Similar Words Matching
Similar words matching is a record-matching method that compares the words inside a value and treats two records as duplicates when they share enough of them. Instead of comparing "Acme Logistics" and "Logistics by Acme" as two whole strings, it compares the words Acme and Logistics, and finds them in both.
The method can work in a few ways. Any-order matching pairs values that share words wherever those words appear. Position-based matching rests the match on the first or last words of each value. A setting for how many words must match controls how strict the comparison is. Some tools also add a list of ignored terms, such as Inc. or LLC, that are left out of the comparison, so words that appear in many names neither create a match nor block one. Similar words matching suits fields where the same name is often written in a different word order or with extra words, such as company names and full names.
See how Dedupely matches names by shared words in your CRM
What Is Similar Words Matching?
Similar words matching is a way to compare values by the words they share rather than by their exact characters. It's built for fields where the same thing gets written in a different shape: reversed, shortened or padded with extra words. A typical example: one record has the full name "Robert Chen" and the company "Acme Logistics Inc.," and another has "Chen, Robert" and "Logistics by Acme." An exact match misses both fields. A shared-words comparison pairs the full names, because both contain Robert and Chen, and pairs the company names, because both contain Acme and Logistics once "Inc." and "by" are set aside. With both fields in agreement, the two records point to one person at one company.
There are three common approaches, and each one has its own strength. Any-order matching catches word order duplicates, such as "Chen, Robert" and "Robert Chen," and suits names that get reversed or rearranged. First-word matching pairs values that start the same way, such as "Acme Logistics" and "Acme Logistics Group," so it handles names that pick up extra words at the end. Last-word matching pairs values that end the same way, such as "Dr. Ana Ruiz" and "Ana Ruiz," which helps with titles and middle names. Each approach can be made stricter or looser by setting how many words must match.
The method applies to name fields across CRMs: company names on companies in HubSpot, accounts in Salesforce and organizations in Pipedrive; lead names in Close; and full names on contacts in Nimble and Google Contacts.
How Similar Words Matching Works
Similar words matching is one of the comparison methods used in the matching stage of deduplication. It decides whether two values in a field count as a match, and it runs alongside the other fields in a rule.
- Split each value into words. "Logistics by Acme" becomes three words: Logistics, by and Acme. Some tools also even out capitalization or punctuation at this point.
- Set aside ignored terms. Words that don't identify anyone, such as legal suffixes like Inc. or Ltd. and generic words like "Group," can be left out, so they neither create a match nor block one.
- Compare the words. In any-order matching, the words are compared regardless of where they sit in each value. In position-based matching, only the first or the last words are compared.
- Count the shared words. A setting decides how many words must match. Requiring one shared word finds more pairs, and also more false matches. Requiring two or more is stricter.
- Combine and group. The field's result is combined with the other fields in the rule, and records that meet the rule are grouped into duplicate sets for review or merging.
Because shared words are a looser signal than an identical value, the method finds near-duplicates that exact matching misses, such as reversed or padded names, and it works best when a second field backs up the match.
Why Similar Words Matching Matters for CRM Teams
Name fields are where many CRM duplicates hide. A company can enter the CRM as "Acme Logistics" from a form, "Acme Logistics, Inc." from an import and "Logistics by Acme" from a rep's notes. To exact matching, those are three different companies, so their contacts, deals and activity stay split across three records. Reports count the account three times, ownership rules can assign it to different reps, and in CRMs that bill by record count, each extra record is paid for.
Full names behave the same way. A contact saved as "Robert Chen" in one record and "Chen, Robert" in another looks like two people to an exact comparison, and so does a contact whose second record adds a middle name or a title. Without word-based comparison, these duplicates stay in the CRM until someone notices them by hand.
The method has a cost when it's set too loosely. Many companies share common words, such as "First," "National" or "Solutions," and many people share a last name. A rule that pairs records on one shared word can group unrelated companies or people into one set, and merging that set mixes their histories and contact details. How many words must match, which words are ignored and which other fields support the match decide whether similar words matching finds real duplicates or creates false ones.
Getting Started With Similar Words Matching
Start with the fields where word order and extra words vary most, often company names and full names. Look through a sample of those values first to see how they differ: reversed names, legal suffixes, added words like "Group" or "Holdings," or titles in front of a person's name. What you find there points to the right approach and the words worth ignoring.
When you dedupe company names, begin strict. Require two or more shared words, ignore legal suffixes and filler words, and pair the name with a second field, such as website domain, phone number or city. Check the duplicate sets the rule finds before merging anything, then lower the number of words only if real duplicates are still being missed.
For people, match names regardless of word order on full-name fields, together with a second field such as email or company. A shared first or last name is a weak signal on its own, so it should never be the only field in a rule. Any pair that turns out to be two different records should be rejected, so it stays out of future results.
