Executive search
Duplicate candidate records: review the merge before losing the history
A candidate-record merge approval checklist: establish identity, compare conflicting fields, inspect CRM behavior and verify the result. Includes a fictional comparison table.
Two candidate records appear to describe the same person. One contains recent research. The other holds an earlier conversation, a different contact address and notes from another assignment. Merging them could make the context easier to find, but only if they really belong together and the resulting record preserves the right meaning.
A candidate-record merge approval checklist should answer those questions before anyone confirms the operation. A similarity warning is a request for review. It is not proof of identity, a field-precedence decision or authority to combine information with different restrictions.
This guide covers a human review of a proposed merge. It does not instruct you to export the whole database, activate a deduplication service or assume that every CRM offers the same recovery options.
TL;DR
- Confirm that the records concern the same person using appropriate evidence.
- Review conflicting values, assignment history and restrictions before choosing the result.
- Read the actual CRM's merge behavior; identifiers and recovery differ across tools.
- Verify the authorized result against the approved comparison and retain unresolved exceptions.
Copyable merge-review record
Keep the comparison in the appropriate restricted working location. Include enough source context for a reviewer to understand the decision without reproducing every historical note.
| Review area | What to establish |
|---|---|
| Proposed pair | Original record identifiers and why the pair was flagged |
| Identity evidence | What supports a common identity, with conflicts and uncertainty visible |
| Field comparison | Conflicting values, their sources, dates and intended meanings |
| Context to preserve | Relevant activities, assignment links, notes and restrictions |
| Product behavior | Resulting identifier, precedence rules, associations and recovery limitations |
| Proposed outcome | Merge, keep separate or investigate further, with a reason |
| Authorization | Who may perform the operation and who has reviewed the consequential choices |
| Verification | Checks on the actual result, exceptions and follow-up ownership |
The checklist is an original operational aid. The firm's actual access controls and record-handling process determine who can authorize and execute a merge.
Establish identity before reconciling values
Begin with why the records appear related. A shared name or employer can be a clue, but colleagues with similar names and people who move between organizations can create misleading matches.
Review appropriate identifying evidence already available for the task. Compare source context and dates. Avoid gathering unrelated personal information simply to make the comparison feel more certain. If identity remains unresolved, keep the records separate and assign the next question.
Flo's duplicate-management documentation says its potential-match detection requires more than a name alone and provides a comparison before merging or dismissing. That is Flo's recruiting-product behavior. It illustrates the distinction between a potential match and a reviewed identity decision; it does not validate every suggested pair.
Do not let confidence in one field settle every other conflict. Confirming that two records belong to one person does not establish which title, employer or contact instruction is correct for the purpose at hand.
Read conflicting fields in their original context
Make a short comparison of the values that matter. Record what each field means, when the underlying event occurred and which source supports it. A current title and an accurately dated former title can both be useful.
The correct resolution may be to retain history rather than overwrite it. A role held during an earlier assignment should not be presented as the person's present role, but deleting it may make the earlier conversation difficult to interpret.
Use the candidate research-note method to distinguish evidence from attributed accounts and consultant interpretation. A merge should preserve those distinctions where the system supports them, rather than compressing them into an apparently authoritative summary.
If the fields use different definitions, resolve that first. One “owner” may mean the relationship contact and another may mean the consultant on a specific assignment. Choosing the latest value without understanding the field could replace two valid responsibilities with one misleading label.
Inspect the tool's actual precedence rules
Merge behavior is product-specific. Read current documentation for the relevant object and check what the preview will do with conflicting values, notes, files, activities and associations.
Attio's native merge documentation says a merge creates a new record with a new identifier and URL. It prioritizes the right-hand record's values when fields conflict. That behavior is specific to Attio; it should not be assumed for another CRM or presented as a Laisy feature.
A reviewer using that workflow would need to check the preview and consider references to the original identifiers. In another system, the surviving record may behave differently. Document the behavior you actually rely on rather than copying a generic “primary record wins” rule.
Identify anything the preview does not establish. You may need a system administrator to clarify a field type or association before proceeding. An unanswered question is a reason to pause that operation, not to infer a favorable result.
Verify recovery before relying on it
The existence of an export, history view or trash feature does not necessarily mean a merge can be undone faithfully. Ask what recovery can restore and what it cannot, using the current tool's documented process.
HubSpot's merge guidance states that merged records cannot be unmerged and that effects vary by object. Recreating a record is not the same as restoring the original state. That distinction matters before an authorized operator confirms a consequential merge.
Use whatever preservation steps the firm's approved process requires and the system supports. Keep any retained material appropriately restricted. A backup should not become an ungoverned parallel database or a reason to collect more information than the task needs.
