How does deduplication work?
- Agree the match rule first: exact email, or fuzzy name plus phone, or company domain plus address.
- Run the rule in report mode and read a sample, because fuzzy matching produces confident wrong answers.
- Decide the survivorship rules: which record wins, and which field values are kept from the loser.
- Merge rather than delete, so calls, emails, orders and notes from both records end up in one place.
- Add a uniqueness rule at the point of entry, or the duplicates rebuild themselves within months.
A worked example
A regional HVAC company had every customer three times over: once from a web form, once from a call taken on a mobile and once from an old billing import. Technicians were arriving without service history. An assistant matched on phone plus street address, reviewed 200 proposed merges by hand before running the rest, and kept the oldest account number with the newest contact details. Service history assembled into single records and repeat visits stopped starting from nothing.
Where does deduplication show up in your tools?
HubSpot, Salesforce, Zoho CRM and Pipedrive all ship duplicate management tools, and Follow Up Boss flags matching leads. For large jobs the merge is often prepared in a spreadsheet or Airtable and reimported, always after an export that can be restored.
Common mistakes
- Running an automatic merge on a loose name match, which joins two different people with the same name.
- Deleting the duplicate instead of merging it, which throws away orders, notes and email history.
- Cleaning up once and leaving forms and imports able to create the same duplicates next quarter.
Why does deduplication matter?
Duplicates break the two things a database is for: seeing one customer's full history and counting how many customers you have. They also produce the same email three times, which costs goodwill. A careful merge with agreed rules restores both, and a uniqueness check at entry keeps it that way.
How does AssistBPO handle deduplication?
Our data assistants agree the match and survivorship rules with you in writing, run the rule in report mode first, and review a sample by hand before any merge is committed. A dated export is taken before each pass so a bad merge can be reversed. The last step is a uniqueness rule on your forms so the problem does not return.