Rerun and rollback are different problems
A rerun loads the same source again after you fix a mapping. It works only if every destination record stores its source ID, so the second pass can update in place instead of creating duplicates. A rollback returns the destination to how it was before the run. That covers more than deleting new records: values on records that already existed have to be restored too. Native importers handle the first half, deleting what they created, and not the second.
What native tools do / where they stop
| Platform | What you can undo natively | Where it stops |
|---|---|---|
| HubSpot import | Filter the records an import touched and bulk-delete them | HubSpot says that if an import updated existing records, “you will not be able to reverse the changes by deleting” |
| Salesforce Data Loader / Import Wizard | Delete inserted records by ID; the Recycle Bin keeps them for 15 days (30 if Support extends it) | Records you updated aren’t versioned. You need your own pre-load export |
| Zoho CRM Migration Wizard | An Undo option in Import History | Zoho’s API-based migrations, such as the one from HubSpot, can be undone or rerun only a limited number of times |
The common thread: no native tool snapshots the values it overwrites. Anyone migrating into a CRM that already holds data, such as a merge or a second wave, needs that snapshot.
How to make a migration reversible
- Export the destination before the run, including every field the migration could touch.
- Write the source ID into a custom field on every record, so a rerun can update in place.
- Tag each batch with a run ID, so you can find and delete exactly that batch.
- Run a sample of a few hundred records that covers every object and association, and compare it field by field.
- Freeze automations in the destination during the load, so workflows don’t send emails or change fields you’d then have to undo.
- Keep the source system read-only, not deleted, until counts and spot checks pass.
Where FlitStack AI fits
Every FlitStack AI migration runs against a sample first, uses hash-based validation, and has one-click rollback to the pre-migration state. Delta sync covers changes made during cutover. See the CRM migration checklist or scope a migration.