Cloud migrations fail less often because of the technology and more often because of what wasn't checked beforehand. This checklist is the pragmatic version — not exhaustive, but the questions that matter most for a mid-size team.
1. Assess
Before migrating, inventory what you actually run today — including the quiet dependencies nobody remembers configuring. Most surprises during migration come from systems nobody thought to list.
- List every workload, including internal tools no one thinks of as 'production'
- Map dependencies between systems, not just the systems themselves
- Identify what genuinely can't tolerate downtime
2. Plan
Decide what 'done' looks like for cost, performance, and security before you start moving workloads, not after. Without this, 'we migrated' and 'we succeeded' can quietly become different things.
3. Pilot
Choose a low-risk workload to prove the migration approach before touching anything customer-facing. The pilot should be boring on purpose — its job is to surface process problems, not handle drama.
4. Migrate
Migrate in phases with rollback plans, and validate each phase against real usage before moving to the next. Resist the urge to migrate everything in one cutover window.
5. Optimize
Once workloads are stable in the cloud, revisit cost and performance with real usage data. The right-sizing that matters happens after migration, not during it.
Written by
Vali LLC
Editorial Team
The Vali editorial team writes about staffing, cloud, and software delivery based on patterns we see across client engagements.