Cloud switching economics is broader than an egress price. A buyer can receive a clean export and still spend months rebuilding data contracts, reconciling records and validating downstream models. The EU Data Act provides a relevant policy and legal framework for switching data-processing services, but practical migration costs must be analyzed separately. This article introduces a transition ledger with explicit assumptions, separate legal and commercial questions, and a testable exit plan. Numbers are illustrations, not market benchmarks or legal determinations.
Identify the service before interpreting the rule
According to the European Commission’s Data Act overview, the regulation entered into force on 11 January 2024 and generally began applying on 12 September 2025. It aims to improve access, fair contractual terms and switching between data-processing services. The first procurement question is whether the relevant product and contract fall within a particular provision. A managed database service, a public cloud infrastructure plan and a separately licensed proprietary dataset are not interchangeable legal objects. Record the legal entity, hosting region, service inventory, contract renewal date and technical dependencies before discussing charges. The regulation text, Regulation (EU) 2023/2854, is the legal source; the Commission’s summary is guidance. Obtain jurisdiction-specific legal review where scope is disputed.
Separate four categories of migration expense
Build four ledger columns: provider-controlled export and transition charges; internal transformation and validation labor; temporary dual-running infrastructure; and demonstrable business interruption. A supplier can affect the first category directly, while the buyer may control much of the testing and parallel operation. Suppose a transition involves $1,000 in applicable provider charges, $9,000 of engineering work, $4,000 for duplicated environments, and a separately modeled service risk. These figures are illustrative and should never be quoted as ordinary EU cloud migration prices. Store invoices, time sheets, architecture assumptions and scenario IDs alongside each budget line. Otherwise procurement may bargain away an export fee while overlooking an engineering task ten times larger.
Normalize portability before comparing vendors
Portability begins with a representative export, not a promise in a sales deck. The destination team should inspect stable record keys, timestamp conventions, documented schema versions, encoding, access controls, audit logs and a reliable deletion procedure. A machine-readable CSV is insufficient if its columns are undocumented or if joins depend on proprietary identifiers. Include downstream queries and features that must be rebuilt, and distinguish the buyer’s data from vendor-owned software. Run a sample cutover in an isolated environment and record the observed reconciliation effort. A successful proof must show equivalent business decisions on both sides, or clearly document accepted differences. This technical evidence gives procurement a defensible basis for negotiating transitional support.
Model base, delayed and failed cutover scenarios
A single migration quote creates false precision. In a base scenario, all tests pass within the planned two-week dual-run window. In a delayed scenario, unresolved mapping errors require an additional week. In a rollback scenario, the buyer returns temporarily to the original provider and pays for revalidation. Assign scenario probabilities only where prior projects supply credible frequency evidence. Without that history, use low, central and high estimates instead of fabricating statistical confidence. Keep a separate incident register so the same outage is not charged twice, once as a contingency and again as a business-loss estimate. Finance should be able to reconstruct the assumptions when actual invoices arrive.
Calculate payback with visible assumptions
Imagine the incumbent costs $120,000 a year and the alternative costs $92,000 for comparable service. Annual nominal savings are $28,000. If a one-off transition costs $24,000, first-year savings are only $4,000 before quality and service differences. Simple payback is $24,000 divided by $28,000 per year, or around 10.3 months. If a failed cutover adds $12,000 of remediation, the result changes materially. These are hypothetical figures and ignore discounting, tax and unknown performance improvements. Put a sensitivity grid in the business case: vary destination price, migration duration and reconciliation effort independently. A procurement decision made solely on headline license savings may select a more expensive operating model.
Contract for an exit drill and evidence package
An effective contract specifies supported export formats, documentation, transition responsibilities, test windows, reasonable assistance, defect response, deletion evidence and escalation contacts. The buyer should practice moving a small real-world sample before the renewal deadline, using checksums and agreed acceptance queries. Keep records of failures that the supplier can fix and those originating in the customer’s migration scripts. Align the right to switch with data-licensing and intellectual-property rights; having technical access does not automatically authorize redistribution of third-party content. At renewal, compare the rehearsed exit costs with staying, not merely with a theoretical free migration. The objective is informed choice and operational continuity rather than an adversarial exit exercise.
The Commission’s framework changes the conversation around switching, but it does not eliminate every internal cost or contractual nuance. A service-by-service ledger turns legal rights into a controlled engineering exercise and makes both the savings case and downside risks visible to the finance team.
Sources
- European Commission, Data Act overview, updated 2 July 2026
- EU Regulation 2023/2854 (Data Act), adopted December 2023
