Table of Contents
- Migration Phases and Exit Gates at a Glance
- Migration phases and the gate that closes each one
- What We Check First Before Touching DNS
- Build the dependency inventory
- Capture behavior, not just configuration
- Authentication and Routing Checks Before Cutover
- Pre-cutover checks and what happens if they fail
- Compare actual messages field by field
- Running a Pilot and Migrating in Waves
- Prove the pilot before expanding
- Keep parallel operation controlled
- Monitoring and Rollback After Cutover
- Post-cutover symptoms, likely causes and fixes
- Separate measurement changes from infrastructure failures
- Common Mistakes That Cost Sender Reputation
- Defaults aren't a migration strategy
- Ownership failures become sending failures
- Frequently Asked Questions
- Does moving to a new email platform reset sender reputation?
- How long should the parallel run last?
- Does a new sending IP or domain need warmup?
- What should happen when Gmail starts deferring mail after cutover?
- Confirm Your Cutover Gates Before You Switch
- When to bring in an expert
Do not index
Do not index
The cutover finishes on schedule, the delivery dashboard stays green, and then the welcome series starts landing in spam before a launch. An email platform migration needs to preserve more than contacts and templates. It needs to preserve authentication, routing, consent, workflow behavior, and the signals mailbox providers use to decide where messages belong.
Table of Contents
Migration Phases and Exit Gates at a GlanceMigration phases and the gate that closes each oneWhat We Check First Before Touching DNSBuild the dependency inventoryCapture behavior, not just configurationAuthentication and Routing Checks Before CutoverPre-cutover checks and what happens if they failCompare actual messages field by fieldRunning a Pilot and Migrating in WavesProve the pilot before expandingKeep parallel operation controlledMonitoring and Rollback After CutoverPost-cutover symptoms, likely causes and fixesSeparate measurement changes from infrastructure failuresCommon Mistakes That Cost Sender ReputationDefaults aren't a migration strategyOwnership failures become sending failuresFrequently Asked QuestionsDoes moving to a new email platform reset sender reputation?How long should the parallel run last?Does a new sending IP or domain need warmup?What should happen when Gmail starts deferring mail after cutover?Confirm Your Cutover Gates Before You SwitchWhen to bring in an expert
Migration Phases and Exit Gates at a Glance
Accepted doesn't mean inboxed. A receiving server can accept a message and still put it in spam, which is why a green delivery dashboard can conceal a migration problem. The distinction between delivery and email deliverability determines what the operations team should measure before declaring success.
Treat the cutover as a reputation event, not a file transfer. The brand domain and subscriber list can stay unchanged while the sending hosts, signing domains, return-path, and tracking infrastructure change underneath them.
Migration phases and the gate that closes each one
The following schedule is a planning example, not a promise. Phases overlap, and the post-cutover watch starts when production traffic moves, not when the project begins.
Phase | Typical timing | Exit gate |
Inventory | Week 1 in the example plan | Every stream, domain, integration, and credential has a named owner |
Authentication and routing | Weeks 1 to 2 in the example plan | Representative messages pass authentication and follow the intended route |
Pilot cohort | Weeks 2 to 3 in the example plan | Records match field by field, and unsubscribe and bounce handling work |
Parallel run and waves | Weeks 3 to 6 in the example plan | Suppression and consent match exactly, with no unexplained hard-bounce rise |
Post-cutover watch | Days 1 to 30 in the example plan | Deferrals, complaints, and placement remain within the agreed baseline range |
Inventory closes only when the team can account for every sending dependency, including the small application nobody considers part of marketing. Authentication closes only after real messages from the new environment pass, not after a platform displays a verification badge.
The pilot proves that data and events survive the move. The parallel run proves that the business can operate, reconcile changes, and recover without re-mailing unsubscribed contacts or losing critical notifications.
For an infrastructure migration, Finchum Fixes IT's guide to moving email to Microsoft 365 provides useful context for mailbox-side planning. Mailbox access and outbound deliverability should have separate acceptance criteria because either can work while the other fails.
Define the evidence required to close each gate before implementation begins. Otherwise, project pressure tends to replace measurable acceptance criteria with screenshots and verbal assurances.
What We Check First Before Touching DNS
Before a migration changes routing, map what sends email and what each stream depends on. The domain and list may stay the same while the sending host, signing domain, and tracking layer change. Treat each change as a reputation risk to measure, not a calendar task to complete.
