Back to all articlesCRM migration

Choose One Cutover or Migration Waves Without Splitting Record Context

Alactic Editorial10 min read

Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.

A phased CRM migration is workable when each wave moves a usable group of related records, leaves a clear owner for every live field, and can be reconciled and reversed without destabilising another wave. If those conditions fail, a single cutover may create a shorter, more manageable risk than weeks of divided customer context.

The choice is not simply between a large migration and several small ones. It is between concentrating change into one controlled window and maintaining temporary boundaries between two operating systems. A small import can still create a large operational problem if the account moves today, its open opportunities move next month, and its activity history remains somewhere else.

Use the worksheet below to test the boundaries before committing to a schedule. The goal is to choose a sequence that keeps records interpretable and gives the team a credible recovery path.

1. Choose the unit of context before choosing the waves

Record counts are useful for estimating transfer work, but they are a weak basis for grouping waves. Start with the smallest collection of records someone needs to understand and act on a customer situation.

For one business, that collection may be an account, its contacts, open opportunities, activities, notes, and attachments. For another, it may include a parent account and several subsidiaries because the same people negotiate across them. A contact linked to multiple accounts can turn two apparently independent groups into one dependency.

Classify each relationship:

  • Must move together: separating the records would make an active workflow incomplete or misleading.
  • Can remain separate temporarily: users can consult a clearly identified, read-only source without losing essential context.
  • Requires a controlled bridge: work depends on information crossing the boundary, with an explicit update process and owner.

A bridge might be a reviewed daily export, a controlled manual handoff, or a validated transfer process. It is not a feature assumption. Check whether the required capabilities are available in the workspace and current plan before designing around them.

The important distinction is between missing data and missing meaning. An opportunity may import successfully yet become misleading if its contact role, account relationship, or latest negotiation note is absent.

2. Define the temporary boundary as an operating rule

For every wave, write down which system owns which records and fields during the overlap. Avoid a rule such as “use the new CRM where possible.” It leaves users to resolve conflicts individually.

A useful boundary statement names the record group, the authoritative system, permitted edits, and the treatment of late changes. For example: “After acceptance of this wave, all new activities for these accounts are entered in the target. The source remains read-only for historical lookup. Any source edit discovered after the final extract becomes a reconciliation exception.”

Then test common boundary crossings:

  • A contact belongs to an account that has moved and another that has not.
  • An opportunity changes account ownership during the overlap.
  • A user records a new note against an older activity.
  • An attachment arrives after its parent record was extracted.
  • A source record is merged, deleted, or reopened after the initial copy.

For each case, specify where the change happens, how it reaches the other system if needed, and how conflicts are resolved. “The team will keep both updated” is not a control unless it includes ownership, a queue, and a verification step.

A single cutover still needs a boundary, but usually for a shorter transition: the source stops accepting changes, the final changes are transferred, acceptance checks run, and the target becomes authoritative.

3. Filled hypothetical wave-planning worksheet

Fictional worked example: The following company, record counts, timings, and limits are hypothetical planning inputs, not customer results or product benchmarks. A business proposes migrating 180 account groups across three waves. Its migration team can allocate six person-hours per day to reconciliation during the overlap.

Each worksheet entry records the proposed group, related records, cross-wave dependencies, temporary boundary, acceptance evidence, and rollback condition.

Wave A: standalone account groups

  • Scope: 60 account groups, 240 contacts, 75 open opportunities, and their related activities, notes, and attachments.
  • Grouping rule: Each group includes all active account relationships and contacts needed for its open opportunities.
  • Dependency finding: Two proposed contacts also belong to accounts assigned to Wave B. Those connected account groups are reassigned to Wave B before the scope is frozen.
  • Temporary boundary: The target owns accepted Wave A records. The source retains a read-only copy. Other account groups remain writable only in the source.
  • Historical treatment: Historical activities move with their account group. Attachments that fail transfer remain exceptions; the group is not accepted until its required attachments are accessible and correctly linked.
  • Acceptance evidence: Expected IDs reconcile; required relationships resolve; all 75 open opportunities have the correct account and contact roles; selected histories are reviewed in chronological order.
  • Rollback condition: New target changes can be identified and transferred back without changing Wave B records. Until that path is demonstrated, Wave A is not isolated enough to launch.
  • Owner and overlap: A migration lead owns acceptance; a data steward owns exceptions. Planned overlap is two working days.

Wave B: connected account families

  • Scope: 40 account groups, including the connected groups removed from Wave A, with 360 contacts and 110 open opportunities.
  • Grouping rule: Parent accounts, subsidiaries, shared contacts, and their active histories move as connected bundles.
  • Cross-wave dependency: A consolidated management report combines Wave A and Wave B pipeline. It must not count both a source copy and a target copy.
  • Temporary boundary: Wave membership and authoritative-system status accompany the reporting extract. One named reporting owner checks that each opportunity appears once.
  • Acceptance evidence: Parent-child links and shared-contact relationships reconcile. Report totals are compared using the same inclusion rules, not just the same extraction date.
  • Rollback condition: Wave B can revert without rewriting Wave A account IDs or shared reference values. A required global remapping would invalidate this rollback plan.
  • Owner and overlap: The data steward owns relationship checks; the reporting owner owns consolidated totals. Planned overlap is three working days.

Wave C: historical and inactive account groups

  • Scope: 80 inactive account groups, 300 contacts, and closed opportunities with their histories.
  • Grouping rule: Records remain grouped by account context rather than being split into separate contact and activity batches.
  • Cross-wave dependency: An inactive contact may become relevant to an active account before Wave C starts.
  • Temporary boundary: A reactivation request triggers a dependency review. The connected group either moves early through the same acceptance process or remains together in the source until scheduled migration.
  • Acceptance evidence: Historical records retain their source identifiers, dates, relationships, and agreed status meanings. Closed records must not appear as newly opened work.
  • Rollback condition: Historical imports can be identified separately from new target activity. If they cannot, bulk removal would risk deleting valid work.
  • Owner and overlap: An archive owner signs off completeness. Planned overlap is two working days.

