Workspace Migration Services

Migration

Recovering a failed Google Workspace migration

A failed Google Workspace migration is recovered by establishing what actually copied before changing anything else. Triage takes about 48 hours: inventory both estates, stop the running tool, then decide between finishing forward and rolling back. Re-running a failed job without that inventory is what turns one problem into duplicated mail.

Recovery engagement
First stepStop the running job; change nothing else
Triage48 hours to a written position on both estates
Commonest faultPermissions and ownership, not message loss
Second commonestDuplicates from a re-run without state
DecisionFinish forward or roll back, stated with reasons
Mail continuityRouting stabilised before any further copying

Stop the job before anything else

The instinct on discovering a broken migration is to restart it. Resist it. Most migration tools keep state, and a restart without knowing whether that state survived is how a mailbox ends up with two copies of four years of mail, which is a far more expensive problem than an incomplete copy.

Stopping costs nothing that cannot be resumed. Staff are already affected, and one further hour of a known-bad position is cheaper than an unknown one.

Establish what actually copied

Recovery is an accounting exercise before it is a technical one. Both estates are counted: mailboxes and message counts per mailbox, files and folders per drive, ownership and sharing state, group membership, calendar resources, and where mail is currently being delivered.

That count produces the only document that matters, which is a per-object statement of complete, partial, or not started. Everything after this is a decision rather than a guess, and it is the document we hand over whether or not the recovery work continues with us.

  • Message counts per mailbox on both sides, not sampled
  • File and folder counts per drive, with ownership recorded
  • Sharing and permission state, which is what usually broke
  • Current MX and routing position, including any split delivery
  • Third-party application grants that pointed at the old tenant

Permissions break more often than mail

Mail migration is a mature, well-understood copy. Permission migration is not, because permissions are a graph rather than a list: a file shared with a group whose membership was migrated separately, inside a folder whose owner left, referenced by a link somebody saved in a document.

This is why a migration can report success while staff report that nothing works. The messages arrived; the access model did not. Rebuilding it is methodical work driven by the inventory, and the sequence is ownership first, then group membership, then direct shares, then link-based access.

Finish forward or roll back

Finishing forward is usually right where mail has already been cut over and most content has landed: complete the remaining copy, repair permissions, reconcile, then close. Rolling back is right where the cutover has not happened, or where the target was built on a structure that will cause the same failure again.

Either way the decision is written down with its reasons before work resumes, because the failure being recovered from was frequently a decision nobody had recorded. We would rather hand you an argued recommendation you can reject than resume work you did not choose.

What we see that others don't say

Almost every migration described as failed is in fact partially complete and still running. The damage usually happens in the hour after the discovery, when somebody restarts the job to fix it, and the second pass duplicates everything the first pass had already delivered.

What this doesn't cover

  • We cannot recover data that no longer exists in either estate and is past its retention window. Recovery works with what is still there.
  • We do not provide expert witness or dispute support against a previous supplier, though the inventory we produce is factual and yours to use.
  • We do not resume a third-party tool's job under its own licence unless you hold that licence and authorise it.
  • Triage is a paid engagement. We will not quote remediation before the inventory exists, because any such number would be invented.

Get a fixed price for this

Send your seat count, current platform and deadline. You get a fixed price and an available cutover date, usually within one working day, from the engineer who would run the work.

Prefer to talk? Call +44 20 7183 3436 (Mon–Fri 08:00–18:00 GMT), or message WhatsApp +44 7403 423563.

Questions we get asked

Our migration failed halfway. What should we do first?
Stop the running job and change nothing else. Most tools keep state, and restarting without knowing whether that state survived is what produces duplicated mail, which is harder to fix than an incomplete copy.
How long does it take to know where we stand?
About 48 hours for a written per-object position: message counts per mailbox on both sides, file and ownership counts per drive, sharing state and current mail routing.
What usually broke, if the mail arrived?
Permissions and ownership. Mail copying is mature; the access graph is not, so a migration can report success while staff find that shared drives, group-shared files and saved links no longer work.
Is it better to finish the migration or roll back?
Finish forward when mail has already cut over and most content has landed. Roll back when the cutover has not happened, or when the target structure would cause the same failure again.

How this page is verified

Reviewed by Workspace Migration Services migrations desk, Google Workspace migration engineers. Last checked .

  • Data transfer and ownership behaviour reflects Google Admin Help documentation as of 2026-09-19.
  • Triage sequence and durations are the Workspace Migration Services recovery runbook as of 2026-09-19.
  • Failure-cause ordering derives from Workspace Migration Services recovery engagements to 2026-09-19.

Sources you can check

Related pages

Triage first, decisions second

Tell us what tool was used, what has already cut over, and what staff are reporting. We will come back with the inventory plan and a fixed triage price.