CRM duplicate records: build a data-quality process that lasts

Distinguish duplicates from legitimate relationships

Two records with similar names are not necessarily the same person or company. A parent organisation can have separate buying entities, and several people can share a generic company email address. Decide what each CRM record is intended to represent before applying broad cleanup rules.

For a fictional industrial supplier, one customer has two branches with different delivery and billing arrangements. Combining them into a single undifferentiated account might simplify the list while damaging the sales process. The cleanup brief should describe the intended account hierarchy and contact relationships, not just a desired reduction in record count.

Map the entry points

List manual entry, forms, imports, integrations and synchronisation with other systems. Identify which source creates records and which updates them. A team can merge yesterday’s duplicates while an unchanged import produces them again tomorrow. Examine a small sample from each entry route.

Give every integration a documented identifier strategy. Where records have stable external IDs, retain the mapping. Do not rely on a display name to reconnect records after migration. Include the person who maintains the integration in the cleanup process so the preventative change reaches the actual source.

Separate matching from the action taken

Salesforce’s matching-rule guidance describes fields and comparison logic for identifying possible duplicates. Detection is a different decision from whether to block creation, warn a user or merge records. Ask the supplier to demonstrate the configured outcome.

HubSpot’s deduplication documentation explains identity behaviour and import identifiers. Check the specific creation path and object type; do not assume every connector, form and import follows one identical rule. Use test records to inspect the behaviour before changing a live import.

Define what survives a merge

Keep a review trail for ambiguous cases and use the product’s documented merge behaviour. Export or otherwise preserve the appropriate pre-change evidence under the organisation’s access controls. A spreadsheet of names alone may not contain enough information to reconstruct lost associations.

Test false matches as well as missed matches

Prepare examples of the same person using a changed address, two people with similar names, related companies sharing a domain and a repeat import with stable IDs. Ask a knowledgeable reviewer to label them before testing. This exposes whether the proposed rule is too aggressive or too weak.

Record the reason for each incorrect match. Tightening a rule can reduce false merges while leaving more records for human review. That tradeoff should be explicit. A headline saying most duplicates removed is less useful than knowing that important customer relationships remain intact.

Prevent recurrence at the source

Use supported unique identifiers and validation where the business process permits them. Teach staff how to search for an existing record and how to report a suspected duplicate. Make the correct path easier than creating another record to bypass a confusing search result.

Review integration retries and import settings. An update that accidentally becomes a create operation can produce a growing problem without any user typing a duplicate. Keep a small exception queue with an owner and review date. Large periodic cleanups are harder to investigate than a steady flow of understood exceptions.

Measure the quality of the working record

Inspect a sample of active accounts after the cleanup. Can the salesperson find the current contact, relevant deal and next action? Have the related systems kept the same identity mapping? These checks connect data quality to usable work instead of cosmetic record counts.

Use the lead-routing guide to review ownership after identity changes. The HubSpot CRM and Salesforce profiles provide starting points for product research. Assign ongoing responsibility for the rules, imports and unresolved cases so the CRM stays useful after the initial project.

Frequently asked questions