Cloud migration has become a strategic priority for the majority of Belgian SMEs. Yet many businesses approach this project without a clear methodological framework, leading to budget overruns, avoidable service interruptions and partial adoption that undermines the expected return on investment.

At ITOPS.be, we have been guiding SMEs of 20 to 250 employees through Azure, AWS and GCP migrations for several years, using a two-step approach: a Blueprint (scoping and design) followed by a Build (phased execution). Here is what we have learned on the ground.

Why Azure for Belgian SMEs?

Microsoft Azure operates data centres across Western Europe — notably in Amsterdam and Dublin — enabling compliance with the data residency requirements imposed by GDPR. For Belgian SMEs that process personal data of European clients, this localisation is not a detail: it is a contractual and regulatory obligation.

Beyond compliance, Azure offers native integration with the Microsoft 365 tools most of your team already uses: Exchange Online, Teams, SharePoint, OneDrive. This continuity reduces the learning curve and accelerates adoption across the organisation. That said, Azure is not a foregone conclusion — we return below to the Azure vs AWS vs GCP choice depending on your context.

The three migration models

The choice of model drives the cost, timeline and risk of the entire project. Three broad approaches coexist, and most SMEs combine all three depending on the application.

1. Lift-and-Shift (rehosting)

You move your physical servers or VMware VMs to Azure Virtual Machines without modifying the applications. This is the fastest and least risky model in the short term. Ideal for proprietary line-of-business applications or legacy ERPs whose source code you do not control.

Limitation: you do not benefit from native cloud advantages (elasticity, automatic high availability, managed services). You often pay more than in a data centre for the same performance, because you are importing an architecture designed for on-premise hardware. Rehosting is therefore a good starting point, rarely a destination.

2. Replatforming

You retain the application logic but adapt certain components to leverage Azure managed services. For example: replacing your dedicated SQL Server with Azure SQL Database, or migrating your scheduled jobs to Azure Functions.

This is the right balance for most SMEs: improved performance, lower operational costs (no more patch management, automated backups), without a full rewrite. The effort-to-benefit ratio is generally the most favourable of the three.

3. Refactoring (cloud-native)

You re-architect the application for execution in containers (Azure Kubernetes Service) or in a serverless architecture. Recommended only for applications with highly variable traffic, differentiating building blocks or new strategic applications. Refactoring delivers the best long-term gains in elasticity and cost, but requires real development investment and cloud-native skills in-house or with your partner.

Cost estimation and FinOps: what nobody tells you

The migration itself is rarely the largest budget item. What surprises SME leaders is the monthly running cost once the migration is complete.

Here are the cost categories to anticipate:

Compute (VMs): on the Lift-and-Shift migrations we run, typically the single largest line item on the monthly bill
Storage and backups: often underestimated, especially if you have significant file volumes (accounting data, archives, CAD files)
Data egress: Azure charges for outbound bandwidth. If your applications transfer large volumes externally, this item can become significant
Support and licences: Azure Hybrid Benefit allows you to reuse your existing Windows Server and SQL Server licences — always check your eligibility before budgeting

The discipline that makes the difference over time is FinOps: the continuous financial governance of your cloud resources. In practice this means rightsizing (matching VM sizes to actual observed usage), Reserved Instances or Savings Plans (1 to 3-year commitments that often cut the compute bill by 30 to 60% depending on the profile), automatically shutting down test environments overnight and at weekends, and systematic tagging to attribute every cost to a service or project.

Our recommendation: use Azure Pricing Calculator for an initial estimate, then refine with Azure Cost Management after 30 days of real-world operation. Precise figures depend heavily on your workload profile — be wary of any flat-rate estimate offered before the discovery phase.

GDPR and data residency

For a Belgian SME, GDPR compliance is not optional. The basic rule: host personal data in cloud regions located within the European Union. Azure offers several EU regions (West Europe in Amsterdam, regions in France, Germany and others); AWS and GCP provide equivalents. During scoping, identify which data is concerned, in which region it will reside and who accesses it.

