A migration business case that shows a saving is easy to produce and easy to disprove twelve months later. The discipline that makes the number trustworthy is including the costs that are invisible at the point of comparison.
What gets left out
- Egress. Moving data out is charged in a way that moving it in is not, and reporting or analytics workloads can make this the largest single line.
- Inter-zone and inter-region traffic between components you have deliberately separated for resilience.
- Storage growth. On-premise storage is a step cost every few years; cloud storage is a continuous line that rises with your data whether or not anyone reviews it.
- Backup retention, which is separately charged and separately growing.
- Non-production environments left running outside business hours.
- The support plan tier you will actually need, which is rarely the free one.
Reserved capacity is a forecast, not a discount
Committing to one or three years lowers the unit price substantially, and it converts a variable cost into a fixed one. That is a good trade if your forecast is sound and a poor one if the workload is being re-architected. Do not commit capacity for systems you intend to change.
Lift and shift is a decision, not a failure
Re-architecting during a migration doubles the risk surface: if something breaks you cannot tell whether the cause was the move or the redesign. Moving as-is, stabilising, then optimising is slower on paper and usually faster in practice. It also gives you real consumption data to optimise against rather than estimates.
Where the money is actually recovered
Genuine savings tend to come from decommissioning, not from the cloud being cheaper per unit. The migration is the only time an organisation systematically inventories what it runs, and the discovery that a material share of servers support nothing anyone can name is common. Capture that saving explicitly, because it is the part of the business case that is real.
Sovereignty and latency are constraints, not preferences
Where data must remain in-country, or where a system serves users over links that make round-trip latency material, those constraints define the architecture before cost does. We have delivered on-premise and hybrid because that was the correct answer, and we will say so before a tender rather than after.
Sources and further reading
External links, provided for verification. CyberEdge is not responsible for third-party content.