How to find duplicate companies in Attio

Last checked September 1, 2026 · Deduptio is our product. Every claim about another tool comes from its public site or documentation and is linked in place — tell us if something is out of date.

What Attio does on its own

Attio surfaces likely duplicate companies using the domain attribute and prompts you to merge them. When both records carry the same domain, this works well and costs nothing. Merging is available on company records, two at a time, and it cannot be undone.

The gap is not in the merge. It is that a large share of duplicate companies never share a domain in the first place.

The five ways duplicate companies escape a domain rule

1. No domain at all. A company record created from a hand-typed name, a conference list, or a person using a Gmail address has nothing to match on. These are the most common duplicates in young workspaces and the ones domain matching never sees.
2. Rebrands and acquisitions. The company changed name and domain; you have the old record from 2024 and a new one from last week, with no overlapping value.
3. Several legitimate domains. Regional sites, a product domain distinct from the corporate one, a marketing domain, an old domain still redirecting.
4. Different sources normalize differently. One integration writes acme.com, another writes www.acme.com or https://acme.com/. Whether these match depends entirely on normalization.
5. Group structures. A holding company, its trading name and a subsidiary may share a domain while being three genuinely separate records — the inverse problem, where domain matching produces false positives.

A rule that holds up

The false positives to expect

• Subsidiaries and parents. Same domain, different entities, different deals. Link them; do not merge them.
• Franchises and locations. Twenty records legitimately called the same thing, distinguished only by address.
• Agencies and their clients.An agency's staff often appear on the client's domain and vice versa, which drags people-based evidence in the wrong direction.
• Generic names. Every market has three companies called Apex, Summit or Northstar. Name similarity is meaningless here without a second signal.
• Shared infrastructure domains. Records whose domain got set to a registrar, a hosting provider or a free-email domain during an import — these will cluster hundreds of unrelated companies into one group.

Choosing the survivor

For companies, the survivor is usually the record other things point at: the one with the people linked to it, the open deals, the list memberships, and the notes. A newer record may have cleaner fields, but if the older one carries three years of relationships, keep the older one and copy the good values across. Field-level decisions are covered in survivor and field rules.

After merging, check the people. A merged company should have every contact from both records attached to it, and any person record still pointing at a deleted company is a sign the association did not follow.

Keeping it clean

Company duplicates rebuild faster than person duplicates because so many tools create company records as a side effect — enrichment, email sync, forms, imports. Two habits fix most of it: make every integration write with a unique attribute so it updates instead of creating, and run a scheduled scan weekly so a new cluster is a handful of groups rather than a project.

Related

• Fuzzy matching company names — the normalization and thresholds behind pass 3 and 4.
• Finding duplicate people in Attio — the same problem where the identifier moves with the job.
• How to merge records in Attio · The full guide