A cloud migration readiness assessment is often delivered as a slide deck with a maturity score. That is not useful on its own. The output you need is a decision: which applications move, in what order, with what changes, at what run cost, and what happens if the migration stalls halfway.
This guide sets out an eight-step assessment that a small or mid-sized business can complete without a large consulting programme. It assumes you already run production systems and cannot afford a long freeze.
Prerequisites
- A sponsor with budget authority. An assessment without a decision-maker becomes a document that nobody acts on.
- Access to inventory: server lists, application owners, vendor contracts, network diagrams and the billing data for what you run today.
- A baseline: current availability, incident history, peak load and the fully loaded cost of the existing environment.
- A stated constraint: the deadline, the budget ceiling or the risk that triggered the review. Write it down before the first workshop.
Step 1: Define the business case
State the reason for moving in measurable terms. "Reduce infrastructure cost" is not a case. "Move the order platform to a managed database so that failover no longer requires a manual restore" is a case. Common reasons include capacity limits, recovery time, end-of-life hardware, vendor support, remote access and the cost of maintaining a data centre room.
Record what would make the migration a failure: a cost increase after twelve months, an availability drop, a lost integration or a team that cannot operate the new platform.
Step 2: Inventory applications and dependencies
List every application, database, file share, scheduled job, integration and DNS record. For each, record the owner, the business function, the technology, the data classification, the users, the peak load and the recovery requirement.
Then map dependencies. The application nobody remembers is usually the one that breaks first: a reporting job that reads a database directly, a printer service, a payment callback that depends on a fixed IP address.
Step 3: Classify data and obligations
Separate personal data, financial records, credentials and public content. Note where each set is stored, who can access it, how long it is retained and whether any contract or regulation restricts its location. For Bangladeshi businesses, this includes the personal data protection law and any sector-specific requirements. A migration that moves data without checking these obligations creates a compliance problem that is more expensive than the hosting saving.
Step 4: Assess architecture fit
For each application, answer four questions:
- Is it stateless or does it store session data on the server?
- Does it depend on a fixed IP, a local file path or a specific operating system version?
- Can it run behind a load balancer and survive an instance restart?
- Does it have a supported upgrade path, or is it already out of support?
Applications that fail these questions can still move, but they need remediation before or during the migration. Estimate that work explicitly rather than hiding it in a contingency line.
Step 5: Model the run cost honestly
Compare the current total cost with the expected cloud cost, including compute, storage, network egress, managed databases, backups, monitoring, support plans and the staff time to operate it. Include the cost of the migration itself and the cost of running both environments in parallel.
Use a range, not a single number. Model a low, expected and high scenario for traffic and storage. The FinOps Framework is useful here because it treats cost as an operating practice with owners, not a one-off estimate.
Step 6: Assess skills and operating model
Cloud changes who does what. Someone must own identity and access, network rules, backup and restore, patching, cost review and incident response. If those responsibilities are unnamed, the migration will transfer risk to whoever answers the phone at night.
Decide the operating model before choosing the technology: in-house, managed service, or a split with clear boundaries. Write the escalation path and the response hours.
Step 7: Choose a migration pattern per application
- Retire: remove applications nobody uses. This is the cheapest win and the most often skipped.
- Retain: keep some systems where they are because the risk or cost of moving is higher than the benefit.
- Rehost: lift and shift with minimal change. Fast, but it carries existing weaknesses into a new bill.
- Replatform: move to a managed database or container platform while keeping the application largely intact.
- Refactor: change the architecture. Justified only when the business case depends on it.
- Replace: move to a SaaS product instead of migrating the application.
Record the chosen pattern, the reason and the estimated effort for each application.
Step 8: Build the plan and the rollback
Sequence the work by business risk, not by technical convenience. Move a low-risk internal tool first to prove the operating model, then a customer-facing system with a tested rollback. Define the cutover window, the data synchronisation method, the acceptance tests and the point at which you stop and revert.
A rollback plan that assumes the old environment is still available must say for how long and who pays for it.
Evidence pack
- Application inventory with owners and criticality.
- Dependency map, including integrations and scheduled jobs.
- Data classification and location constraints.
- Per-application migration pattern and effort estimate.
- Current and projected run-cost model with assumptions.
- Operating model, named roles and escalation path.
- Cutover, acceptance and rollback plan for the first wave.
Decision gates
- Gate 1: the business case is measurable and the sponsor accepts the success and failure definitions.
- Gate 2: the inventory and dependency map are complete enough to price the first wave.
- Gate 3: data obligations are understood and the target location is acceptable.
- Gate 4: the run-cost range and the operating model are approved together.
- Gate 5: the first wave has a tested rollback and a named on-call owner.
Common mistakes
- Pricing only compute and forgetting storage, egress, backups and support.
- Migrating an unsupported application unchanged and calling it progress.
- Leaving the operating model until after cutover.
- Testing the application but not the restore, the DNS change or the payment callback.
- Treating the migration as a project with an end date rather than a change in how systems are run.
Metrics
- Percentage of applications with a named owner and an agreed pattern.
- Actual versus modelled run cost after the first full month.
- Availability and recovery time against the agreed target.
- Migration wave duration and rollback rate.
- Number of manual operations removed by the new platform.
If you need an assessment that produces decisions rather than a maturity score, start with Managed Technology & Cybersecurity. The pricing page explains how assessment, migration and support are scoped. For the commercial side of the same decision, read how cloud vendor lock-in changes as you grow.
Sources
- AWS, Overview of the AWS Cloud Adoption Framework, accessed 17 November 2026.
- Microsoft, Cloud Adoption Framework for Azure, accessed 17 November 2026.
- NIST SP 800-145, The NIST Definition of Cloud Computing, accessed 17 November 2026.
- FinOps Foundation, FinOps Framework, accessed 17 November 2026.
Responses (0)
No comments yet. Add the first one below.
Leave a response
Your email address will not be published. Required fields are marked *.