Two profiles share an email address. One candidate opted out of contact. The other has a final interview tomorrow. An automatic merge turns both histories into a single, perfectly tidy record and nobody can explain which decision belongs to which person.
How should a recruiting CRM merge candidate records?
A recruiting CRM should merge candidate records only after checking multiple identity signals, conflicting values, source reliability, contact restrictions, mandate relationships and recruiter ownership. Uncertain matches need human review; approved merges need an audit record, retained provenance and a controlled way to undo a mistake.
A duplicate suggestion is not proof of identity. Shared email inboxes, recycled phone numbers, married names, transliteration and multilingual resumes all create plausible matches that deserve different decisions.
Design merge rules around harm. Sending the wrong candidate’s private interview notes to a client or reopening an objection is more serious than leaving a low-confidence duplicate unresolved for another day.
Separate duplicate detection from identity confirmation
Detection identifies a possible relationship; identity resolution decides whether the records represent the same person.
A matching email can be useful evidence, but it is not universally unique or permanent. Consultants encounter shared agency inboxes, temporary work addresses, family phone numbers, names that change and profiles created from incomplete CVs. Rank evidence by reliability rather than treating every matching character as equally meaningful.
Capture why the system suggested a match: identical verified email, matching professional identifier, overlapping employment history, same phone or similar name and location. Show disagreements as prominently as agreements. A confident reviewer should be able to explain the decision without appealing to an opaque percentage.
Set match thresholds and mandatory review cases
Automatic action is appropriate only for explicitly defined, low-risk situations with a safe reversal path.
Write separate outcomes for exact confirmed identity, likely match, weak resemblance and direct contradiction. A verified identifier plus supporting history may justify a proposed merge; two common surnames and the same city should not. Treat client-facing submissions, confidential mandates and sensitive notes as reasons to require a person regardless of score.
Block automation when one profile has a contact objection, a different legal name without supporting context, contradictory current employers, inconsistent birth-related information or separate active client processes. Thresholds should be reviewed with real agency examples, not borrowed from a generic CRM demonstration.
- Require more evidence when a match could expose confidential client or candidate information.
- Send conflicting verified contact details to manual review.
- Never let enrichment or parser freshness overrule an explicit candidate objection.
Choose field survivors without deleting provenance
The newest value is not automatically the most accurate value.
Define a precedence order per field. A number confirmed by the candidate yesterday can replace a stale imported number; a guessed location from a resume parser should not replace a recruiter-confirmed preference. Keep the source, verification date and prior meaningful value when the business needs to explain the change.
Separate facts from opinion. “Worked in Munich until 2024” is a dated factual statement; “unlikely to relocate” may be a consultant’s outdated interpretation. The UK ICO’s accuracy guidance distinguishes current facts, historical records and opinions, which is especially relevant when several recruiter histories are combined.
Carry submissions, ownership and communication chronology
Merging a person must not silently rewrite who handled each mandate or what the client was told.
Preserve links to applications, submitted shortlists, interview feedback, documents and client-specific restrictions. Attach each event to its original author and mandate. If two consultants legitimately own different relationships, record the conflict and decide future ownership rather than attributing all historic work to the last editor.
Check whether a client is off-limits for one mandate, whether a candidate has already been represented and whether a recruiter promised not to contact a particular employer. The merged record should make those facts visible without combining confidential notes across desks that are not authorised to see them.
Give objections and restrictions priority
A suppression, objection or withdrawn permission must survive record consolidation and immediately control downstream workflows.
Test every contact channel separately. One candidate may object to promotional email while still discussing a specific live mandate by phone; another may withdraw from all future search activity. Preserve the purpose, scope, timestamp and decision rather than collapsing everything into a single misleading “consented” flag.
Propagate the protective state before enabling sequences, exports, client sharing or enrichment. If two records disagree, pause outreach until the recruiter or privacy owner resolves the conflict. Silence, a blank field or an old imported tick is not permission to assume the broader interpretation.
When a merge combines one “contact allowed” record with one objection, the safe immediate state is review—not contact allowed.