MailAdept's Week 1 deliverability audit reviews infrastructure, authentication, domain setup, routing, the provider environment, and key deliverability risks, then sets priorities. The inventory below gives teams a shared view of the systems and business rules that can affect a cutover.

Build the dependency inventory
- List every message stream. Separate marketing, lifecycle, transactional, and outbound traffic, recording the domain or subdomain each uses. A password reset has a different operational owner and failure response from a newsletter.
- Record every sending host. Include SMTP relays, application servers, shared-mailbox devices, printers, and scanners. An overlooked relay can interrupt invoice delivery while employee mail continues normally.
- Map credentials and endpoints. Record which system uses each API key, SMTP credential, and webhook endpoint. Store secret references rather than passwords in the inventory to avoid creating a security risk.
- Export suppression and consent data. Identify where each status lives and which system takes precedence if records conflict. Without a clear rule, an import can reactivate someone who has opted out.
- Map templates, automations, and replies. Document entry and exit conditions, delays, reply-to addresses, shared-mailbox access, and delegate permissions. Messages can send successfully while customer replies land in an unattended mailbox.
- Trace CRM and analytics synchronization. Identify which events create contacts, update preferences, record purchases, and stop campaigns. Broken synchronization can trigger duplicate journeys or promotional messages after a conversion.
- Assign ownership and status. Give each row an accountable person, a test, a cutover status, and recovery instructions. A shared task without a named decision-maker slows incident response.
Capture behavior, not just configuration
A field called “subscribed” can mean different things in different systems. Record its allowed values, how blanks are handled, timestamps, and the rule for resolving conflicting updates.
Apply the same scrutiny to automation. A contact import should not count as a fresh signup unless that behavior is intentional. Otherwise, existing customers may receive onboarding again and submit avoidable complaints.
Include small streams too. A single-purpose form mailer may rely on an old credential that only becomes visible when it stops working.
Keep the inventory as a shared operational sheet, not a static presentation. Marketing, product, IT, and support should confirm their dependencies and ownership before the pilot begins.
Authentication and Routing Checks Before Cutover
A cutover can keep the same From address and recipient list while changing the sending host, signing domain, envelope-from domain, and tracking host. Check each layer before switching traffic. The domain and list may stay constant; the reputation signals do not.
Microsoft's Outlook.com requirements apply to domains sending more than 5,000 messages per day to its consumer services. They require SPF and DKIM to pass, a published DMARC policy of at least p=none, and alignment through SPF or DKIM. Measure Outlook-specific traffic, not total global sends.
Pre-cutover checks and what happens if they fail
Check | Pass condition | Why it matters | If ignored |
SPF | The new sending hosts are authorized, and representative messages pass | Receivers check whether the host can send for the envelope domain | Legitimate messages can be rejected or filtered |
DKIM | The new platform's signature validates | Receivers can verify the signing domain and message integrity | Authentication failures can undermine trust |
DMARC | A published policy exists, and an authenticated identifier aligns with the visible From domain | Alignment connects authentication to the address recipients see | Messages can fail provider requirements |
One-click unsubscribe | Bulk marketing messages support the required unsubscribe mechanism | Recipients need a working way to stop unwanted mail | Complaints and processing failures become more likely |
Spam complaints | Provider-reported complaints stay within the approved operating range | Complaint changes can signal audience or migration problems | The team may expand a failing wave |
Hard bounces | No unexplained increase against comparable pre-migration traffic | An increase can point to stale data or missing suppressions | Repeated invalid sends can harm reputation |
For Gmail bulk traffic, Google's sender requirements require authentication and applicable unsubscribe controls. Spam rates must stay below 0.3% and ideally below 0.1%. Treat the lower figure as the operating target, not the higher limit as a buffer.
Compare actual messages field by field
Inspect Authentication-Results, Return-Path, DKIM-Signature, and Received in representative messages. Compare the envelope domain, signing domain, sending host, and authentication results across the old and new environments.
Check the destinations behind tracked links, too. A changed redirect domain alters the message's infrastructure footprint even when the creative and sender address remain unchanged.
Check SPF, DKIM, and DMARC records alongside the corresponding message headers. Use this SPF, DKIM, and DMARC authentication guide to review the records and their roles. Record-level checks complement, but do not replace, inspection of a delivered message.
Test marketing, transactional, and lifecycle traffic separately. They may use different credentials and signing settings, so a successful newsletter does not confirm that a password-reset email follows the same authenticated route.
