The Migration That Missed Twice
A financial services company budgeted a cloud migration at $14M over 18 months in 2023. The migration finished in 28 months at $26M and the cloud bill came in 35 percent higher than the on-premises bill it replaced. The board called the migration a failure. The CIO disagreed but could not produce a counterargument that survived scrutiny.
This is the modal cloud migration failure in 2026. McKinsey's enterprise cloud research consistently shows roughly 60 percent of cloud migrations deliver below their projected ROI (McKinsey, "Three actions CEOs can take to get value from cloud computing," 2024). The pattern is consistent. The migration picked the wrong R for most of the workload portfolio.
If your migration is being scoped today, the framework you pick determines whether you join the 40 percent that succeed or the 60 percent that struggle.
Coined Frame: The Six Rs Picked Honestly
Gartner's Six Rs framework (Rehost, Replatform, Refactor, Repurchase, Retire, Retain) is well-known. The framework is not the problem. The honest application of the framework is the problem. Most migration plans default to Rehost for everything because it produces the shortest plan. The shortest plan is rarely the highest-ROI plan.
R1 - Rehost (lift and shift). Move workloads to cloud with minimal changes. Fastest path to migration completion. Lowest ROI because cloud-specific cost optimization opportunities are missed. Defensible only for workloads scheduled for retirement within 24 months.
R2 - Replatform (lift, tinker, shift). Move with small optimizations (managed databases instead of self-hosted, containerization). Moderate effort, moderate ROI. Defensible for workloads with long planned lifespan but not core to differentiation.
R3 - Refactor. Rebuild for cloud-native architecture. High effort, high ROI when justified. Reserved for workloads that are strategic, will run for years, and benefit from cloud-native capabilities.
R4 - Repurchase. Replace with a SaaS product. Cost is license-based rather than infrastructure-based. Defensible when commodity functionality (CRM, HR, finance, marketing) is being self-hosted.
R5 - Retire. Shut down. The most overlooked R. Most enterprise portfolios contain 10-25 percent of workloads that nobody depends on and that can simply be turned off.
R6 - Retain. Keep on-premises. Defensible for workloads with specific regulatory, latency, or hardware requirements that cloud cannot meet cost-effectively.
The migration plans that succeed assign each workload honestly to the right R. The plans that struggle default everything to R1 or R3.
The 12-Month Reality
Cloud migrations consistently take longer than planned. The 28-month migration in the story above ran 56 percent over plan. This is closer to average than to exceptional.
Three patterns explain the overruns.
Pattern 1 - Discovery surprises. Teams underestimate what they will find during discovery. Undocumented dependencies, unused but apparently-critical systems, integrations nobody knew about. Discovery should run for two to three months before the migration plan finalizes, not in parallel with the migration.
Pattern 2 - Refactor scope creep. Workloads scoped as R2 (Replatform) turn into R3 (Refactor) during migration because the team sees opportunities. The opportunities are real but turn 6-week migrations into 6-month rebuilds. The scope discipline has to hold.
Pattern 3 - Cost surprise. Cloud cost estimation pre-migration is unreliable. Workloads that ran efficiently on-premises sometimes run inefficiently in cloud because the team did not redesign for cloud cost patterns. The cost optimization work has to happen during migration, not after.
The teams that hit their timelines build discovery, scope discipline, and cost optimization into the migration plan explicitly rather than treating them as parallel concerns.
What Has Changed in 2025
Three shifts affect modern migration ROI.
AI workload considerations. Migrations planned in 2022 did not need to account for AI workloads. Migrations planned in 2026 do. GPU capacity, inference cost patterns, AI-specific data flows all affect target architecture decisions.
Repatriation as a real option. Some workloads that were cloud-migrated in 2018-2021 are being repatriated to on-premises or colocation in 2024-2026 because the cloud cost did not match the workload pattern. The decision is workload-specific. The pattern shows that cloud is not always the cheaper option at steady state.
Sustainability requirements. EU CSRD and similar regulations require carbon reporting that affects cloud vendor selection and architecture decisions. The major cloud providers offer carbon-aware regions and reporting; most enterprise migration plans have not incorporated this yet.
The ROI Math That Works
Cloud migration ROI has four components. Reports that include all four are defensible. Reports that include one or two are not.
Component 1 - Infrastructure cost differential. Cloud cost versus on-premises cost. Usually positive (cloud cheaper) for variable workloads, often negative (cloud more expensive) for steady high-utilization workloads.
Component 2 - Operational cost differential. Headcount required to operate. Usually positive (cloud reduces operations headcount) but the savings are often offset by new specialized roles (cloud engineering, FinOps).
Component 3 - Capability enablement. New capabilities possible after migration that were not before. Auto-scaling, global deployment, AI service access, managed analytics. These are real but often overstated in pre-migration business cases.
Component 4 - Risk and resilience differential. Operational risk, security posture, disaster recovery capability. Usually positive for cloud but difficult to quantify in dollar terms.
The migrations that show positive ROI usually do so on Components 1, 3, and 4 combined, with Component 2 often net neutral. The migrations that struggle usually counted Component 2 savings that did not materialize.
What Logiciel Does Here
Logiciel works with engineering and IT leadership on migration planning, particularly on the R-assignment exercise that determines whether the migration produces ROI. The work is typically a discovery and assessment phase before any migration commits, followed by execution support on the highest-stakes workloads.
The Cost-Performance Tradeoffs framework covers the cost-side analysis that has to inform R-assignment. The Hybrid Delivery Model framework covers how external engineering capacity fits into migration execution.
A 30-minute working session is enough to evaluate your migration plan against the Six R framework and identify the workloads where the R assignment is likely wrong.
Frequently Asked Questions
Should I do a big-bang migration or phased?
Almost always phased. Big-bang migrations have higher failure rates and produce business disruption. Phased migrations let teams learn from early waves and adjust later waves accordingly.
How do I prevent scope creep during migration?
Explicit R-assignment frozen at the start of each migration wave, with change control if R-assignment needs to change. The discipline is uncomfortable but produces predictable delivery.
When should I consider repatriation from cloud?
For workloads with very high steady utilization (above 70 percent), predictable load patterns, and cost-sensitive economics. The math sometimes favors colocation or on-premises for these workloads. Repatriation is rare but legitimate for specific workload patterns.
How do I plan for AI workloads in a migration?
Add an explicit AI workload track to the migration plan. GPU capacity, inference patterns, AI data flows have different requirements than general compute. The target architecture should accommodate these explicitly rather than treating them as ordinary workloads.
What is the right team composition for a migration?
Migration program manager, cloud architecture lead, security architect, FinOps lead, and engineering leads from each migrating domain. Plus dedicated change management and communication capacity for organizational impact. Migrations fail more often on coordination than on technology. Sources: - McKinsey, "Three actions CEOs can take to get value from cloud computing," 2024 - Flexera, "2024 State of the Cloud Report"