If the consequences cannot be checked adequately, retain the proposed pair for investigation. The cost of another visible unresolved duplicate can be easier to manage than an incorrect combined identity whose history now looks coherent.
Check the boundary between accounts and systems
Records may exist in different accounts or tools for legitimate operational reasons. The right result may be a reviewed reference between contexts rather than a physical merge.
Google Contacts' merge help says contacts saved in different Google Accounts cannot be merged directly. That is a product boundary. It is not an instruction to copy contacts between accounts to work around the limit.
Before proposing a cross-system consolidation, establish the purpose, permission and applicable restrictions. A colleague's working address book and the firm's assignment record are not interchangeable just because both contain the same name.
Keep this review separate from candidate rediscovery. A record may become relevant to another search while still requiring a current-context and use check. Resolving a duplicate does not establish suitability for a new role or permission for fresh outreach.
Fictional example: three pairs, three different outcomes
This example is synthetic. The people, fields and pair counts are invented to demonstrate the review. They are not measured duplicate rates, a customer case or evidence of a cleanup saving.
A fictional researcher brings three proposed pairs to a reviewer. The first has consistent identity evidence but different dated role information. The second has the same name and company but conflicting identity evidence. The third appears to concern one person, yet an access restriction on a source note has not been clarified.
| Pair | Reviewed finding | Proposed outcome |
|---|---|---|
| A | Common identity established; earlier and current roles describe different periods | Prepare a merge comparison that preserves the historical context, then obtain authorization |
| B | Name similarity does not resolve conflicting identity evidence | Keep separate and record why the suggestion was rejected |
| C | Identity appears consistent; source-note restriction remains unresolved | Hold the merge and assign the restriction question to the appropriate owner |
Only pair A proceeds to a product preview. The operator checks the actual conflict rules and the proposed result before any change. Pairs B and C stay visible with different reasons; neither is counted as a completed merge.
These fictional outcomes do not estimate how many real duplicate suggestions should be merged.
The Government Data Quality Framework guidance recommends communicating data limitations. Applied to a proposed merge, keep the unresolved conflict beside the value it qualifies rather than concealing it in a tidy combined profile.
Google Contacts’ editing guidance explains its contact history and notes limits for older contacts. Check what history is actually available in the current tool; an absent edit record should not be replaced with an invented account of how the value arose.
Give the operator an executable decision
An approval should identify the exact records, consequential field choices and the condition under which the operation may proceed. “Clean up duplicates” is too broad to settle a specific conflict.
If the record changes after the comparison, review the affected choice again. A new note or instruction can alter what the operator should preserve. The approved comparison should describe the state the reviewer assessed.
Keep the roles clear. A consultant can clarify assignment meaning, while an administrator may be needed to explain system behavior. Neither should be assumed to hold every kind of authority simply because they can see the records.
Include unresolved pairs and approved-but-unexecuted work in an appropriate consultant handover. The incoming colleague needs to know what remains a proposal and what has actually changed in the system.
Verify the result before calling the cleanup complete
After an authorized operator performs the merge, compare the actual record with the approved outcome. Check identity, relevant field values, dated history, source context, assignment associations and any important access boundary.
Test the references that colleagues need for ongoing work. A changed identifier can affect a saved link or a handoff document. Verify those references rather than assuming the visible profile is the only consequence. Where the product processes related activities asynchronously, complete verification after the relevant processing has settled instead of signing off on an incomplete view.
If the result differs, record the discrepancy and follow the appropriate recovery or correction process. Do not make additional speculative changes to force the result to resemble the preview. Preserve enough evidence for the responsible person to investigate.
Count reviewed pairs, authorized operations, completed operations and accepted results separately. Those numbers describe different stages. A tool's “merged” notification does not establish that the working context survived correctly.
Test one review task alongside current tools
Choose a suitable proposed pair and agree the evidence, reviewer and acceptance check. Include the effort of investigating conflicts and verifying the result when assessing the method.
Use the pilot-success framework to specify the evidence, reviewer and acceptance check for one proposed pair.
Frequently asked questions
Are matching email addresses enough to approve a merge?
Evaluate the identity evidence and current context appropriate to the records. A matching field can be useful, but the reviewer still needs to examine conflicts and understand what the operation will combine.
Which record should be the primary one?
Check the actual tool's behavior and the approved field choices. A more complete or more recently edited record is not automatically correct in every field. Some products create a new record instead of retaining an original.
Can the team merge first and correct mistakes later?
Establish the consequences and recovery limits first. Some systems cannot unmerge records. An incorrect combination can affect context and associations beyond the most visible field.
Can AI decide which candidates are duplicates?
Treat a suggestion as a review input. Keep identity decisions, authorization and result verification with appropriate people. This guide does not claim an automatic identity or merge capability for Laisy.