In an Exchange Online migration, choosing the method matters as much as the move itself. Cutover, staged, IMAP and hybrid all reach the same destination, but downtime, the ability to roll back and the number of calls to the help desk on migration morning depend on the method. Below we explain how to choose the method based on user count, source system version and tolerance for downtime, and in which order to carry out the method you choose.
Four migration paths and where they really fit
There are four migration paths using Exchange Online’s native tools. Each was designed for a particular source environment and scale; forcing one in place of another causes problems.
- Cutover migration: All mailboxes on on-premises Exchange are moved in a single batch. Outlook Anywhere must be configured on the source server and reachable from the internet. Microsoft documents this method for Exchange 2003, 2007, 2010, 2013 and 2016 sources. Directory synchronization must be off; the batch itself creates the users in the cloud. The technical upper limit is 2,000 mailboxes, but because of creation and migration time Microsoft considers this method reasonable for 150 users or fewer.
- Staged migration: Mailboxes are moved in groups, and directory synchronization is required. It is supported only for Exchange 2003 and 2007; the method doesn’t exist for Exchange 2010 and later. Since support for those versions has ended, it is rarely used today.
- IMAP migration: Only email is moved from Zimbra, cPanel-based hosting or an old IMAP server. Calendars, contacts and tasks don’t come across. If the source is Google Workspace, look first at Exchange Online’s separate Google Workspace migration batch, which moves mail, calendar and contacts together. The limits Microsoft documents: at most 500,000 items per mailbox and individual messages up to 35 MB.
- Hybrid migration: On-premises Exchange and Exchange Online run side by side for a while; mailboxes are moved in groups, without users noticing and, if needed, in a way that can be reversed. A shared GAL, free/busy information and mail flow under a single domain are provided. This is the standard path for medium and large environments and where long-term coexistence is needed. Microsoft’s decision guide points to hybrid for more than 150 mailboxes on Exchange 2010, 2013 and 2016 sources, and to cutover or express (minimal hybrid) migration for moving fewer than 150 mailboxes in a few days. For Exchange 2019 and Subscription Edition, see the supported version list in the hybrid deployment prerequisites.
Once SPF and DKIM have been verified after the mail migration, prepare a separate plan for moving DMARC to a reject policy.
For sources where the native tools aren’t enough (systems where calendars and contacts must also be moved and that offer no access other than IMAP), third-party migration tools are considered. Moving between two tenants is a separate job; we cover it in our tenant-to-tenant migration plan article.
Decision table: user count, version and downtime tolerance
| Criterion | Cutover | Staged | IMAP | Hybrid |
|---|---|---|---|---|
| Source system | Exchange 2003, 2007, 2010, 2013, 2016 (Outlook Anywhere) | Exchange 2003/2007 only | Any IMAP server; Exchange 2000 and earlier | Exchange 2010 and later |
| Suitable scale | Small organization | Intermediate | Small–medium | Medium–large |
| Directory synchronization | None | Necessary | Optional | Necessary |
| Data moved | Mail, calendar, contacts | Mail, calendar, contacts | Mail only | The entire mailbox |
| Outlook profile | Recreated | Recreated | Recreated | Usually preserved |
| Rollback | Difficult | Difficult | Difficult | Can be moved cloud → on-premises |
The decision order is simple. If the source isn’t Exchange, use IMAP or a third-party tool. If the source is Exchange and users will continue to be managed from on-premises AD, use hybrid. If AD sync isn’t wanted, the user count is low and a migration over a single weekend is acceptable, use cutover. Cutover’s 2,000-mailbox limit stays on paper; in an environment approaching that number, changing Outlook profiles overnight exceeds the support team’s capacity. The limit is operational, not technical.
Pre-migration preparation: DNS, licensing, UPN and mailbox sizes
Whether the migration goes smoothly depends largely on the readiness of the on-premises server and systems infrastructure. The items below should be closed before the migration, whatever the method.
- Domain verification: All SMTP domains in use are added to Microsoft 365 and verified with a TXT record. The MX record is not touched at this stage.
- UPN alignment: It is recommended that a user’s UPN match their primary SMTP address; it isn’t mandatory for every setup. If they differ, the user signs in to Microsoft 365 with a name different from their email address; tell users in advance. Non-routable suffixes such as
.localmust be replaced with a verified domain before synchronization; otherwise the user is created in the cloud with an onmicrosoft.com name. The details are in our Entra Connect Sync versus Cloud Sync comparison. - Licensing: Before the mailbox is moved, the user is assigned a license that includes Exchange Online. Shared mailboxes can stay unlicensed if they don’t exceed 50 GB and don’t need archiving/retention.
