A duplicate suggestion is a candidate for review. It is not proof that two records represent the same person or organization. Similar names can hide different legal entities; changed email addresses can belong to the same person.
The July HubSpot roundup added duplicate similarity scoring. Treat the score as a way to organize investigation, not as permission to merge automatically.
Define identity before setting a review threshold
Agree which identifiers matter for the object and business model. A company domain may be shared by multiple entities or subsidiaries. A customer identifier may be reliable in one source but absent from prospect records.
Write down disqualifying evidence as well as matching evidence. Two active contract references for different legal entities should prompt investigation even when the display names look identical.
Use a structured review queue
Prioritize candidates using confidence, operational impact and the availability of evidence. Records involved in active deals or service commitments may deserve a more careful review than inactive test data.
For each candidate pair, record the identity decision, preferred surviving record and any fields requiring resolution. Keep uncertain cases separate instead of forcing a binary answer to complete a cleanup count.
Decide which values should survive
A matching identity does not resolve every property conflict. One record may contain the current owner while the other carries a verified external identifier or relevant preference history. Review the fields that drive routing, communication and reporting.
Use the field-ownership model to determine authoritative sources. Recency alone is not a sufficient rule for every property.
Inspect operational dependencies
- Active opportunities and customer relationships.
- Workflow enrollment and scheduled actions.
- External references used by integrations.
- Associations, activity history and reporting classifications.
- Communication eligibility and ownership exceptions.
Validate actual merge behavior with approved representative records. Do not infer what will happen to every association or workflow from the fact that the merge interface accepts the action.
Measure cleanup quality
A useful cleanup reduces ambiguity and recurring duplicate creation. Counting merged records measures activity. Also review incorrect matches, unresolved candidates and the sources that continue creating duplicates.
For an illustrative prevention measure, inspect whether forms and integrations use the agreed identifiers consistently. Correcting the creation process can matter more than running another large merge batch.
Continue with the CRM data-quality guide and audit checklist.
Resolve the identity rules behind duplicates.
I can review matching logic, property ownership and the processes creating recurring data-quality problems.
Explore a data-quality audit →