In Dedupely, Similar Words is one of the Match Options in a Duplicate Search: it compares shared words in Any Order, or the First or Last words, with a Number of Words to Match setting that decides how strict the comparison is. Ignored Terms, up to 50 per field, keep words like Inc. or LLC out of the comparison.
Key Takeaways: Similar Words Matching
Similar words matching pairs records by the words their values share, so names written in a different order or with extra words can still match. It works in any order or on the first or last words, with a set number of words that must match. It suits company names and full names, and works best with ignored terms and a second field to keep false matches low. In Dedupely, Similar Words is one of the Match Options in a Duplicate Search, with Any Order or First or Last words and a Number of Words to Match setting. Ignored Terms leave out words like Inc. or LLC, and Rejected Matches keep false pairs out of future results.
Frequently Asked Questions About Similar Words Matching
When should you use similar words matching instead of exact or fuzzy matching?
Use similar words matching when the same name is written with its words in a different order or with extra words. Exact matching needs identical values, so it misses "Acme Logistics" and "Logistics by Acme." Fuzzy, or phonetic, matching catches names that sound alike but are spelled differently, such as Sean and Shawn, but reordered or added words can still keep two values apart. Similar words matching fills that gap by comparing the words inside each value. It fits company names and full names, and it's a poor fit for unique identifiers like email addresses, where exact matching is stronger. Many rules combine the methods, with one strong field on exact matching and the name field on a looser one. In Dedupely, each field in a Duplicate Search gets its own Match Option: Exact, Similar, Fuzzy, Similar Words, Domain Root or None. Similar Words compares shared words in Any Order, or the First or Last words.
How do deduplication tools handle duplicate accounts with name variations?
Deduplication tools handle name variations by comparing account names with looser methods and confirming the match with a stronger field. Account and company names vary through legal suffixes, abbreviations, word order and typos, so an exact comparison misses many of them. Word-based matching handles reordered or extra words, ignored-term lists strip suffixes like Inc. or Ltd., similar matching catches typos, and domain-based matching compares the company's website or email domain. Some tools also use AI duplicate matching to compare many field combinations at once. The approach is the same whether the accounts sit in Salesforce, HubSpot or Pipedrive: loosen the name comparison and back it with a field like domain, phone or city. In Dedupely, Similar Words compares shared words in Any Order, or the First or Last words, and Ignored Terms leave out up to 50 words per field. Domain Root matches the core domain of an email or URL, and Dedupely Intelligence, Dedupely's AI duplicate matching, checks companies for field combinations your searches don't cover.
Which fields work best with similar words matching?
Similar words matching works best on multi-word name fields, such as company names, full names and deal names, where the same name gets reordered, shortened or padded with extra words. Company names pick up legal suffixes and words like "Group," full names get reversed or gain a middle name or title, and deal names often add a year or a word like "Renewal." Comparing shared words handles all of these. It adds little on single-word fields, where there's only one word to compare, and it's a weaker fit for identifiers like email addresses, phone numbers and IDs, where exact matching or a format-aware method is stronger. Pairing a name field with one of those identifiers often gives the best balance. In Dedupely, Similar Words is one of the Match Options, so it can sit on a company name or full-name field next to an Exact email field in the same Duplicate Search. Nicknames such as Robert and Bob are a separate "Match nicknames" setting on first-name fields, which works with whichever Match Option the field uses.
How do you match names regardless of word order, such as "Chen, Robert" and "Robert Chen"?
Compare the name by its words in any order, so the position of each word doesn't matter. Full names get reversed when one system stores "last name, first name," and company names get rearranged when people type them from memory. Any-order matching splits each value into words and checks how many they share, so "Chen, Robert" and "Robert Chen" pair on both words. Setting how many words must match keeps it from pairing everyone named Chen. Position-based matching is an alternative when names share a stable start or end: first-word matching pairs "Acme Logistics" with "Acme Logistics Group," and last-word matching pairs names with and without a title. Back the name with a second field, such as email or phone, to keep the match tied to one record. In Dedupely, Similar Words compares shared words in Any Order, or the First or Last words, and the Number of Words to Match setting decides how many of them must match.
How do you stop word-based matching from pairing different companies?
Require more than one shared word, ignore words that don't identify a company, and confirm the match with a second field. Common words like "First," "Global" or "Solutions" appear in many unrelated company names, so a rule that pairs on one shared word can group "First Solar" with "First Bank." Raising the number of words that must match, adding those common words to an ignored list and pairing the name with a website domain, phone number or city all cut false matches. Check a sample of each rule's results before merging, and reject the pairs that aren't duplicates. In Dedupely, the Similar Words Number of Words to Match setting controls how many words must match, and Ignored Terms leave out up to 50 words per field. If a search compares three fields, all three must meet their criteria. Safety Settings are on by default and hold back matches with more than 20 records, and Rejected Matches keep false pairs out of current and future Duplicate Searches.
