Fuzzy Matching
Fuzzy matching is a way of matching values that are close but not identical, such as names with a typo or names that sound alike. Phonetic matching is the most common type for names: it compares how values sound instead of how they're spelled, so "Sean" and "Shawn" can pair. In some tools, fuzzy matching is also the name of the phonetic method itself.
In deduplication, fuzzy matching finds the records that exact matching misses because one value differs slightly from the other. Different methods measure "close" in different ways: some count the letters that differ, some compare how the values sound, and some compare the words inside them. Because every fuzzy method allows some difference, it also pairs more records that aren't duplicates, so it's often used on one field while a stronger field, such as email or phone, confirms the match.
See how Dedupely catches sound-alike names and typos in your CRM
What Is Fuzzy Matching?
Fuzzy matching is the family of record-matching methods that pair values that are close without being identical. Two kinds of difference are the most common in CRM data, and each has its own approach.
- Typos and spelling slips. Edit-distance matching counts how many letters must be added, removed or changed to turn one value into the other. "Catherine" and "Cathrine" are one change apart. Its strength is that it catches mistakes in any kind of text, names or not.
- Names that sound alike. Phonetic matching turns each value into a code based on how it sounds and pairs values with the same code. "Sean" and "Shawn" share one, even though they're two changes apart. Its strength is that it catches alternate spellings that look quite different but sound the same.
Take "Kathryn Smyth, Acme" and "Catherine Smith, Acme." The last names are one letter apart and sound the same, so either approach pairs them. The first names are four changes apart, which is more than many edit-distance rules allow. Phonetic algorithms also differ here: Soundex, one of the oldest, keeps the first letter as written and keeps C and K apart, while newer ones, such as Metaphone, encode the first sound and can pair the two.
Fuzzy matching works on contacts in HubSpot, Nimble and Google Contacts, people and organizations in Pipedrive, accounts and leads in Salesforce, and leads in Close. It's used most on names, and it's least useful on values that should match exactly, such as email addresses or IDs.
How Fuzzy Matching Works
Fuzzy matching is a comparison step inside the matching stage of deduplication. It works in a few steps:
- Prepare the values. Values are often evened out before they're compared, for example by ignoring capitalization or punctuation, so the method only looks at differences that matter.
- Measure how close they are. Edit-distance methods count the changes between two values and often turn that count into a score. Phonetic methods convert each value into a sound code, so two values either share a code or don't. Word-based methods compare the words a value contains, so word order or an extra word doesn't break the match.
- Decide what counts as a match. A score-based method pairs values above a set threshold, and a phonetic method pairs values with the same code. A looser threshold finds more duplicates and more false matches.
- Combine it with other fields. In rule-based matching, every field in a rule has to match, so fuzzy matching on a name is often paired with a stricter field, such as email, phone or company.
- Group and review. Records that meet the rule are grouped into duplicate sets, and a sample is checked before anything is merged.
Each method has blind spots. Phonetic codes don't capture every typo, and edit distance doesn't know that two very different spellings sound the same. Neither one pairs a nickname with a full name, such as Bob and Robert, because those differ in both spelling and sound. That's why tools often offer several fuzzy methods side by side.
Why Fuzzy Matching Matters for CRM Teams
Names are often among the least consistent values in a CRM. They arrive from web forms, imports, event lists, sales tools and syncs, and they're often typed by someone who heard the name on a call. The same person can end up as "Sean" in one record and "Shawn" in another, or as "Cathrine" after a quick form fill. Exact matching treats those as different people, so the duplicates stay in the CRM.
Each missed duplicate has a cost. It inflates contact counts and reports, splits emails, calls and notes across two records, and can put the same customer in front of two reps. In CRMs that bill by contact or record count, it's also a record paid for twice. Finding them takes sound-alike name matching or another fuzzy method, because the names never match exactly.
The same leniency brings risk. A fuzzy rule on name alone pairs different people who happen to share a similar name, such as Jon Smith and John Smith at two different companies. If that set is merged, two customers' histories, owners and contact details are mixed into one record, and in many systems a merge is hard or impossible to reverse. That's why fuzzy matching works best when a stronger field confirms the match, and why its results get a closer look before they're merged.
Getting Started With Fuzzy Matching
Use fuzzy matching after the exact matches are merged, so the results show mostly near-duplicates. Pick the method for the kind of difference you see in your data: phonetic matching for names spelled by ear, edit-distance or similar matching for typos, and word-based matching for company names with extra or reordered words.
Keep a strong field next to every fuzzy one. A fuzzy first and last name with an exact email, phone or company pairs the same person spelled two ways, without pairing everyone with a similar name. If a field is often blank or full of placeholder values, leave it out instead of loosening the rest of the rule.
Check a sample of the new duplicate sets before you merge any of them, and reject the pairs that turn out to be two different records, so they stay out of future results. Change one field at a time, so you can tell which change found the new duplicates and which one let false matches in.
In Dedupely, Fuzzy is the phonetic Match Option: it uses a phonetic algorithm to pair alternate spellings, such as Sean and Shawn, while the Similar Match Option handles typos and mixed or missing letters. Each field in a Duplicate Search gets its own Match Option, so a name field can use Fuzzy while email or phone stays on Exact or Similar.
