Discovery ran for two weeks before anything moved. It surfaced four dependencies nobody had documented, including a licence server with no owner and a nightly batch job that read from a share three teams believed had been decommissioned. Finding those in week two rather than on cutover night is the entire value of doing this properly.
The lease terminated on schedule. Total planned downtime across all six waves was 42 minutes, measured at the application rather than the hypervisor, and no wave needed its rollback. Provisioning a new environment went from three weeks of tickets to under an hour of Terraform, and the monthly run rate came in 18% under the pre-migration datacentre cost once rightsizing was done.
Rather than one heroic weekend, we built a landing zone first and then moved workloads in six reversible waves, lowest risk first. Each wave had a rehearsed cutover, a rollback that had actually been tested, and a go/no-go call that Northbank's own team made. Aura now runs the platform under a managed contract.
