Something changed the economics, or the deadline
- A data centre lease, hardware refresh, support contract or licensing model has changed the economics of staying put.
- A platform, managed service or vendor is being discontinued, or a contract renewal needs a credible alternative behind it.
- Cloud spend is above plan and the estate needs replatforming rather than lifting again.
- Regulatory or sovereignty requirements mean the workload has to sit somewhere else.
- A carve-out, acquisition or divestment has a deadline attached.
A rehearsal before the real thing
Inventory of servers, applications, data stores, managed services and licences, with the real traffic flows and dependencies rather than the diagram from three years ago. Data volumes, change rates and egress get measured, because they set both the cutover window and the cost. Output: a disposition per workload, a cost model built on measured utilisation, and a wave plan with downtime budgets agreed per workload. Retirement usually funds a meaningful part of the programme.
Landing zone built or extended, connectivity established, identity and permissions translated to the target platform’s own model, and the security, logging and backup baseline in place before the first workload arrives.
One representative workload — not the easiest one — taken to production, proving the runbook, the cutover procedure, the rollback and the cost assumptions against reality before the full plan is committed.
Replication and data migration, database migration with the schema and performance work that comes with it, containerisation and pipeline changes where a workload is being replatformed, testing against the agreed targets, then a rehearsed cutover with a tested rollback at every wave.
The old estate switched off and the contracts closed, then rightsizing and commitment-based discounts applied against real consumption. This is where the business case is realised or lost, and it is the stage most migrations skip.
A different problem
Virtual machines port between providers. Almost nothing else does — managed databases, identity models, networking primitives, serverless runtimes and observability have no clean equivalents, and infrastructure code has to be rewritten rather than repointed.
The two line items that break these business cases are egress from the provider you are leaving, and the dual-running period while both platforms are live. We size both before you commit — and if the move does not pay back inside a defensible period we will say so at assessment rather than at wave four.
Obligations met during the move, not after it
Data residency and lawful transfer while data is in flight, DORA and NIS2 requirements, notification of changes to material outsourcing, and evidence that each cutover was controlled and reversible. Firms increasingly have to demonstrate that a provider exit is executable rather than merely documented — the discipline that makes a migration safe is what produces that evidence, so we build it during delivery instead of reconstructing it for an audit later.
Four ways in
Migration assessment
Fixed scope, producing the dependency map, dispositions, cost model and wave plan. Most clients start here, including those whose programme has already stalled.
Service or vendor replacement
A single platform service, CDN, WAF or managed database swapped out: compatibility review, parallel setup, traffic shifted progressively, rollback at the routing layer. Days or weeks, not a programme.
Migration delivery
Pilot wave at a fixed price, then wave-by-wave execution alongside your team, scaled with delivery partners where the volume calls for it.
Exit and portability review
What it would take, cost and break to leave your current provider — whether you are planning to move or need to evidence that you could.
Is your deadline realistic?
Send us what you need to move and the deadline you are working to, and we will tell you in one call.
- Inventory, dependency map, data volumes, egress and licensing position
- A recorded disposition and target design per workload, with service-equivalence mapping where the platform changes
- Cost model and wave plan with agreed downtime budgets
- Target environment delivered as infrastructure code
- Runbooks and tested rollback procedures per wave
- Decommissioning plan and post-migration optimisation
- Operational handover and team enablement