CRM duplicate records: how to merge them safely

Short answer

Merging duplicate customer records in a CRM takes four steps: export the database, sort duplicates into three groups by confidence (exact same number, probable, doubtful), decide by rule which record is the master and which field values survive, and merge only the confident group automatically. Doubtful pairs are reviewed by a person. The reason is simple: in many CRMs a merge cannot be undone, and a wrong merge mixes two customers' data together.

This article is about cleaning a database that has already accumulated duplicates. Preventing new ones is a separate job.

How duplicates fill a database

• Old Excel lists have been imported several times.
• Numbers are written in different formats: 055…, +99455…, 99455….
• The same customer wrote from different channels and each channel opened its own record.
• Managers created new records without searching first.
• When the company moved from a previous CRM, data from two systems overlapped.

The result: leadership says "we have 12,000 customers" when the real number may be far lower. Every duplicate distorts reports and sends the same person two messages in a campaign.

Step 1: export and back up

Before cleaning, export the whole database and keep that copy untouched. The CRMs' own documentation states why: HubSpot, for example, says merged records cannot be split again. A wrong merge can only be repaired by hand, from the backup.

Step 2: sort duplicates into three groups

1. 1ConfidentThe same phone number in normalised form, or the same email. This group can merge automatically.
2. 2ProbableSame name and the same last 7 digits of the number, or the same name with a similar email spelling. A person confirms with a quick look.
3. 3DoubtfulOnly the same full name, or only the same company name. Never merged automatically — these are often different people.

Normalising numbers into one format first is the step that finds the most duplicates. Without it, "055 234 56 78" and "+994552345678" look like different customers.

Step 3: choose the master record

In a merge, one record becomes the master and the other's data moves into it. Choose the master by rule, not by chance. A practical rule: the record with an open lead or an active owner; failing that, the record with the most activity; if still tied, the oldest. In most CRMs the master record's field values win in a merge — HubSpot works this way — so choosing the master decides which data survives.

Step 4: field rules

Write down in advance which value survives for each field:

• Name: the more complete value (with surname).
• Phone and email: all are kept — one as primary, the others as additional.
• First source: the oldest record's source. This field is never replaced by a newer one.
• Status and stage: the furthest along — but if there is an open lead, its status.
• Activities, notes, orders: all combined, none deleted.
• Consent: the most restrictive value — if any record has an opt-out, the opt-out stays.

Why the first source must not change is explained in lead source attribution.

B2B: contacts and companies are separate

In a B2B database, three employees of the same company are three separate contacts, and merging them is wrong. Duplicates sit at another level here: "Alfa LLC" and "Alfa MMC" may be two records of one company. Compare company records by tax number or website domain, and keep the contacts under the company.

Who does it and who signs off

The cleanup should be one person's project — usually the CRM admin or whoever owns sales operations. The head of sales signs off the rules: the master record, the field rules, what happens to doubtful pairs. Managers are brought in to check their own customers in the probable group, because they know them.

Checking after the merge

The work is not over when the merge finishes. Run three checks within a week. First, does the total number of activities match the figure from the backup? Second, open 30 randomly chosen merged records and check: did the name, numbers, source and consent survive according to the rules? Third, ask the managers: has anyone seen another person's order or note on their customer's record? If an error turns up, those records are restored by hand from the backup and the rule is tightened. Write the result up as a short note, so the next cleanup does not repeat the same mistake.

An illustrative example

This is an illustrative example. A clinic is combining two years of Excel lists with its new CRM. The database has 9,400 records. Once numbers are normalised, 1,300 records fall into the confident group and merge automatically. Another 420 pairs are probable: administrators review them over two days and confirm 350. The 180 doubtful pairs are left alone.

After the merge the database drops to about 7,700 records, while the number of activities stays the same. In the next campaign, patients stop receiving two reminders each.

Common mistakes

• Merging without a backup.
• Merging automatically by full name.
• Deleting duplicates — the activity history and the first source are lost.
• "Restoring" an opted-out customer's consent from another record.
• Not changing intake rules after the cleanup — three months later the database fills up again.

Limitations

No algorithm finds every duplicate: the same person recorded with a different number and a different name can only be confirmed by the customer. Family numbers and office lines carry their own risk. Cleaning a database is processing of personal data — check retention and deletion rules against local law.

Duplicates in Vexvon

In Vexvon duplicates are mainly prevented at intake: if an open lead exists in the same company for the same number, the new lead is linked to it as a duplicate and the earlier history is shown; when the same number writes again it is added to the customer's existing timeline. When an old database is uploaded by CSV, repeated numbers are checked during upload. We do not claim a tool for merging individual records by hand — cleaning the old database before uploading it is recommended. More on Vexvon CRM.

Next step

Export the database, normalise the numbers and count the rows sharing a number — that is the size of your confident group. So that the database does not fill up again after the cleanup, apply the intake rules from duplicate lead prevention. More articles are in the lead management section, and we can review your migration plan together during a demo.