Lift and shift vs re-platform vs re-architect: choosing the right cloud migration strategy
Choose re-host (“lift-and-shift”) when you need the fastest, lowest-risk path to exit a data-centre lease; re-platform only the middleware and database layers (“re-platform”) when you want 15–30 % cloud cost savings without touching core code; re-architect into microservices only when the business case shows ≥40 % agility gain or ≥50 % infra-cost reduction within three years. Gartner’s 2025 CIO survey shows 68 % of Southeast Asian enterprises that pick the wrong pattern overspend by an average USD 1.8 M and add 9.5 months to ROI breakeven.
What is “lift and shift” (re-host) and when does it make financial sense?
Re-host, better known as lift-and-shift, moves VMs or bare-metal images into cloud IaaS with ≤5 % application change. According to AWS’s 2026 economic impact report, enterprises complete re-host projects in 4–8 weeks and cut infra TCO by 11 % on average—ideal when a Jakarta bank needed to exit two co-lo racks before a 30 % rent hike in 2025. TechNext Asia helped the bank move 193 VMs to Amazon EC2 using CloudEndure; the only code change was a DNS update, saving USD 420 k in hardware refresh.
Because lift-and-shift adds no cloud-native features, only pick it when (1) technical debt is low, (2) the app is slated for retirement within three years, or (3) regulatory deadlines (e.g., Malaysia’s PDP Act data-locality rules) trump optimisation. Our cloud migration assessment guide contains a 19-question checklist that scores apps ≥70/100 as re-host candidates.
How is re-platform different and why do 42 % of ASEAN workloads land here?
Re-platform (a.k.a. “lift-tinker-and-shift”) swaps legacy runtime components for managed cloud services—think SQL Server to Azure SQL or WebLogic to Amazon ECS Fargate—without altering business logic. IDC’s 2025 Southeast Asia Cloud Pulse shows 42 % of migrated workloads adopt this middle path, capturing 27 % infra savings plus 38 % faster patching cycles. Thai retailer Central Retail replaced Oracle RAC with Azure PostgreSQL Hyperscale, removed 14 000 Oracle core licences and trimmed five FTEs, freeing USD 1.1 M annually.
Pick re-platforms when apps need quick scalability but code is too brittle for full refactor. Typical changes: switch from file to Amazon S3, swap JMS queues for Azure Service Bus, containerise with no micro-service split. ROI breakeven lands in 6–9 months—half the time of re-architect. The key trap is scope creep; lock the exit criteria in a service blueprint before the first VM is shut down.
What does it really mean to re-architect—and why do only 13 % succeed?
Re-architect (refactor) turns a monolith into cloud-native services—usually micro-front ends, serverless components, and polyglot data stores—to unlock elastic scale and DevOps velocity. Gartner’s 2026 app-modernisation report finds only 13 % of ASEAN re-architect projects hit planned ROI; the rest overrun by 160 % because of underestimated data-consistency and testing effort. Singapore insurer NTUC Income rebuilt its core policy suite into 52 Kotlin micro-services on AWS Lambda, cutting release cadence from quarterly to 47 times a day and saving USD 2.4 M yearly in unplanned outage cost—yet the 22-month programme spent USD 8 M before first dollar of benefit.
Use re-architect when (a) product roadmap demands ≥3 releases a month, (b) peak-to-trough utilisation varies >10×, or (c) you need multi-region active-active for compliance (MAS TRM guidelines). Mandatory enablers: automated testing ≥80 % coverage, a domain-driven design map, and product owners who can fund a 12-18 month runway. Anything less and you should re-platform first, then refactor in increments—exactly the strangulation pattern we outline in How to Build a Software Modernisation Strategy (2026).
Decision matrix: how do risk, cost, and value line up across the three styles?
| Dimension | Re-host | Re-platform | Re-architect |
|---|---|---|---|
| Avg duration (Gartner 2025) | 6 wks | 4 mths | 14 mths |
| Capex vs on-prem baseline | –11 % | –27 % | –55 % (yr-3) |
| Failure rate SEA | 8 % | 18 % | 67 % |
| Skills needed | VMware, scripting | Containers, managed PaaS | Micro-services, CI/CD, Chaos Eng |
| Regulatory risk | Low | Medium | High (data-residency splits) |
Run each candidate application through a 100-point score: business lifespan (20), change frequency (15), scale variance (15), compliance complexity (15), budget runway (20), and engineering skill (15). Apps scoring <50 → re-host; 50–70 → re-platform; >70 → re-architect. TechNext Asia’s Cloud Migration Canvas automates the scoring in a 90-minute workshop—clients such as PetroVietnam refined 212 apps into three waves in a single afternoon.
Which ASEAN compliance traps appear in each model?
MAS, BI, and BIMA regulations turn seemingly technical choices into board-level risk. For example, Indonesia’s BI requires disaster-recovery failover ≤30 km if customer data > IDR 5 B is involved—lift-and-shift to Jakarta-only AWS AZs can breach that rule unless you also turn on cross-region replication (doubling spend). Re-platform data-modernization playbook shows how to map data-class to sovereignty tier before selecting the target pattern.