Keep inbound routing distinct from outbound sending. Changing MX routing does not migrate an application's SMTP credentials, and authorizing a new sender does not confirm that replies reach the right mailbox.
Running a Pilot and Migrating in Waves
Start with a pilot, retain both platforms for a defined fallback window, then advance in waves only after measured throughput and error rates pass the agreed gates. Calendar-based expansion is convenient, but it doesn't establish that the new route is safe.

Prove the pilot before expanding
- Select representative cohorts. Include active, consented recipients from low-volume and high-volume streams, with coverage across receiving providers. Don't build a pilot entirely from employee addresses or unusually enthusiastic subscribers.
- Reconcile records and behavior. Compare contact fields, consent, suppressions, timestamps, and segment membership. Then test event processing, unsubscribe propagation, bounce handling, rendering, tracking, and reply-to behavior.
- Exercise human and application mail. Use controlled transactions to validate employee messages, receipts, notifications, and reset flows. Compare placement and latency, not merely the platform's accepted status.
- Apply explicit release gates. The UnixDev migration checklist supports 100% agreement for suppression and consent, zero unexplained hard-bounce increases, and documented parity for critical workflow events before expansion. The same checklist describes email-only migrations for 500-user environments as commonly taking 2–4 weeks, with larger staged deployments requiring additional support and remediation capacity.
- Move a bounded wave. Change a stream or domain that the team can diagnose and reverse independently. Hold it whenever a release gate fails, even if throughput looks healthy.
Keep parallel operation controlled
Parallel operation means both environments remain available. It doesn't mean both should send the same production event to the same customer.
Define which platform owns each event during the fallback window. Use stable event identifiers and reconciliation logs so retries don't produce duplicate receipts or welcome messages.
A migration also needs explicit email automation acceptance tests. Verify that a purchase stops the intended sequence, an unsubscribe reaches every relevant system, and an event retry doesn't re-enroll the contact.
The trade-off is operational effort. Smaller waves slow the transfer but narrow the failure domain, while larger waves reduce coordination overhead at the cost of harder diagnosis and a wider customer impact.
Reserve people for support, reconciliation, and remediation throughout the rollout. A schedule that allocates everyone to implementation leaves nobody available to investigate the first production failure.
Monitoring and Rollback After Cutover
An unexplained increase in provider deferrals should stop expansion before the team increases volume again. The first response is containment and diagnosis, not an automatic full rollback that may introduce another routing change.
Capture the pre-migration baseline by recipient provider, sending domain, and message stream. Include deferrals, hard bounces, complaints, unsubscribes, webhook failures, queue latency, and placement so a healthy aggregate doesn't hide a broken transactional route.

Post-cutover symptoms, likely causes and fixes
These are investigation starting points, not diagnoses established by the symptom alone.
Symptom | Likely cause | Fix |
Gmail deferrals immediately after cutover | Authentication trouble, a changed sending route, or an overly aggressive ramp | Pause the wave, inspect the provider response and authentication results, then retest |
Outlook.com rejections for a particular domain | Alignment failure or another sender-requirement violation | Validate representative messages and correct the stated failure before resuming |
Open rates drop while clicks hold | Changed tracking, privacy behavior, or reporting definitions | Compare clicks, placement, and business events before changing sending strategy |
Hard bounces rise after cutover | Incomplete suppression transfer or list synchronization | Reconcile affected records and hold the stream |
Webhooks fail and unsubscribes remain unprocessed | Missing endpoint configuration, expired credentials, or rejected events | Restore access, replay safely, and confirm end-to-end processing |
Read the complete SMTP response, including the enhanced status and provider explanation. A 4xx temporary failure requires different handling from a 5xx permanent rejection; repeatedly retrying permanently rejected recipients isn't a recovery plan.
Preserve the original response text and the platform's event classification. Otherwise, a reporting translation can hide the actual reason messages were deferred or rejected.
Separate measurement changes from infrastructure failures
A new tracking host, pixel behavior, or reporting definition can move open rates without an equivalent change in placement. Compare unique clicks, completed transactions, reply activity, and controlled placement observations before concluding that the migration damaged reputation.
Conversely, stable opens don't excuse a growing deferred queue. A small delayed segment can contain password resets or account alerts with immediate customer consequences.
Rite NRG's discussion of controlling downtime and rollback offers useful broader migration context. For email, recovery must preserve current consent and suppression state, not merely restore yesterday's configuration.