Beyond the region choice, document any transfers outside the EU, sign a data processing agreement (DPA) with your cloud provider, and enable encryption at rest and in transit. These elements belong in your record of processing activities. For regulated sectors (healthcare, finance), additional sector-specific requirements may apply — to be verified from the Blueprint stage.

Zero-downtime strategies

Migrating without halting operations is the strongest expectation among business leaders. Several techniques make it possible.

Blue-green deployment keeps the current environment (blue) and the new Azure environment (green) running in parallel. You switch traffic over in one go once green is validated, and you retain the ability to revert instantly to blue if a problem arises. Canary deployment refines the switchover: you first direct a small fraction of traffic (5 to 10%) to the new environment, observe the metrics, then increase progressively.

For databases, continuous replication with a synchronisation mechanism reduces the cutover window to a few minutes. In every case, a documented and tested rollback plan is non-negotiable: knowing how to revert cleanly is worth more than hoping everything goes well.

Phased migration plan

We structure every migration into four phases, following the Blueprint-then-Build logic:

Discovery and assessment (2 to 4 weeks) — a complete map of your estate: application dependencies, open ports, data volumes, peak loads. Without this step, you are migrating blind. We use Azure Migrate to automate this discovery.
Proof of Concept — migrating a limited, non-critical scope to validate the target architecture, the real costs and connectivity.
Migration in waves — each wave groups applications with close dependencies, with its own rollback plan. Each wave is validated before moving to the next.
Post-migration optimisation — rightsizing, Reserved Instances, governance.

A recurring pitfall: underestimating connectivity. A consumer internet connection at 50 Mbps may suffice at the start of migration, but will quickly become a bottleneck once most of your workloads are in the cloud. Plan for scaling (ExpressRoute, higher bandwidth) from the planning phase.

This phased approach allows you to maintain business continuity and validate each step before proceeding to the next. It connects directly with your network infrastructure, which often needs modernising in parallel.

Post-migration operations

The migration is not an end in itself. Once in the cloud, you must establish continuous operations: monitoring (Azure Monitor, proactive alerts), patch management on remaining VMs, backups tested through regular restores, and a monthly FinOps review of the bill.

It is also the moment to harness what the cloud makes possible: the automation of repetitive tasks and agentic AI workflows to relieve your teams. Azure without governance is a bill that can spiral within weeks — establish from day one budgets with alerts, Azure Policy to restrict allowed regions and resource types, and a consistent naming convention.

For scoping and implementation, discover our Cloud & DevOps services.

Frequently asked questions

How much does an Azure migration cost for an SME?

The cost breaks down into two parts: the one-off migration project (scoping, execution, support, which depends on the number of applications and the chosen model) and the recurring monthly Azure bill. For an SME of 20 to 250 employees, compute is usually the dominant share of that monthly bill. Any serious estimate requires a prior discovery phase — be wary of figures quoted blind.

Lift-and-shift vs refactor — which is right?

Think of it as an effort-versus-payoff slider. Lift-and-shift asks little and delivers fast, but leaves cloud savings on the table: great to get started, or for frozen legacy applications. Refactoring sits at the other end — lasting gains, but a development budget to match. In practice, most of our SME clients set the slider in the middle (replatforming) for the application core and refactor only the building blocks that genuinely set them apart.

Will migration cause downtime?

With a wave-based approach and techniques such as blue-green or canary, the interruption window can be reduced to a few minutes, or even zero for applications designed for it. The condition is a documented and tested rollback plan. A "big bang" migration without a rehearsal is the leading cause of avoidable downtime.

Azure vs AWS vs GCP for a Belgian SME?

If you are already on Microsoft 365, Azure offers the best continuity and the Azure Hybrid Benefit on your existing licences. AWS remains the broadest service catalogue and suits cloud-native architectures. GCP excels at data and machine learning. For a Belgian SME, the decisive criterion is often not technical but operational: the skills available and data residency in an EU region. We stay neutral and recommend based on your real context.

If you are considering an Azure migration for your SME, contact us for a complimentary discovery audit. We will provide a map of your current infrastructure and a realistic cost estimate before any commitment.