Cleaning up a messy CRM

TL;DR:

• Match duplicates on something stable like email address, not on names, and review before merging.

• Keep the record with the richest history, copy over any fields the others have, and never delete without a backup export.

• Stop the mess coming back: required fields, one owner per contact and a short rule for who may create records.

Your contact list has three entries for the same customer. Sales calls her twice, support emails the old address, and a report says you have more customers than you do. A customer database (a CRM) gets messy because many people add to it over time with no shared rules. Cleaning it is mostly deciding rules, then applying them carefully.

Export a backup first

Before changing anything, export the contacts to a file and save it with today's date. Most CRM tools offer an export to CSV; look in the contacts list or the settings. If a merge goes wrong, that file is how you recover. Merging is often hard to reverse inside the tool, so assume you cannot undo it.

Write the rules before you touch a record

• Dedupe rule. What makes two records "the same person"? Email address is the best key because it is usually unique. Names are weak: two real people can share one, and one person can appear as "Bob" and "Robert".

• Survivor rule. Which record do you keep? A sensible default is the one with the most history (deals, notes, emails), and if tied, the most recently active.

• Field rule. When two records disagree, which value wins? Usually the most recent, unless the older value is one a person typed on purpose, like a preferred name.

• Required fields. Decide the minimum every contact needs, for example email, company and owner. A record missing them gets fixed or removed.

• Ownership. Every contact has exactly one owner who is responsible for it. Merging two records with different owners means choosing one.

Worked example: merging duplicate contacts

Say the export shows three rows that look like one customer. These are made up for the example.

1. Normalize the key: Lowercase emails and trim spaces. Rows 1 and 2 then match exactly: row 2 had a capital letter and a trailing space that hid the match. In a spreadsheet, =LOWER(TRIM(C2)) does both.

2. Handle the near match: Row 3 has a different email and a slightly different company name. Same name, same company, different address: that is a candidate, not a certainty. Look at the activity before deciding, or ask the owners.

3. Pick the survivor: Row 1 has the latest activity and the most complete fields. It survives.

4. Copy across what is missing: Row 2 adds nothing new. Row 3 holds a second email. If it is Dana's work address, add it as an extra email on row 1 rather than losing it. Company stays "Harbor Print" if that is how the company writes it; check their website.

5. Resolve the owner: Sam and Priya both touched her. Message them, agree on one owner, and note who in the record.

6. Merge and verify: Merge rows 2 and 3 into row 1. Open the result and check that notes, deals and emails from all three are present, and that the contact count dropped by two.

Counting check: you started with 3 rows for one person and should end with 1, so the total number of contacts should be exactly 2 lower than before. If it dropped by more, something merged that should not have.

The same method works for any job with a long list of near duplicates: normalize first, find candidates, decide by rule, then check the count. Do a small batch of ten, look at the result in the tool, and only then do the rest.

Mistakes to avoid

• Merging on name alone. Two real people named Alex Kim will become one.

• Letting a tool merge everything automatically on its first run. Review a sample by hand first.

• Deleting records that look empty. A bare email may still be on a live mailing list or a past invoice.

• Cleaning once and walking away. New duplicates arrive daily from forms, imports and manual entry.

Keeping it clean afterwards

• Make the fields from your list required where the tool allows it, so a record cannot be saved without them.

• Search for the email before adding a new contact.

• Review ownerless contacts every month; unowned records are where mess builds.

• Pick one format for phone numbers, company names and countries and write it down for everyone.

Note: Contact data can include personal information. Check your company's privacy rules, and any law that applies where your customers live, before exporting or sharing a list. For legal questions, ask a professional.

Doing it faster

Sort the export by email in a spreadsheet and the duplicates sit next to each other, which makes the review quick. Once you have a method that works, run it the same way each time, on a schedule, so the list never gets as bad again.

Frequently asked questions

Should I delete or merge?

Merge when two records are the same person, so their history is kept. Delete only records that are clearly junk, and only after exporting a backup.

What if I am unsure two records match?

Leave both, add a note, and ask the owner. A duplicate is a small cost; a wrong merge can lose real history.