- Mailbox sizes: Exchange Online Plan 1 provides 50 GB of primary mailbox storage and Plan 2 provides 100 GB. Mailboxes over the quota are first split into an archive or cleaned up.
- Related data: PST files, public folders, room/resource mailboxes and applications that send mail via SMTP (printers, ERP, monitoring systems) go into a separate inventory. A frequently forgotten dependency is application SMTP relays; if noticed late, invoice and alarm emails stop going out.
For the size inventory, this command in the on-premises Exchange Management Shell is enough:
Get-Mailbox -ResultSize Unlimited |
Get-MailboxStatistics |
Select-Object DisplayName, ItemCount, TotalItemSize |
Sort-Object TotalItemSize -Descending |
Export-Csv C:\Migration\mailbox-size.csv -NoTypeInformation -Encoding UTF8
The few mailboxes at the top of the resulting list determine the total migration time. Plan them separately.
Hybrid Configuration Wizard and the migration batch flow in a hybrid migration
Hybrid setup is done with the Hybrid Configuration Wizard (HCW). HCW sets up the organization relationship, OAuth, send/receive connectors, remote domains and the migration endpoint on both sides together. When the wizard fails, one of the following prerequisites is usually missing; check them one by one before running it:
- On-premises Exchange must be up to date and supported by Microsoft; Microsoft’s hybrid deployment prerequisites page is the reference for supported versions and CUs.
- The Exchange hybrid deployment option (Exchange hybrid writeback) must be enabled in directory synchronization. In Entra Connect Sync, this is one of the optional features in the setup wizard; Cloud Sync also supports Exchange hybrid writeback, and it is enabled separately in the configuration.
- In the classic topology, the MRS Proxy endpoint must be enabled on the EWS virtual directory (
Set-WebServicesVirtualDirectory -MRSProxyEnabled $true). HCW can configure this, but if you have more than one server, check yourself that it is consistent on all of them. - A valid certificate from a public CA must be assigned to Exchange for the Autodiscover and EWS names.
- In the classic topology, Exchange Online must be able to reach the EWS/Autodiscover addresses on 443. If inbound access can’t be opened, modern hybrid (Hybrid Agent) is chosen; in that case an outbound 443 connection is used, but port 25 must still be planned for mail flow.
- The account running HCW must have the Organization Management role on premises and Global Administrator (or equivalent Exchange rights) in the cloud.
Once hybrid is set up, mailboxes are moved with remote move batches. A typical flow in Exchange Online PowerShell:
Connect-ExchangeOnline -UserPrincipalName [email protected]
# Verify that the migration endpoint is reachable
Test-MigrationServerAvailability -ExchangeRemoteMove -RemoteServer mail.example.com -Credentials (Get-Credential)
# CSV: users to move in the EmailAddress column
New-MigrationBatch -Name "Wave-01" `
-SourceEndpoint "Hybrid Migration Endpoint - EWS (Default Web Site)" `
-CSVData ([System.IO.File]::ReadAllBytes("C:\Migration\wave-01.csv")) `
-TargetDeliveryDomain "example.mail.onmicrosoft.com" `
-AutoStart
# Track progress
Get-MigrationBatch "Wave-01"
Get-MigrationUser -BatchId "Wave-01" | Get-MigrationUserStatistics |
Select-Object Identity, Status, PercentageComplete, TotalItemsInSourceMailboxCount
# Once pre-sync is done, complete at the planned time
Complete-MigrationBatch -Identity "Wave-01"
The endpoint name must match the one HCW created; check it with Get-MigrationEndpoint. If you don’t let the batch complete automatically, data is synchronized incrementally and the user’s switchover closes with a final sync of a few minutes outside working hours. Plan waves by department: a manager and their assistant, and teams that use shared calendars, should be in the same wave, because some delegate and folder permissions work only partially between the cloud and on premises. A delegate relationship split across waves generates complaints on day one.
MX, Autodiscover and client switchover order
The switchover order depends on the method. The wrong order usually doesn’t lose messages; it causes Outlook connection problems and a pile-up of support requests on migration morning.
With the hybrid method
- MX and Autodiscover can stay on premises throughout the migration. On-premises Autodiscover redirects the client to Exchange Online for a moved mailbox via the
targetAddressattribute. - When the last wave is complete, the MX record is changed to the value shown for the domain in the Microsoft 365 admin center. The on-premises receive connector is kept open until the old MX’s TTL expires.
