When a merge fails

A merge is several writes to Attio, and any of them can fail — a rate limit, a network blip, an expired connection, an attribute Attio refuses. When that happens the merge stops where it is and the row in the Merge Log gets a failed badge. Nothing is silently half-done and left for you to find later.

Retrying is safe

Failed rows have a Retry button. Retrying does not start a new merge and does not re-plan anything: it resumes the same merge with the same stored plan, from the step it stopped at. Deduptio records which steps completed and which records were already merged, so a step that already ran cannot run twice.

That is what makes retrying an operation that deletes records safe to press more than once.

Automatic retries

You usually won't need the button. Failures that look transient are retried automatically in the background — up to three attempts, spaced 15 minutes, then 1 hour, then 4 hoursapart. The row's status changes on its own when one succeeds.

Failures that are not transient are not retried. A merge that fails because the Attio connection is dead waits for you to reconnect; a merge that fails because you are at the Free merge cap waits for next month or an upgrade. Retrying those before fixing the cause just fails again.

Merged, with a warning

Some merges land and then hit a problem on the way out. Attio's merge itself is irreversible: once it runs, the original records are consumed and a new merged record exists. Everything Deduptio does after that — applying your field rule to the merged record, clearing a value Attio filled in against a survivor wins rule — is a refinement, not the merge.

When one of those follow-up writes fails, the row reads merged, with a warningrather than failed. The records are merged; the group leaves your queue; the 24-hour restoreworks as usual. Hover the badge for what didn't land. Usually it is a single attribute Attio rejected — the other field decisions are written one at a time so one bad value cannot discard the rest — and you can set that one field by hand on the merged record in Attio.

“This record no longer exists in Attio”

A duplicate group is a photograph of your data taken when the scan ran. If a record in it was merged or deleted in Attio afterwards — by you, by a teammate, by another tool, or by an earlier merge here — the group describes something that is no longer there, and no amount of retrying will change that.

Deduptio retires that group instead of offering it again: it disappears from the Duplicates queue and the Merge Log row explains why. Run the scan again to rebuild the group from what Attio holds now.

Inside a bulk run

In a bulk merge, one failing group does not stop the run. The failure is reported on that group's card and the remaining groups carry on. A plan limit or an expired connection is different — those stop the run cleanly and tell you exactly how many of N were merged, with an Upgrade or Reconnect link.

If it still won't merge

1. Open the group's records with View in Attio ↗ and check they still exist — a record deleted in Attio after the scan cannot be merged, and the group is retired rather than retried.
2. Run Review merge again. The dry run is read-only and re-reads the records live, so it will surface a problem that has appeared since the scan.
3. Reconnect if any page is showing a reconnect banner.
4. Still stuck? Send us the merge row — we keep error records for 90 days and can see exactly which step failed and why.

Failed rows and rollback

A failed row offers Retry, not Restore. Rollback is offered on merges that completed, because a snapshot is only complete once the merge is. Finish the merge with Retry and the row becomes restorable like any other — with the usual 24-hour window, counted from the merge, not from the retry.

If a merge is going to sit failed for a long time, that matters: the recovery window is running while it waits.