Key Takeaways: Fuzzy Matching
Fuzzy matching pairs values that are close but not identical, which lets it find duplicates that exact matching misses. Edit-distance methods catch typos, phonetic matching catches names that sound alike, and word-based methods catch reordered or extra words. Because every fuzzy method allows some difference, it works best alongside a stronger field and with a review before merging. In Dedupely, Fuzzy is the phonetic Match Option for alternate spellings such as Sean and Shawn, and Similar handles typos. Each field in a Duplicate Search gets its own Match Option, Rejected Matches keeps false pairs out of future results, and Merge Rules decide what each merge keeps.
Frequently Asked Questions About Fuzzy Matching
How is fuzzy matching different from phonetic matching?
Phonetic matching is one type of fuzzy matching, and in some tools the two names mean the same thing. Fuzzy matching covers every method that pairs values that are close but not identical. Edit-distance methods count the letters that differ, so they catch typos like "Cathrine" for "Catherine." Phonetic methods compare how values sound, so they catch alternate spellings like "Sean" and "Shawn," which look more different than a typo but sound the same. Many tools offer both, and some use "fuzzy" as the label for the phonetic method specifically. When weighing fuzzy vs phonetic matching in a tool, check which method its fuzzy option uses, because that decides which duplicates it finds. In Dedupely, Fuzzy is the phonetic option: it uses a phonetic algorithm for alternate spellings, such as Sean and Shawn. Typos and mixed or missing letters are handled by the Similar Match Option, and each field in a Duplicate Search can use either one.
How do you match names that sound alike but are spelled differently?
Use phonetic matching on the name fields, and pair it with a stronger field such as email, phone or company. Phonetic matching turns each name into a code based on how it sounds, so "Sean" and "Shawn," or "Smith" and "Smyth," get the same code and pair even though their spelling differs. Phonetic algorithms vary: some keep the first letter as written, so "Catherine" and "Kathryn" stay apart, while others encode the first sound too. Comparing first and last name separately often works better than comparing a full-name field, where a middle name or initial can break the match. Nicknames, such as Bob and Robert, don't sound alike, so they need a nickname list instead. In Dedupely, the Fuzzy Match Option uses a phonetic algorithm for alternate spellings such as Sean and Shawn. For nicknames, first-name fields have a "Match nicknames" setting that works with any Match Option, including Fuzzy, so Robert and Bob can pair.
How do you use fuzzy matching to find CRM duplicates without false matches?
Combine the fuzzy field with at least one strict field, and review the results before merging. A fuzzy rule on name alone pairs different people with similar names, while fuzzy name + exact email, or fuzzy name + exact company + phone, points to one person. This applies in any CRM, whether the records are contacts in HubSpot, leads in Salesforce or people in Pipedrive. Keep fuzzy methods off fields that many records share, such as city or job title, and check a sample of each new rule's sets. Reject the pairs that are two different records, so they don't come back. In Dedupely, the KB recommends combining 2 to 4 key fields in a Duplicate Search, and if a search compares three fields, all three must meet their criteria. Rejected Matches removes a set from current and future searches. Safety Settings are on by default and hold back matches with more than 20 records or placeholder data, and you can send uncertain sets to Manual Review.
Which CRM fields should you use fuzzy matching on?
Use fuzzy matching on names, and a different method on most other fields. First and last names vary through typos and alternate spellings, so phonetic or edit-distance matching suits them. Email addresses and IDs should match exactly, because a "close" email often belongs to someone else. Phone numbers differ mostly in formatting and country codes, so a method that strips formatting works better than a phonetic one. Company names often carry suffixes like Inc. or Ltd., or list their words in a different order, so word-based matching fits them, and websites compare best on the root domain. In Dedupely, Fuzzy fits name fields, where alternate spellings such as Sean and Shawn are common. Similar catches typos and, on phone numbers, strips formatting and country codes and compares the last seven digits. Similar Words compares shared words, which suits company names, Domain Root compares the core domain of an email or URL, and Exact suits email addresses. Up to 50 Ignored Terms per field can be added on Fuzzy and every other option except None.
Why does fuzzy matching miss some duplicates you can see?
Often, the method doesn't measure the kind of difference the two records have. A phonetic method compares sound, so a typo that changes how a name sounds, such as "Vatherine" for "Catherine," can produce a different code. An edit-distance method counts letters, so two spellings that sound the same but differ in several letters can fall outside its threshold. Neither one pairs a nickname with a full name. In a rule with several fields, one field that doesn't meet its criteria, such as a blank email, also breaks the match. To find the cause, compare the two records field by field and switch the field that differs to a method built for that difference. In Dedupely, Fuzzy compares sound, so switch the field to Similar to catch typos and mixed or missing letters, or turn on "Match nicknames" on first-name fields, which works with any Match Option, including Fuzzy. Typo allowance, one of the search's Enhancements, prevents small differences from blocking matches when stricter matching is used, and it applies to any Match Option.
