How to merge duplicate CRM records without losing data

Direct answer

Merging is destructive: in most CRMs, once two records are merged you cannot cleanly split them again. So the work is in the preparation. Define which fields identify a duplicate, define which record survives and which value wins in a conflict, export a backup, run the merge on a small batch, send anything ambiguous to a person, and only then scale up. Then fix the source of the duplicates, or you will be doing this again in three months.

1. Choose match keys deliberately

A match key is the combination of fields that says two records are the same real-world person or company. Different keys have different risks:

• Exact email address: the strongest key for contacts, though shared mailboxes such as info@ or sales@ can belong to several people
• Phone number after normalising format: useful, but a switchboard number is shared by everyone at a company
• Name plus company: helpful as a secondary signal, unsafe alone because common names collide
• Company website domain or registered number for companies: strong, but watch for subsidiaries and free-mail domains that are not a company at all
• Fuzzy name matching: use it to suggest candidates for review, never to merge automatically

2. Write survivor-record rules

When two records merge, one is kept and the other folds into it. Decide the rule before you start, and write it down so it is not decided ad hoc by whoever is cleaning that day. Typical rules:

• Keep the record with the most recent activity, because that is usually the one people are actually working
• Or keep the oldest record, to preserve the original creation date and source; choose one and use it consistently
• Prefer the record that has an open deal or an assigned owner
• Prefer the record created by a human over one created by an import or a form, if the human record has cleaner data
• Never let a blank value overwrite a filled one

3. Back up first

Before any bulk merge, export the affected objects with their record IDs, owners, creation dates and every field you care about. Also export associations, meaning which contacts belong to which company and deal, since these are often not part of a plain contact export. Save the file somewhere versioned with a date. Check what your platform offers for restoring merged records; do not assume it exists. A spreadsheet backup will not restore a merge for you, but it lets you reconstruct what was lost and lets you answer "what did this record look like before".

4. Handle field conflicts

Conflicts are where quality is lost. For each field with a competing value, choose a rule by field type rather than deciding record by record:

• Contact details such as email and phone: prefer the value verified most recently, and keep the other as a secondary field if your CRM has one
• Job title and company: prefer the newest, since people change roles
• Lifecycle stage or lead status: prefer the most advanced status, but check by hand when one record is marked as a customer and the other as unqualified
• Consent and unsubscribe flags: keep the most restrictive value; never re-subscribe someone through a merge
• Owner: keep the owner of the surviving record unless there is an open deal with a different owner, in which case ask
• Free-text notes: never drop them; keep both, ideally attached to the surviving record

5. Merge associated deals, activities and history

A merge should carry the timeline with it: emails, calls, notes, tasks, form submissions and deals. Test this on a couple of records before scaling. Then look for the second-order effects. Two deals on the merged contact may now be the same opportunity entered twice, or two real deals. Treat those separately from the contact merge and review them by hand. Company merges are riskier than contact merges because they move every contact and deal underneath, so do company merges in smaller batches and check the associations afterwards.

6. Build a manual review queue

Set up a simple sheet or a CRM view for records the rules could not decide. Give it four columns: the two record links, why they were flagged, the reviewer's decision, and the date. Assign one person, and set a rule for what happens to records that are still unresolved after a set time; usually the answer is to leave them alone. A slow, clean queue is better than a fast, wrong merge. If you find that the queue is very long, the underlying cause is probably intake, not cleanup.

7. Stop the duplicates coming back

Merging without prevention is a treadmill. Look at where duplicates originate: forms that create a new record on every submission, imports without a match key, integrations that do not check before creating, or reps who add a contact instead of searching. Those are covered in how to prevent duplicate leads in your CRM. For the recurring, rule-based tidying that follows a merge, see manual CRM cleanup automation. Also agree who owns data quality, because "everyone" means nobody.

What automation can and cannot do here

Rules can find candidates, apply the survivor logic and merge high-confidence matches, leaving the rest for review. They cannot guarantee that every duplicate is found, since two records for the same person with different emails and no shared phone will look unrelated, and we would not claim otherwise. If you want this built into your workflow rather than done as a one-off, it belongs in CRM & Sales Workflow Implementation, which starts from a review of your current data. The CRM migration complexity estimator can give you a rough sense of how messy your data is before you decide how much to do by hand.