1. Why a migration must be planned as a project
An IT migration is the move of applications, data or infrastructure from one environment to another: from an on-premises server to the cloud, from one hypervisor to another, from an old management system to a new platform. It is often presented as a technical operation, but in practice it also affects people's daily work, supplier contracts, licences and obligations over personal data. For this reason it should be treated as a project, with written objectives, owners, dates and success criteria, rather than as a series of interventions carried out whenever there is time.
The most common risk is not the technical failure itself, but the late discovery of something that was not considered: an integration with another system, an overnight service, a certificate about to expire, a network printer connected to an application. A serious plan exists to make these dependencies visible before the cutover. It is worth discussing it with the people who know the systems before fixing the timetable.
2. Prerequisites: what must be true before you start
Before choosing a strategy some conditions are needed. A verified and restorable copy of the data must exist, not a presumed one. The scope must be approved by the people who use the systems, because they know which functions really matter. If any of these conditions is missing, the project stops at the first surprise, or proceeds on decisions taken by people who do not know the consequences.
The testing budget is the item most often cut, because it produces nothing visible until it is needed. Yet it is exactly what allows problems to be found before production. Ask your provider how it intends to demonstrate that the target environment behaves like the source, and include this test in the contract, not in a marginal note.
- A project manager with decision-making authority and a point of contact for each business area involved.
- A complete backup of the data to be migrated, with at least one successful restore test.
- A minimum inventory of systems, even if incomplete, to be completed during the work.
- A budget that includes test environments, verification time and a margin for rollback.
- A written agreement with the provider on responsibilities, maintenance windows and support.
3. Inventory and priorities: what to move, keep or retire
The inventory is the document that turns a vague intention into a verifiable list. For each system note the name, the owner, the users, the data processed, the integrations with other systems, licence expiry dates and the maximum tolerable downtime. Whoever knows a system rarely can tell you everything on their own, so the inventory should be built through interviews, network traffic analysis and reading the working processes, not only through existing documents, which are often out of date.
Once complete, the inventory must be classified. Some systems are critical and require a planned transition with agreed windows. Others are little used and can be retired, provided the data required by law or by the business is preserved. This exercise often reveals systems nobody uses any more but which continue to cost money, and single-user applications that keep an important process running. Both cases must be decided explicitly, not left to chance.
4. Choosing the strategy for each system
There is no single strategy for the whole migration. Each system is assigned to one of the possible paths, based on its value to the business, the cost of changing it and the technical risk. The table summarises the most common options with their practical meaning. The differences are covered in the dedicated guides, but the underlying criterion remains the same: how much the system changes and how much the way people work must change. For each system it is also worth noting who manages it today, because the choice often depends on that person.
A wrong choice costs more than a slow one. Moving quickly an application that should be rethought carries technical debt into a new environment. Rewriting a marginal system consumes months without proportionate benefit. Decide system by system, document the reason and review the choice if requirements change during the project. When the choice is uncertain, trial the system in a test environment before deciding.
| Strategy | What it means in practice | When it makes sense |
|---|---|---|
| Rehost | Moving the system as it is to a different infrastructure | Tight timelines, stable system, no need to change it now |
| Replatform | Moving it with minor adjustments, for example to a managed database service | You want to reduce maintenance without rewriting the code |
| Refactor | Substantially modifying or rewriting the application to exploit the new environment | The system is strategic and its limitations block the work |
| Replacement | Adopting an external product or service and moving the data into it | An adequate solution already exists and the current system is obsolete |
| Retirement | Decommissioning the system after archiving the data that must be kept | No process uses it any more or its functions are covered elsewhere |
5. The phases of the project, in order
The phases follow a precise sequence. Skipping a step does not save time: it moves the cost to the phase where fixing it is most expensive. The sequence below adapts to the size of the project, but the logic remains the same: first understand, then test, finally move. Each phase closes with a written approval from whoever is accountable, and with a maximum time within which to decide whether to proceed, not just a date.
The most delicate point is the transition between testing and production. Tests show that the system works under controlled conditions. The real transition introduces updated data, real users and different loads. For this reason verification after cutover matters as much as the initial tests and should be considered part of the migration. Also appoint a contact for anomalies, with a maximum time within which they must be reported.
- Complete the inventory and have the business owners approve the classification of systems.
- Assign each system a strategy and record the reason for the choice.
- Prepare a test environment that reproduces realistic data, integrations and user accounts.
- Carry out the first transition on a non-critical system and record every problem encountered.
- Test the data restore and the return to the source system before any cutover.