Worksheet decision: The revised waves are candidates for rehearsal, not approved production waves. Wave A depends on demonstrating change replay; Wave B depends on isolated identifiers and reporting rules; Wave C depends on a controlled reactivation path. Failure of any gate changes the sequence or the migration approach.

4. Price the reconciliation work, not just the transfer

Phased migration creates recurring work for as long as both systems remain operational. Estimate it from a rehearsal rather than assuming that smaller waves automatically mean less effort.

Use this planning calculation: daily reconciliation demand equals expected exceptions multiplied by average handling time, plus fixed verification work. Include investigation, correction, rechecking, and communication—not just editing the affected record.

In the hypothetical worksheet, suppose rehearsal produces 24 exceptions per day at eight minutes each, plus 90 minutes of fixed checks. That is 282 minutes, or four hours and 42 minutes, against six available person-hours. The remaining 78 minutes is limited headroom, not spare capacity that should immediately be assigned elsewhere.

If the exception count doubles, demand becomes 474 minutes, or seven hours and 54 minutes. The plan no longer fits the available capacity. Possible responses include changing the grouping, removing a repeated exception cause, shortening the overlap, adding trained reconciliation capacity, or reconsidering a single cutover.

Track both queue size and oldest unresolved exception. A stable queue can conceal one old relationship defect that blocks an important account. Define a maximum tolerable age by business consequence before launch, rather than accepting whatever backlog emerges.

5. Compare failure scenarios on the same terms

Compare the same defect under both approaches. Otherwise, teams often compare an ideal wave plan with a worst-case single cutover.

Hypothetical failure: contact relationships import incorrectly

Wave-based outcome: Fewer accounts may be affected initially, and later waves can pause. However, fixing shared contacts may touch records already accepted in an earlier wave. Rollback is only local if the relationships and subsequent edits are genuinely isolated.

Single-cutover outcome: The defect may affect the entire migrated population. If detected before release, the team can keep the source authoritative and delay opening the target. If detected after live work begins, recovery must also preserve that new work.

Hypothetical failure: the final change transfer is incomplete

Wave-based outcome: The team must identify missed changes for the affected wave while maintaining the boundaries for every other active wave. Repeated final transfers create repeated opportunities for the same defect.

Single-cutover outcome: There is one main final-change boundary to reconcile, but a failure may extend the interruption for everyone. The team needs a defined decision point for delaying release or restoring source operation.

Hypothetical failure: an automation runs twice

Wave-based outcome: Copies in both systems may trigger duplicate tasks or external actions if trigger ownership is unclear.

Single-cutover outcome: Duplicate actions remain possible during activation, but the team can use one coordinated source-disable and target-enable sequence.

For either approach, document which authorised person controls activation, what evidence confirms the old trigger is inactive, and how duplicate actions will be identified. Reversing data does not undo an external action already taken.

6. Prove rollback isolation with a rehearsal

A backup is useful, but it does not by itself demonstrate rollback. A workable rollback must preserve changes made since the migration and restore a coherent operating state.

  1. Establish identity. Retain source-to-target ID mappings and a durable way to identify wave membership.
  2. Create realistic changes. In a rehearsal environment, add a note, update an opportunity, change a relationship, and create a new record after migration.
  3. Stop writes. Rehearse who can pause edits, imports, and automated actions for the affected scope.
  4. Extract the changes. Identify what must return to the source and what needs manual resolution.
  5. Restore relationships. Apply changes in a dependency-aware order and check that references resolve.
  6. Check neighbouring waves. Confirm that accepted records outside the rollback scope remain unchanged.

If reverting one wave requires restoring the whole database, call it a whole-system rollback. Do not label it isolated recovery simply because only one wave caused the failure.

7. Make the sequencing decision with explicit gates

Choose waves when connected record groups can be bounded, authoritative ownership is unambiguous, measured reconciliation fits available capacity, and rollback does not disturb accepted groups. Give the overlap a planned end date and a rule for handling exceptions that threaten it.

Prefer a single cutover when most active records form one connected dependency group, shared edits cannot be controlled across systems, or repeated reconciliation would exceed the team's capacity. This choice still requires a rehearsed transfer, an acceptable interruption window, final-change checks, and a release decision owner.

Choose neither yet when dependencies are unknown, identifiers are unstable, or recovery cannot preserve new work. More waves will not repair an undefined record model; one cutover will not make an untested import reliable.

A limited rehearsal can clarify the decision without becoming the first production wave. Its purpose is to expose the boundary failures before users depend on that boundary.

8. Run one boundary test before approving the schedule

For the next planning session, select one proposed wave and complete the worksheet fields with real inventory data. Include a shared contact, a late activity, a reactivated account, and a post-migration edit in the rehearsal.

Record four results: unresolved cross-wave dependencies, reconciliation person-hours, oldest unresolved exception, and whether rollback changed anything outside the intended scope. Set acceptable thresholds before seeing the results.

Approve the wave only when its evidence meets those thresholds. Otherwise, regroup connected records, simplify the temporary boundary, or test a single-cutover plan. The schedule should follow the boundary the team can operate—not the other way around.

CRM migrationData integrityCutover planningReconciliation

Put the next conversation to work.

Explore Alactic Communications for WhatsApp conversations, campaigns and AI-assisted workflows. Choose the product and plan that fit your team.

Explore Communications