How to Keep CRM Data Clean: A Guide to Duplicates

How duplicates appear in a CRM, how to detect them and how to merge them: data hygiene, matching rules and preventive controls.

Dirty data never announces itself. One day somebody says “this report is wrong”, and from that day nobody looks at the reports again. A CRM is only as trustworthy as the data inside it.

How duplicates appear

  • Entry from different channels: the same person fills in a web form, then calls; two separate records are created.
  • Format differences: “0532 123 45 67”, “+905321234567”, “5321234567” — identical to a human, different to a system.
  • Spelling differences: “Ahmet Yılmaz”, “AHMET YILMAZ”, “Ahmet Yilmaz”.
  • Bulk imports: an uncontrolled spreadsheet upload writes alongside existing records rather than into them.
  • Company/contact confusion: the same business is opened both as a contact and as a company.

Detection: which field do you match on?

The looser the matching rule, the higher the risk of merging the wrong records; the stricter it is, the more duplicates survive. A tiered approach works in practice:

  1. Exact match: an identical normalised phone number or email is almost certainly the same person.
  2. Strong match: name plus the last seven digits of the phone, or name plus email domain.
  3. Weak match: name similarity alone — never merge automatically; send it for review.

Merging without losing anything

Merging is hard to undo. Three rules:

  • Back up first. Keep a copy of the records before merging so there is a way back if something goes wrong.
  • Carry the history, do not delete it. Calls, messages, quotes and tickets must attach to the surviving record.
  • Make the survivor rule explicit. For example “the most recently updated address wins”, “a filled field overwrites an empty one”.

Prevention: the cheapest method

Cleaning is not a one-off task but a recurring cost. The real saving is in prevention:

  1. Warn at entry. The system should flag it when a record is about to be created with an existing phone number.
  2. Normalise phone numbers. Convert to a single format the moment the number is entered.
  3. Ask for fewer fields, but validate them.
  4. Put imports through a gate. Make a matching preview mandatory before bulk uploads.
  5. Use external IDs in integrations. A unique external ID on records arriving from e-commerce or ad forms prevents the same record being created twice.

A regular maintenance rhythm

Thirty minutes a month costs far less than an annual clean-up project. The checklist:

  • Has the list of new duplicate candidates been reviewed?
  • Is the share of records without a phone or email rising?
  • Are there unassigned records?
  • Have long-dormant records been archived?

Frequently asked questions

Should I delete duplicates?

No — merge them. Deleting takes the history with it; merging brings the history into one record.

Is automatic merging safe?

For exact matches (same phone or email), yes. Weak matches must always pass through human approval.

How many duplicates are normal?

The target is zero, but 1–2% is generally considered manageable. What matters is the direction: if it is rising, your preventive controls are missing.

Next step: to see how duplicate warnings work at the point of entry, request a demo.