode and date of birth provides stronger supporting evidence.

A stronger matching rule combines the phonetic comparison with checks on other fields. It can then distinguish between a likely match, a possible match that needs review, and records that should remain separate. The weight given to the phonetic code should reflect the quality of the data and the consequences of a false match.

Using Double Metaphone in AdaptiveMatch™

AdaptiveMatch™ is designed to apply matching algorithms to individual columns. A rule could compare surnames with Double Metaphone, addresses with fuzzy matching and email addresses exactly, using the combined evidence to identify potential matches. The planned Double Metaphone configuration uses one code per value rather than comparing every primary and alternate code combination. This may reduce false matches, but could miss some name pairs. Test the rule on representative data before applying it more widely.

Frequently Asked Questions

Written by

Team WinPure

The WinPure Team shares official updates on our products, features, and company news. From new releases and enhancements to behind-the-scenes developments, this space keeps you informed on how WinPure continues to deliver secure, reliable, and innovative data quality solutions.