Re-platform projects often overlook encryption-at-rest defaults: Thai PDPA considers unencrypted personal data in object storage a “control failure” even if the CSP encrypts disks underneath. Re-architect efforts face cross-border data flow: Malaysia’s PDP Act needs Minister consent to move personal data out of the country—so architecting a regional analytics lake in Singapore may need tokenisation or sovereign DB services like Azure SQL Malaysia. Budget 6–8 weeks for regulatory design in each model; skipping this step is the #1 reason cloud migrations fail in Southeast Asia, according to Yugabyte’s 2026 failure analysis.
How do you finance the journey—spend vs. lease vs. cloud-native SaaS?
Re-host fits existing CapEx budgets because you can capitalise lift licences and treat cloud as “rented hardware.” Re-platform blurs OpEx: Gartner notes 54 % of CFOs reclassify managed PaaS as OPEX from month-1, hurting EBITDA ratios unless offset by parallel hardware depreciation drops. Re-architect forces full OpEx but unlocks consumption economics—Vietnamese e-commerce leader Tiki budgeted USD 0.08 per shipped order after moving to Lambda vs. USD 0.21 on VMs, enabling pay-as-you-grow during 11-11 sales spikes.
Use cloud vendor credit programmes: AWS Migration Acceleration (MAP), Azure Migrate & Modernise, and Google Cloud Rapid Assessment fund up to USD 3 M each in partner services. TechNext Asia clients captured USD 11.4 M in such credits during 2025, cutting net migration cost by 32 %. Layer in green financing—HSBC’s Sustainable Tech Loan discounts 25 bps if projected carbon drops >20 %, common in re-platform and re-architect due to shared tenancy.
Real-world ASEAN case studies: three enterprises, three patterns, three outcomes
Re-host: Malaysia’s Mayflower Airways moved Navitaire NewSkies reservation system to AWS in 5 weeks to exit a Cyberjaya co-lo lease expiring 30 June 2025. Zero code change, 18 % infra saving, and compliance with CAAM cloud security supplement. Payback: 3 months.
Re-platform: Vietnam’s largest spice exporter, Hapro, shifted SAP ECC 6.0 from IBM Power to Azure Mv2 series and swapped SAP ASE to Azure SQL Managed Instance. Duration: 11 weeks; database licence reduced 41 %; batch jobs that took 6 hrs now finish in 1.8 hrs using Azure Premium SSD v2. CFO reported USD 680 k annual saving, funding a future S/4 conversion.
Re-architect: Philippine fintech PayMaya split its monolithic wallet into 71 Golang micro-services on Google Cloud Run, backed by Cloud Spanner. Results: 99.99 % uptime (from 99.5 %), P99 latency 120 ms (was 1.8 s), and PCI-DSS audit time cut by 45 %. Cost neutral for 14 months, then 48 % infra reduction once traffic doubled—proving the business case.
Frequently Asked Questions
Which strategy delivers fastest ROI for regulated banks in SEA?
Re-platform hits breakeven fastest—8.2 months on average—because managed database and middleware services remove licence and patching cost without waiting for micro-service redesign. MAS and BI both accept lift-and-shift for legacy systems provided you add guardrails like WAF, VPN, and log shipping within 90 days.
Can I start with lift-and-shift and refactor later without double billing?
Yes. Use “strangle” traffic routing: keep the monolith in one VPC, peel off micro-services behind an API gateway, and shift traffic %-based. AWS VPC Lattice, Azure API Management, or Google Cloud Cloud Service Mesh let you migrate in 10 % slices, so cloud cost rises linearly with value, not twice.
How do we estimate cloud egress before choosing a pattern?
Model user personas: employees (intra-VPC), customers (internet), regulators (VPN), and partners (API). Run a 30-day PCAP in your DC; multiply bytes by CSP egress fee (AWS: USD 0.085/GB, Azure: USD 0.087, GCP: USD 0.08). Re-architect with edge caching (CloudFront, Azure Front Door) can cut egress >60 %.
Is containerisation mandatory for re-platform?
No. Containers help portability but swapping WebLogic for Azure App Service or VMware to Amazon RDS Custom achieves similar savings. Pick containers only if you need CI/CD portability across clouds or your dev teams already use Kubernetes; otherwise managed PaaS is faster.
What organisational metrics prove we’re ready for re-architect?
Target three KPIs: (1) automated test coverage ≥80 %, (2) mean time to restore (MTTR) <1 hr, and (3) product squad release frequency ≥1 per week. When all three hold for six consecutive months, re-architect risk drops to <25 %, per TechNext Asia’s 40-client dataset.
Ready to pick the right migration pattern? Contact TechNext Asia at https://technext.asia/contact for a complimentary Cloud Migration Canvas workshop and unlock vendor credits up to USD 3 M.