Candidate Deduplication: Same Person, Different Resumes
How to find duplicate applicant profiles without merging different people. Compare contact signals, preserve job history and review evidence.
Two applications can belong to the same person. Two matching details can also belong to different people. The costliest mistake is confusing a possible match with a verified identity.
Why duplicate detection is more than a matching formula
A candidate may apply with a personal email, then return with a work address and a revised resume. Another applicant may share a common name, employer or household phone number. Blind merging can mix private notes, past decisions or consent records.
Current ATS products handle this differently. Ashby documents duplicate warnings and manual merge previews. Greenhouse offers configurable auto-merge rules. This makes a universal one-field merge rule a poor default.
Rank identity clues by strength
| Signal | How to interpret it |
|---|---|
| Same verified profile URL | Strong clue; verify ownership and other conflicts |
| Same email | Useful clue; aliases and shared inboxes exist |
| Same phone | Possible clue; numbers may be reused or shared |
| Name and employer | Weak clue; never enough alone |
| Similar resume wording | Content similarity, not identity proof |
A conflicting unique profile link deserves manual inspection even when another field matches. If the original source is incomplete, label the case unresolved rather than guessing.
What must survive any confirmed merge
Preserve every resume version, original identifiers, application for each vacancy, candidate communications, timestamps, data retention obligations and the origin of manual notes. Keep job-specific evaluations attached to the job, not permanently written into a universal person score.
The ability to undo a mistaken merge matters. Ashby and Greenhouse both document limitations on reversing merges. Test a recovery procedure before enabling any automatic rule, especially after bulk imports.
Compare two example profiles
Two synthetic examples start with a shared email and phone. Change any field to see how the evidence changes. The tool never confirms identity or combines data.
A safe review procedure
- Review normalized identifiers and the original documents side by side.
- Check for contradictory high-strength signals before recommending a merge.
- Confirm that both records belong to the same person through authorized evidence.
- Preview what will be retained, including applications and retention dates.
- Record the reviewer, decision and recovery path; only then apply a merge in the ATS.
The comparison tool on this page does not prove identity and does not write to any database. Keep personal contact information out of demonstration environments and use synthetic data for testing.
Measure false merges as well as duplicate finds
Track suspected duplicates, confirmed matches, confirmed distinct profiles, reversals and unresolved cases. Review a sample of automatic merges separately. A high number of merges is not inherently a quality outcome if one accidental merge can expose or lose candidate history.
Treat deduplication as a data-quality control with human accountability. It should make records more trustworthy without creating new privacy risks.
One practical test uses deliberately confusing synthetic pairs: two people with the same name, one person using two addresses, a shared phone number and a reused resume template. Verify that none is silently combined. Then inspect the merge preview and the documented recovery process for the ATS actually in use.