Keep an audit trail and a practical split path
Every material merge should preserve the reviewer, identity evidence, affected objects and changes needed to unwind the decision.
Record both original identifiers, the surviving identifier, time, reviewer, match rationale, conflicting fields, moved relationships and applied restrictions. Restrict access to the audit data. The purpose is to explain a correction and respond to a candidate challenge, not to create an unlimited hidden archive.
Test a controlled unmerge before rollout. The agency needs to restore the correct notes, documents, mandate links, restrictions and ownership if two real people were combined. Where reversal is technically incomplete, prevent automatic merging and require a higher-confidence approval threshold.
Review parser-driven duplicates at the source
A resume parser can create duplicate candidates when it extracts different spellings, addresses or employers from separate versions of a CV.
Compare parser extraction with the source document before treating a mismatch as evidence of two people. Accents, double surnames, different language versions and changed contract titles are common. Track whether a duplicate originated from parsing, manual entry, imported lists or an integration so the team can fix recurring causes.
Measure useful operational signals: review queue age, confirmed duplicates, false matches, prevented outreach, reversed merges and missing candidate history. Avoid claiming that fewer records automatically means better quality. A smaller database produced by unsafe merges is worse than a larger one with clear review flags.
Candidate merge decision matrix
The right response depends on evidence and potential harm, not on whether two CRM rows look similar at first glance.
| Candidate record situation | Recommended decision | Evidence to keep |
|---|---|---|
| Verified identifier and matching history | Propose a controlled merge after policy checks | Verified identifier, reviewer and source record IDs |
| Same name and broad location only | Keep separate or request more evidence | Similarity reason and unresolved ambiguity |
| Contradictory current employers | Pause and ask a recruiter to inspect both timelines | Original employment claims and source dates |
| Objection on either record | Suppress outreach until scope is resolved | Objection purpose, timestamp and decision |
| Two confidential client submissions | Require mandate-aware human approval | Client boundaries, owners and submission history |
| Parser spelling or language variation | Check the underlying CV before deciding | Document reference, extracted values and correction |
| Incorrect completed merge | Split records and restore affected relationships | Merge audit, incident owner and remediation outcome |
Limits of automatic identity resolution
No similarity model can guarantee that two candidate records belong to the same individual. Matching confidence is a prioritisation aid, not a substitute for reviewing contradictions, confidentiality and the person’s own corrections.
Audit logs and recovery paths vary between software setups. A boutique executive-search team should choose a workflow it can actually supervise; agencies needing high-volume, fully automated identity resolution may require different tooling and dedicated operations support.
Candidate record merge FAQ
These are the decisions that matter after duplicate detection has found a possible match.
Is one matching email address enough to merge candidates?
Not always. Check whether the address is verified, unique in context and supported by other identity evidence. Shared inboxes, reused addresses and imported errors can connect two different people.
Which value wins when candidate records disagree?
Apply a documented field-specific precedence rule using verification, source reliability, purpose and recency. Candidate-confirmed facts generally deserve more weight than unreviewed parser or enrichment output.
What happens to an opt-out when records are merged?
Keep the restriction and apply the safest appropriate contact state while its scope is reviewed. Never allow a blank field or a second historic record to erase an objection or withdrawn permission.
Can a wrong candidate merge be reversed?
Only if the software and agency procedure retain enough original identifiers, relationships, notes and permission states. Test the split process before allowing automatic or bulk merges.
Official sources on accuracy and candidate rights
Use the accuracy and rights principles below when designing review, correction and accountability controls for a recruitment database.
- UK Information Commissioner’s Office: Accuracy principle and correcting personal data
- European Data Protection Board: Respect individuals’ rights
- EUR-Lex: General Data Protection Regulation, including accuracy, storage limitation and data-subject rights
Improve candidate database quality
Evaluate AI CV parser accuracy · Set candidate ownership rules · Audit candidate relationship records · See candidate data enrichment
Keep the candidate history that actually matters
Inspect real agency records, relationship ownership and correction paths before deciding how duplicates should be handled in your recruiting workflow.
See Yena data enrichment and review