What does a Microsoft 365 migration rollback plan look like?

Key facts

• Most rollbacks are partial: one user, one mailbox or one site, not the whole migration.
• Lowering MX and Autodiscover TTLs days ahead makes DNS changes reversible within hours.
• Source systems should stay intact until the business signs off, often for weeks after cutover.
• Every migration has points of no return, such as moving a domain between tenants; name them in advance.
• Go/no-go criteria agreed before cutover prevent decisions made under pressure.

What a rollback plan is for

No one plans to roll back, which is exactly why the plan matters. Without it, a problem on cutover night turns into an improvised debate. With it, the team knows what triggers a rollback, who decides, and what steps to take.

In practice, full rollbacks are rare. Wave-based migrations limit the blast radius, and most issues are fixed forward. The plan is still essential, because it forces the decisions that make a migration safe: keeping the source alive, sequencing DNS changes, and agreeing when there is no going back.

Go/no-go criteria

Write criteria for each wave and get the business sponsor to approve them. Review them at a short go/no-go meeting before cutover, and again the morning after.

• Pre-stage sync complete for all mailboxes or files in the wave, with errors reviewed.
• Pilot users confirmed mail, calendar, shared mailboxes, files and mobile access.
• Identity, MFA and Conditional Access tested for the wave's users.
• Support staff and floor walkers scheduled.
• No conflicting business events, such as month-end close or a major deadline.
• Rollback owner and decision maker named and reachable.

Rollback by workload

Each workload has different reversibility. The table summarizes typical approaches; your plan should list exact steps and commands.

DNS and timing

DNS is the switch for mail and many client experiences. Lower TTLs on MX, Autodiscover and related records several days before cutover so changes in either direction take effect quickly. Record the original values in the plan.

Keep the source system running, and where possible set it to read-only after cutover rather than switching it off. For mail, keep forwarding or coexistence in place until the wave is stable. For files, make the source share read-only so users cannot keep editing old copies, but it remains available if needed.

Points of no return

Some steps cannot be easily undone. Removing an email domain from a source tenant so it can be added to a target tenant is the classic example in tenant-to-tenant work. Decommissioning Exchange Server, deleting a file share, cancelling a carrier contract or removing on-premises Active Directory are others.

Name these steps in the plan, schedule them only after explicit sign-off, and separate them from cutover by enough time to find problems. Once past a point of no return, the plan switches to fix-forward procedures and vendor escalation paths.

How Accred Consulting approaches it

Accred Consulting writes a rollback plan into every migration runbook. Discovery identifies workloads, dependencies and points of no return, and the proposal sets scope, fixed price and timeline. The US-based specialists assigned to your engagement define go/no-go criteria per wave, lower DNS TTLs ahead of cutover, keep source systems intact until sign-off, and document exact rollback and fix-forward steps. In one consolidation of about 1,200 users after an acquisition, 6 waves over 8 weeks avoided an all-hands outage. Handover includes documentation and knowledge transfer, with optional support retainers.

Track record: 475+ client projects, 2,000+ migration workstreams and 210,000+ users migrated by a US-based team. Company facts · Case studies

Related questions