- The Autodiscover CNAME record is changed to
autodiscover.outlook.comlast; make sureinclude:spf.protection.outlook.comhas been added to the SPF record.
With the cutover and IMAP methods
- Once the initial sync is complete, the MX record is pointed to Exchange Online and you wait for the TTL period.
- During this time, messages arriving at the old system continue to be transferred by the batch’s incremental sync; the batch is then completed.
- Autodiscover is updated and users’ Outlook profiles are recreated. On mobile devices the account is removed and added again.
In both cases, lowering the TTL (for example to 300 seconds) a few days before the MX change narrows the rollback window. To diagnose undelivered messages after the migration, use the message trace steps in our Exchange Online mail flow troubleshooting guide.
The state of on-premises Exchange after the migration
Removing on-premises Exchange right after the last mailbox moves is a common mistake that is hard to undo. As long as users are synchronized from on-premises AD, the source of authority is AD: attributes such as email addresses, proxyAddresses and recipient type are changed on premises, not in the cloud. Editing these attributes by hand with ADSI Edit isn’t supported and creates inconsistencies.
There are three options:
- Keep one management server: A single Exchange server is kept for recipient management and SMTP relay and kept up to date as part of regular server and cloud management. This is the lowest-risk path.
- Continue with the management tools only: The management tools in Exchange Server 2019 CU12 and later let you manage recipient attributes with PowerShell after the last Exchange server is shut down. In this model there is no Exchange Admin Center and no hybrid mail flow.
- Move to cloud-sourced identity: Directory synchronization is turned off and users are managed in the cloud. This usually isn’t suitable if other applications depend on on-premises AD.
Whichever path you choose, don’t shut the server down until the devices and applications that send mail via SMTP have been moved to Exchange Online (SMTP AUTH, direct send or relay through a connector). Take a final backup of the Exchange databases and decide how long to keep it. For data protection needs on the cloud side, see our article on Microsoft 365 backup.
Common mistakes
- Renewing the certificate at the last minute: If the certificate HCW uses changes during the migration, the connectors and OAuth trust break. Verify from the start that the certificate will remain valid throughout the migration.
- A reverse proxy or WAF cutting off EWS: MRS proxy traffic creates long-lived, heavy connections. The first things to check in move failures are timeout and pre-authentication settings.
- Assigning licenses late: The mailbox of a user moved without a license becomes disabled after a while. The license is assigned before the batch.
- Forgetting application SMTP traffic: Invoice, alarm and scan emails quietly disappear after the migration. The sending IPs and accounts should be written into the inventory.
- Skipping user communication: If mobile device reconfiguration and the Outlook profile change aren’t announced in advance, support requests pile up on migration morning.
- Leaving the identity decision for later: The authentication method (PHS, PTA, federation) should be decided before the migration starts. If it changes mid-migration, sign-in problems mix with migration problems and diagnosis becomes harder.
Microsoft’s Exchange hybrid deployment documentation should be the reference for the official prerequisites and supported topologies of a hybrid configuration. If you need support with planning, wave management and the migration stages, you can review the scope on our Microsoft 365 and cloud services page.
Frequently asked questions
How many mailboxes can be moved in a cutover migration?
Microsoft states that up to 2,000 mailboxes can be moved in a cutover migration, but because of the time it takes it recommends 150 users or fewer as a reasonable limit. Above that number, recreating Outlook profiles and managing a single migration day become difficult. If you have many users or the migration will span several weeks, the hybrid method gives more control.
Does the Exchange server need to be reachable from the internet for a hybrid migration?
In the classic hybrid topology, Exchange Online must be able to reach the EWS and Autodiscover endpoints on port 443, and port 25 for mail flow. With modern hybrid (Hybrid Agent), inbound 443 access isn’t needed for mailbox moves and free/busy information; an SMTP connection must still be planned for hybrid mail flow.
Do users need to recreate their Outlook profile after the migration?
In a hybrid migration, when a mailbox is moved Autodiscover redirects the user to Exchange Online and the profile is usually preserved. In cutover, staged and IMAP migrations the target is a different organization, so the Outlook profile is recreated and the OST file downloads from scratch.
Can on-premises Exchange be removed after the last mailbox is moved?
If users are synchronized from on-premises AD with Entra Connect, recipient attributes continue to be managed on premises. With the management tools in Exchange Server 2019 CU12 and later, the last Exchange server can be shut down; this step is taken after hybrid mail flow, application SMTP relays and Autodiscover records have been cleaned up.