A practical checklist before you migrate to the cloud

Cloud migration rarely fails because of the technology. It fails because of decisions that were never made explicit — who owns what, what "done" looks like, and what happens the first time something breaks. Before your team moves a single workload, walk through the following.

1. Know why you're moving

"Everyone else is doing it" is not a migration strategy. Cost reduction, elasticity, disaster recovery, and faster releases all call for different architectures and different providers. Write the goal down in one sentence and test every later decision against it.

2. Inventory before you architect

You can't right-size what you haven't measured. Catalogue your workloads, their actual (not assumed) resource usage, their dependencies, and their compliance requirements. This single step surfaces most of the surprises that otherwise show up mid-migration.

3. Pick a migration pattern per workload, not per project

Rehost, replatform, and refactor aren't stages of the same journey — they're different tools for different workloads. A legacy system nearing retirement rarely deserves a rewrite. A core product that will run for the next decade often does.

• Rehost when the goal is speed and the application is stable.
• Replatform when small changes unlock meaningful savings.
• Refactor when the application's growth is currently capped by its architecture.

4. Decide how you'll roll back

Every migration plan we're asked to review has a cutover date. Fewer have a rollback plan. Define the conditions that would trigger a rollback and rehearse it before go-live, not during an incident.

5. Budget for the after, not just the during

Cloud costs are easy to underestimate once autoscaling, data egress, and monitoring tooling are added to the bill. Model a realistic first quarter of usage rather than a best-case one, and revisit it monthly until the pattern stabilizes.

None of this requires a large team or a long timeline — it requires making the implicit decisions explicit before the first workload moves. That's usually where we start when we work with a client on a migration: not with the architecture diagram, but with this checklist.