DTC CRM migration playbook

1. Data mapping

• Profile fields: every attribute, where it is stored, how it is computed, and whether it is still needed.
• Events: event names, payloads, timestamps, and how they map to the new schema.
• Lists and segments: rebuild logic, not just import CSVs.
• Message history: what is portable and what is not.

2. Identity and consent

• The canonical customer ID for each contact, and how duplicates are resolved.
• Consent state per channel, per region, with timestamp and source.
• Suppression lists and global unsubscribes must be preserved and auditable.

3. Event continuity

• Historical events: backfill where possible, document gaps where not.
• Real-time events: map every stream to the new ingestion point.
• Triggers: identify every flow that fires on an event and test it before cutover.

4. QA and parallel run

• Profile counts match within a tolerance.
• Segment membership overlaps with the legacy system.
• Test flows fire at the right moments with the right payloads.
• A small live segment runs in both systems for a validation period.

5. Cutover

• Hard cutover: switch all flows on a defined date. Fastest, highest risk.
• Phased cutover: move one lifecycle moment at a time. Slower, easier to debug.
• Rollback plan: what triggers it, who executes it, how long it takes.

6. Risk register

7. Measurement plan

• Baseline cohort behaviour 30 days before cutover.
• Holdout group kept on the old system for comparison, if feasible.
• Daily reconciliation for the first 14 days, weekly for the first 90.
• Attribution limits documented: what is measured, what is inferred.

When not to migrate

Do not migrate if the underlying data layer is still broken. A migration will not fix identity, consent or event gaps. Fix those first. Do not migrate to escape a problem that lives in strategy or operations rather than tooling.