How duplicate matching works
The signals DeDupe compares to decide two records are the same.
DeDupe compares records across several signals and combines them into a single confidence score. No single field decides a match on its own.
Matching strategies
When you run a scan you choose which strategies to use. Each looks at a different signal:
Companies
- Similar Domain — compares website domains, tolerant of typos
(
acme.comvsacme.co). - Similar Name — fuzzy company-name comparison ("Acme Inc" vs "Acme, Inc.").
People
- Similar Email
- Exact LinkedIn — matching LinkedIn URLs.
- Exact Phone — normalized to a common format before comparing.
- Similar Name — fuzzy person-name comparison.
- Name + Company — the same name at the same company.
A pair is flagged if any enabled strategy fires. Turning on more strategies finds more potential duplicates; turning on fewer keeps results tight.
Custom rules
Beyond the built-in strategies, admins and operators can write custom matching rules to ignore middle names, normalize fields, treat values as equivalent (e.g. "Inc." = "Incorporated"), or exclude records that shouldn't match.
Why some obvious duplicates rank lower
If two records share only a name and nothing else, DeDupe stays cautious — common company and person names produce false matches. Filling in more complete data (domains, emails, phone numbers) improves match quality over time.
Next
See Understanding confidence scores for how the signals combine into the High / Medium / Low bands you see in the queue.