One migration owner
This person signs off record counts, exceptions and the cutover time. They maintain the source-to-destination ledger and are the only person who authorizes the full batch.
CONTROLLED CUTOVER
A migration is not finished when the upload ends. It is finished when dispatch, crews and the office agree on what happens next, and can recover if it goes wrong.
This person signs off record counts, exceptions and the cutover time. They maintain the source-to-destination ledger and are the only person who authorizes the full batch.
This person checks tomorrow’s work with dispatch and the crew leads. They decide whether schedules, addresses and access instructions are usable in the field.
Also name an administrator who can restore access and a communications owner who can stop duplicate notifications. Record their contact method before starting; do not leave the escalation plan inside the system being replaced.
| Gate | Evidence to collect | Do not proceed when… |
|---|---|---|
| 1. Preserve | Dated source exports, row counts, access permissions and a verified readable copy. | An export is incomplete or nobody can access the source. |
| 2. Rehearse | Small import, mapping ledger, duplicate outcomes and representative record checks. | Warnings are unexplained or relationships cannot be reconciled. |
| 3. Validate operations | Owner and crew complete a quote-to-job-to-invoice rehearsal with fictional or authorized test records. | A crew cannot find its job or office-only information is exposed. |
| 4. Freeze and reconcile | Written cutover timestamp and ledger of records changed since export. | Both systems are still accepting uncontrolled edits. |
| 5. Release | Tomorrow’s jobs checked, integrations intentionally enabled or disabled, rollback owner reachable. | Critical exceptions remain or customer messages might fire twice. |
Write a short operating note: “Until [time and timezone], edits belong in [source]. After approval, new work belongs in [destination]. [Owner] handles changes made during the freeze.” A parallel pilot should not mean two independent dispatch schedules.
Keep recurring jobs, deposits, open invoices and in-flight quotes on an exception list until each has a documented destination treatment. A migrated invoice balance must reconcile with payment records; an imported amount alone does not establish whether a customer paid.
Before customer communications begin, inspect scheduled reminders and automation in both systems. Test with approved recipients only. Disabled messages are safer than sending the same appointment reminder twice.
Examples: missing appointments, incorrect customer links, wrong balances, or inappropriate access. Set the decision deadline and name who can call a stop.
Prevent further divergence. Preserve logs and the failed batch mapping before changing anything else. Do not delete the evidence to make the dashboard look clean.
List new bookings, edits, cancellations and recorded payments. The migration owner determines how to carry legitimate changes back to the agreed operating system.
Confirm crew access and tomorrow’s schedule, then tell the team which system to use. Keep the incident and unresolved exceptions visible.
Rollback cannot unsend a message or reverse a real payment automatically. Those actions need their own authorized remediation. A demo reset is not a tested production backup or restore.
Use the staging schema before this checklist and the pilot scorecard for the operational rehearsal. Archive the signed review with the batch ledger rather than treating a green upload banner as final approval.
CrewViper is in private pilot. The public demo uses fictional data and resets on reload. It is not an importer, a backup test, or proof that a production integration is connected. Do not upload real customer data to the demo.