Use a proposed 30-day watch window, with daily operational reviews during the opening week and twice-weekly reviews afterward if signals stabilize. This is an operating cadence, not an industry benchmark, and active alerts should continue between reviews.
Keep the old configuration available until reconciliation, reporting parity, and recovery procedures have been demonstrated. If the fallback platform has stale suppressions or revoked credentials, it isn't a fallback.
Common Mistakes That Cost Sender Reputation
The costs are missed customer actions, unnecessary support tickets, and avoidable reputation damage. A vendor's setup guide won't catch every problem because it describes the new system, not the sending habits and hidden dependencies carried into it.

Defaults aren't a migration strategy
- Accepting unreviewed defaults. Shared sending pools, click-tracking hosts, retry behavior, and daily caps can differ from the old platform. Review them against the inventory before the pilot, or the team may discover a material infrastructure change only after performance drops.
- Bundling every stream into a single wave. Promotional volume can conceal a failing receipt or lifecycle route. Give critical streams independent tests and release decisions so customer-facing failures remain visible.
Ownership failures become sending failures
- Counting only global volume. Consumer mailbox requirements and provider responses need provider-specific analysis. Count traffic by receiving service and sending domain, or a healthy global average can conceal a compliance problem.
- Closing the ticket on cutover day. A technically complete transfer hasn't demonstrated stable production behavior. Keep the migration open through the observation window, with named owners for deferred queues, complaints, and integration failures.
- Replaying everything after recovery. Blind replay can duplicate notifications or send messages that are no longer appropriate. Check event identifiers, current consent, and business relevance before releasing queued traffic.
Refact's discussion of how publishers boost email reach provides relevant context for audience-led sending programs. A migration doesn't make an unwanted campaign welcome, and moving stale contacts into a cleaner interface doesn't improve their consent or engagement.
The strongest fix is often restraint. Hold the questionable segment, repair the missing dependency, and resume only when the evidence supports it.
Frequently Asked Questions
Does moving to a new email platform reset sender reputation?
A migration changes parts of the sending setup, not necessarily the reputation associated with an established domain. Sending hosts, signing domains, return paths, or tracking links may be unfamiliar to receiving providers, so reputation can shift as traffic moves. Treat the cutover as a reputation event and judge each wave by its results, not by the calendar.
How long should the parallel run last?
Keep the old environment available until the team has reconciled customer state and observed the relevant lifecycle, transactional, and promotional events. Retire it only when recovery remains practical and both systems agree on current suppressions and consent.
Does a new sending IP or domain need warmup?
New or unfamiliar sending infrastructure may need a controlled ramp. The right approach depends on the sending setup and audience, so avoid treating a fixed warmup schedule as a migration gate. Mailwarm, a sister brand of MailAdept under Mailwarm Inc., covers warmup as a separate discipline.
What should happen when Gmail starts deferring mail after cutover?
Hold the affected wave and inspect the full deferral response before changing volume. Check for unexpected routing or rate changes, address the stated cause, then resume with a bounded test. Reducing volume may ease pressure, but it does not resolve an authorization or configuration fault.
Confirm Your Cutover Gates Before You Switch
Before cutover day, assign an owner to each dependency and define who can hold traffic. The release decision depends on evidence that the new route works and the team can respond if it does not.
- Inventory ownership is complete. Every dependency has a named owner, a cutover status, and a tested destination.
- Authentication and routing are proven. Actual messages use the intended route, pass the required checks, and meet the receiving provider's requirements.
- The pilot reconciles. Consent, suppression, and critical workflow behavior match without unexplained exceptions.
- Fallback remains usable. Recovery access, credentials, routing, and current customer state are available.
- Incident ownership is explicit. The on-call owner knows the stop conditions, escalation path, and authority to hold traffic.
When to bring in an expert
Repeated incidents, a scheduled migration, provider blocks, persistent deferrals, or unclear deliverability ownership are reasons to consider outside help. MailAdept embeds a dedicated deliverability expert to audit the setup, guide fixes, and provide a step-by-step roadmap.
On cutover day, first confirm who has authority to pause the next wave. Then use measured provider responses, complaints, and placement to decide whether traffic can expand.
MailAdept offers a free deliverability audit to identify authentication, routing, reputation, and ownership risks that can undermine a cutover. Want a second pair of eyes on the setup? Visit MailAdept to start before the next migration wave.
