Dedupe

Dedupe is a duplicate-management platform built specifically for Attio.

Attio’s native merge handles two records at a time and leaves you to pick a winner; Dedupe resolves the entire cluster — three, five, nine records at once assembles a surviving record attribute by attribute, and carries every note, task, file, list membership and relationship across before the duplicates are retired.

It runs in two modes: a one-time cleanup that corrects an existing mess, or continuous monitoring that catches new duplicates as forms, imports and enrichment create them. Detection is transparent (per-object, per-attribute match rules with weights and guardrails), automation is confidence-gated (auto-merge above your threshold, ambiguous clusters queued for review), and every merge is logged and reversible for 30 days. Works across Companies, People and any custom object.

Detect: Fuzzy and exact matching per object and attribute (domain, email, phone, name), with normalization for legal suffixes, punctuation, subdomains and free-mail addresses. Related records are grouped into clusters, not just pairs.

Resolve: Each cluster arrives with a smart-merge suggestion: the most complete, most recent, most-trusted value per field. Accept the row or override any cell. Save a resolution as a rule and it applies to every future cluster.

Migrate: Notes, tasks, files, list memberships, linked people/deals and activity move to the surviving record. Nothing is silently dropped, if something can’t be moved, the merge stops and tells you why.

Keep it clean: Continuous monitoring checks new and updated records on arrival. Obvious duplicates auto-merge above your confidence threshold; the rest land in a review queue with the evidence beside them.

Undo anything: Every resolution writes an audit entry. Roll back a single merge or a whole run for 30 days; retired records return exactly as they were.

Connect Attio: Authorize via Attio’s API with scoped access. Restrict Dedupe to specific objects; every write is attributed and logged.

Choose objects: Enable Companies, People and/or any custom object you want monitored.

Set match rules: Per attribute, pick a comparison (exact / fuzzy / normalized) and a weight. Mark identity-critical fields (e.g. billing country) as guardrails so franchises, subsidiaries and regional entities are never grouped.

Set thresholds: Define the auto-merge score (e.g. ≥ 0.95) and the review band (e.g. 0.80–0.95). Above auto-merges, in-band queues, below is ignored.

Run a read-only scan: Get the cluster count, affected objects and proposed resolution plan. Nothing is written until you approve.

Resolve or automate: Work the queue manually, or turn approved resolutions into saved rules and enable continuous monitoring.

Safety options: Dry-run mode previews outcomes before any write; writes are rate-limited so it’s safe against a live workspace; rollback stays available for 30 days.