- Agree the transition window with the people who will use the system and with the suppliers involved.
- Carry out the cutover following a written list, with one owner for each step.
- Verify the result against measurable criteria and keep the old system available for a defined period.
6. Recurring risks and how to reduce them
Some risks recur in almost every migration. The first is the hidden dependency: a process that reads a shared file, an integration that uses a fixed address, a service account with a password written years ago. The second is data loss during synchronisation, especially when the old system continues to receive changes while the new one is being loaded. The third is a licence that does not follow the application, because it is tied to hardware or to a contract that was not renewed. Each of these risks has an early warning sign worth looking for during testing.
Reducing these risks comes down to simple checks. Compare counts and data samples between source and destination. List the integrations and test each one in the test environment. Verify that every licence is valid in the new environment before the transition date. These checks are not spectacular, but they are what prevents the most expensive surprises. Record the results in a single document, so that residual risk stays visible.
7. People, suppliers and communication
A migration changes how people work, even when the technical side remains invisible. People need to know when the service will be unavailable, what to do during the transition, where to find answers and whom to contact if something does not work. A clear notice sent in advance reduces many support calls, while a missing notice turns every small slowdown into an emergency. A point of contact for each group of users greatly simplifies communication.
With suppliers, the contract must describe who does what during the transition: maintenance windows, available support, responsibility for data and escalation procedures. Ask to see the supplier's rollback plan, not only the forward plan. If the supplier cannot explain how to go back, it is not yet ready for a cutover. Agree also how a problem during the transition will be communicated.
8. The deep dives: questions and reading order
This guide covers the method. For operational detail, read the deep dives in the migration cluster in this order. Start with the pre-migration checklist, which helps you complete the inventory and dependencies. Then read the comparison of application migration strategies, in order to assign a path to each system. If the project includes virtual machines, study the conversion between hypervisors before fixing the transition date.
The other topics answer specific questions. How to move mail and collaboration tools without interrupting communications. Whether it is better to keep files on-premises or in the cloud. How to transfer a database with the downtime you can afford. What to do with legacy applications: rebuild them, move them or retire them. How to build the cutover and rollback checklist. Finally, what to do after the transition, to verify the result and decommission the old environment safely.
9. How to start
The first action this week is simple: choose a system you consider important and ask the people who use it which processes depend on it, which integrations it has and what would happen if it were down for a day. Note the answers in a shared document. This short exercise already produces the basis of the inventory and shows how distributed knowledge is across different people. Write the outcome on a single page, with the decisions taken and the questions still open.
After the inventory, arrange a meeting with your IT supplier and a business manager to decide the scope of the first phase. There is no need to plan everything at once. What you need is to choose an initial group of low-risk systems on which to trial the method, and use what you learn to prepare the rest of the project. Close the meeting with an owner for the first phase and a review date.
Frequently Asked Questions
How long does it take to plan a migration?
It depends on the number of systems, the degree of integration between them and the availability of the people who know the applications. Do not fix transition dates before completing the inventory and tests, because dates imposed in advance tend to produce hasty decisions and improvised rollbacks.
Can I migrate everything in one go?
Usually this is not advisable. Migrating everything together increases the number of dependencies that can break at the same moment and makes it hard to work out which change caused a problem. A group-based approach, from the least critical systems to the central ones, lets you learn from the first transition and apply corrections to the later ones.
Who must approve the migration plan?
The plan should be approved by whoever in the company has the authority to accept the risk, together with the business owners of the systems involved and the IT supplier. If the project involves personal data, the data controller must check that the transfer is consistent with the record of processing activities and with the agreements with suppliers. For contractual details, consult a legal adviser.
Do I need an external provider, or are internal staff enough?
It depends on the complexity and on the time internal staff can devote. A project with many systems, sensitive data or tight downtime windows almost always requires external skills or shadowing. A useful provider does more than execute: it contributes to the inventory, proposes the test plan and documents the rollback. Assess its skills through concrete examples of similar projects.
What matters most when choosing a migration provider?
What matters most is the ability to explain the rollback plan, to demonstrate how it verifies data and to take on written responsibility for the critical phases. Also ask what access to your systems will be needed during the work, for how long and how it is revoked. A provider that talks only about speed but not about rollback should be treated with caution.
Does the migration require an assessment of personal data?
Often yes. Moving personal data to a new provider or a new storage location should be recorded in the record of processing activities and governed by appropriate agreements. A data protection impact assessment is required only in some cases, so check with your data protection adviser whether your project falls within them.