Duplicate Prevention Is Easier Before the Import Than After It

A bulk file can contain duplicates for many reasons: two source systems were combined, a branch appears under an old and new name, or a previous export was appended to a current one. Once those rows become live profiles, cleanup becomes slower and more visible.

Compare more than the business name

Names alone are unreliable. “Smith & Co.” may represent several independent firms, while one real business may appear as “Smith and Company” in another source.

Use several identifiers together, such as phone number, website domain, full address, branch code, and coordinates where available.

Create a review group for uncertain matches

Some records can be merged automatically when every core field agrees. Others need human review. A same-address, different-phone pair could be two departments, two businesses in shared premises, or an outdated duplicate.

Mark uncertain pairs instead of forcing an automatic decision.

Keep historical variants for matching, not publishing

Old names and previous phone numbers can help identify duplicates, but they should not necessarily appear in the live profile. Store them as matching clues or audit history.

Run the check again after transformation

Formatting changes can reveal duplicates that were hidden in the raw source. Two phone numbers may become identical after punctuation and country-code normalisation.

A good deduplication step protects the directory from creating two public answers to the question, “Which listing is the real business?”

Deduplication works best before records enter the directory

A bulk import can multiply a small source problem into hundreds of visible duplicates. Before loading data, compare records using more than one signal. Business name alone is too weak, especially for chains, while address alone can incorrectly merge independent tenants. Phone numbers, website domains, branch identifiers, and unit details can help establish whether two rows represent the same entity.