Background
A mid-sized Australian deposit-taking institution hosted the majority of its estate in a data centre scheduled for closure: banking adjacents, payments orchestration, ledger integration, document management and analytics. The exit date was fixed by the facility, and the executive required staged migration over a single cut-over.
The engagement covered landing-zone design, wave sequencing, residency engineering and delivery leadership. The institution's engineers delivered alongside the firm.
Work performed
- Six weeks of dependency recovery: network flows, scheduled tasks, operator interviews and month-end tracing. Nineteen undocumented integrations were found; three were retired as unused.
- Landing zone built as code: identity, network segmentation, guardrails, logging and cost policy, reviewed by the bank's risk function before the first workload.
- Residency engineered at platform level: Australian regions for customer data, disaster recovery within Australian borders, key custody held domestically.
- First three waves selected for low failure impact, each with a rehearsed rollback before its live window.
- Two full month-end closes run in parallel against the migrated estate before the ledger cutover; eleven discrepancies found and fixed in rehearsal.
- Source estate decommissioned on completion.
Outcome
The estate exited the facility three weeks ahead of the deadline, and the year-end close ran on the new platform without material incident. Run cost fell 38% against the combined baseline of the old facility and its licences. The dependency graph is now maintained as a standing artefact by the bank's architecture function.
Two waves slipped: one by a fortnight on an application control issue, one on a vendor licence misunderstanding that predated the engagement. Both were absorbed within the schedule buffer.