Duplicates are the most common and most damaging kind of bad CRM data. Two records for the same customer split their history in half: one colleague logs a call on one, another sends a proposal from the other, and nobody sees the full picture. A duplicate rule is how a CRM catches them.
The two parts of a duplicate rule
What counts as a match
A rule compares one or more fields between records. Common choices:
| Matching method | Example |
|---|---|
| Exact | sam@fernhill.example equals sam@fernhill.example |
| Exact after normalising | Sam@Fernhill.example equals sam@fernhill.example once case and spaces are ignored |
| Normalised company name | ”Fernhill Studio Ltd.” matches “Fernhill Studio” |
| Normalised phone number | ”(512) 555-0147” matches “512.555.0147” |
Normalising matters more than people expect. Most real duplicates differ only by capitalisation, punctuation, a legal suffix or the way a phone number was typed.
What happens on a match
- Flag. The record saves, and the possible duplicate is put in front of a person to review. Best for fields where matches are likely but not certain, such as company names.
- Block. The save is refused and the user is pointed to the existing record. Best for fields that should be unique, such as an email address.
Merging safely
Finding a duplicate is half the job. Merging well means:
- Choosing the best value for each field, not simply keeping one record wholesale. The newer record may have the right phone number while the older one has the right address.
- Moving everything related (tasks, deals, files, emails) onto the record you keep.
- Keeping an audit trail of what was merged, by whom and when.
- Having a person decide. Automatic merging on weak matches can join two different people who share a name.
Where duplicates come from
Web forms, imports, email capture, business cards and several team members working the same account. Rules that run when a record is saved catch most of them; a periodic scan catches the ones that arrive in bulk.
Common mistakes
- Matching on name alone. There are many Sam Riveras.
- Blocking on a fuzzy match. People will get stuck, then create the record with a typo to get past the block.
- Never reviewing the queue. A flag nobody reads does nothing.
How GuardenerAI handles it
- Duplicate rules in GuardenerAI Automation apply to persons, companies and leads, and either flag a likely match or block the save.
- Matching can be exact, exact after normalising, by normalised company name, or by normalised phone number.
- Flagged pairs wait in a review queue. Merging lets you pick each field from either record and moves tasks, opportunities, files and email to the record you keep.
- A nightly scan flags duplicates that came in through a CSV import, which does not pass through the save-time checks.
- Nothing merges on its own: a person reviews each pair and decides. See also the CRM page.