How to Roll Back a Failed Salesforce Import (Without Restoring the Whole Org)

A bad Salesforce import does not always announce itself. The job says success. Then an admin notices the wrong Opportunity stages, blanked phone numbers, or a lookup that now points at the wrong Account.

Restoring the whole org from a weekly backup is the nuclear option. Most of the time you only need the rows that job touched. Here is what to do in the next hour.

1. Stop digging the hole

Do not run another update to fix the first one until you know the scope. A second CSV often overwrites the remaining good values and makes rollback harder.

• Note the job start time and which objects/fields were in the mapping
• Export the success and error files from the load tool if you still have them
• If the load is still running, cancel it

2. Identify which records changed

You are looking for the set of Ids the job wrote, not everything that looks wrong.

• Filter on LastModifiedDate around the job window (and LastModifiedBy if you used a dedicated integration user)
• Compare against the CSV Ids you loaded
• Watch for related records that cascaded (Contacts under Accounts, OpportunityContactRoles, etc.)

3. Do not mass-delete blindly

Deletes feel clean and often are not. Soft-deleted records, file links, and roll-up fields can leave you worse off. Prefer restoring prior field values on the affected rows when you can.

4. Prefer row-level rollback over a full org restore

A full sandbox refresh or weekly export restore puts yesterday data on top of todays good work from other teams. If you snapped the records before the job, you can put only those rows back.

That is what Dataimporter Backup is for: snapshot the records a job is about to touch, then roll back just that set if the mapping was wrong. The rest of the org stays put.

5. When Data Loader leaves you stuck

Desktop Data Loader has no undo. If you did not export a pre-job copy of those rows, your options shrink to:

• Reconstruct from a recent export or report (slow, incomplete)
• Manual fix for a tiny set
• A full restore you would rather not run

Next time, take a snapshot before the update. Pair it with a sample load and the checks in our migration checklist.

6. After you are stable

• Fix the mapping (dates, 15 vs 18 digit Ids, lookups) before you re-run - see Salesforce date format and 15 vs 18 digit Ids
• Re-load into a sandbox first
• Keep using an integration user so LastModifiedBy stays filterable

If you want the product flow for snapshot + rollback before the next risky CSV import, it is on the Backup page.