
Data migration: a plan that loses nothing
A data migration rarely fails on the copying. It fails on what nobody wrote down: a folder with unusual permissions, a spreadsheet that points to a drive letter, the history the old package refused to export. This is the order we follow when a file server or a legacy application moves, and how you show afterwards that nothing was left behind.
A data migration moves more than files: permissions, paths, links and history go with them. Whatever you do not inventory, you discover when someone goes looking for it.
Clean up before you move, with a business owner for every folder. A migration is the cheapest moment to delete old data, and the only moment anyone is looking at it anyway.
You prove nothing is missing with counts, checksums and a sample checked by the owner, and the old system stays read-only until that proof is signed off.
What a data migration actually moves
Data migration means moving information from an old system to a new one: a file server to a new server or to the cloud, or the records from an unsupported package to its successor. The word suggests it is about files. That is roughly half of it.
The other half is everything attached to those files:
- Permissions. Who may open which folder, and often who may not: payroll, HR records, the board folder.
- Paths. Drive letters, shortcuts, spreadsheets linked to other spreadsheets, templates that pull a logo from the server.
- Integrations. A scanner that saves to a share, an application that keeps its attachments on the file server, a script that collects something every night.
- History. Previous financial years, closed files, earlier versions of a document.
A plan that only counts files works on cutover day and fails in the weeks after, each time for a different colleague. Mailboxes and Teams are out of scope here; how those move is covered in our Microsoft 365 migration playbook.
Clean up first, then move
Every file server that has run for ten years carries folders nobody opens, duplicate copies of the same project and the personal folder of someone who left long ago. Moving all of it is the safe choice that turns out most expensive: the migration takes longer, the new storage is full on day one, and the clean-up is postponed for another decade.
Give every top-level folder an owner in the business, not in IT. IT can see how large a folder is and when it last changed; only the department knows whether it is still needed. For each folder the owner decides one of three things: move, archive (kept read-only) or delete.
Two things limit the deleting. Financial records fall under HMRC record-keeping rules and company law, so what is covered goes to the archive, not the bin. And UK GDPR says you should not keep personal data longer than you need it. A migration is the moment those two rules land on the table together, often for the first time.
Permissions and paths: where file server migrations break
Any tool can copy files. The problems sit around them, and you can find them in advance if you look:
- Permissions granted to individuals rather than groups. On the new platform that becomes a hundred separate exceptions. Convert them to groups before you migrate, and a structure moves instead of a tangle.
- Folders the administrator cannot open. There are almost always some. A migration that skips them reports no error; they are simply missing.
- Paths that are too long and characters the target rejects. A cloud platform has different limits from a Windows server. A scan beforehand produces the list, and you resolve it with the owner, not quietly with a script.
- References to the old location. Linked spreadsheets, macros, desktop shortcuts and scan profiles. Inventory them with the departments; they only break when someone opens them.
What you find here shapes the schedule more than the number of terabytes does.
The legacy package: exporting is not yet migrating
When a package is retired, such as an old accounts system, an order application or sector software, the data does not live in a folder but in the database behind it. The question is then not how to copy files, but what the new software can read and what the old one can hand over.
- Ask the vendor of the old package which exports exist, in which format and with which fields. An export of open items only is very different from one with the full history.
- Ask the vendor of the new package what it can import and who does the conversion between the two. That is their job or an implementation partner’s; it is not IT support.
- Decide what stays behind. Often only master data and open items move across, and the history stays in the old system.
Then that old system has to remain readable. A package that only runs on a server you want to cancel is an archive with a sell-by date. Choose in advance: an isolated virtual machine kept running read-only, or a complete export to a format you can open without the package. What HMRC or your accountant expects from that is a question for them, not for us.
How you prove nothing is missing
“It all seems to be there” is not a check. A data migration is only finished when you can show it, and you build that proof in three layers:
- Counts. Number of files and folders and total size, per top-level folder, on both sides. A difference is not always an error (deliberately deleted), but every difference has an explanation.
- Checksums. A hash per file, before and after. That proves the content matches, not just the name. The migration tool should also log what was not transferred, and why.
- A sample checked by the owner. Ten files they know, opened in the new location, with the right permissions. For a package: the same balances, the same open items and the same customer record in the old and the new system.
Record all three in a short report the owner signs off. That document is also what you reach for when someone says six months later that something is missing.
Cutover day: freeze, catch up and keep a way back
You do not move a large file server in one night. What works: a first full copy well before cutover, while everyone carries on working, followed by one or more delta passes that move only what has changed. On the day itself only the final delta is left.
- Freeze the source. From the agreed moment the old location is read-only. Anyone still saving there is working on a copy that will not come along.
- Test the backup before cutover, and we mean a restore test, not a green tick. If the new platform is not yet in the backup, the old server is your only copy.
- Agree a rollback, with a criterion: what has to go wrong before you go back, and who decides. A rollback you only think of when things go wrong is not one.
- Keep the old environment, read-only, until the report from the previous step is signed off and the first month-end close on the new system has worked.
Where our work stops
What we do. We inventory the file server, convert permissions to groups, scan for paths and references, migrate the files with checksums and a log, set up the new platform and its backup, and keep the old environment available read-only for as long as needed. We are ISO/IEC 27001 certified, and that is how we handle the data that passes through our hands on the way.
What we do not do. We do not convert data inside the database of a third-party package; that is done by the software vendor or its implementation partner, and we provide the server, the access and the planning for it. Nor do we decide which records you must keep or for how long: that is a question for your accountant and, where personal data is involved, your data protection adviser. And whether a folder can go is decided by its owner, not by us.
Questions we get about this
The five that come up most often.
What is data migration?
Data migration is moving information from an old system to a new one, for example from a file server to the cloud or from a retired package to its successor. Alongside the files, permissions, paths, integrations and history move too. Those four decide whether the migration still causes questions in the weeks after cutover.
What are the steps of a data migration?
Inventory and clean up with an owner per folder, prepare permissions and paths, make a first full copy, run delta passes, and on cutover day freeze the source and move the final changes. Then verify with counts, checksums and a sample checked by the owner. The old system stays read-only until that check is signed off.
How do you prevent data loss during a migration?
By copying the source rather than changing it, running a backup restore test before cutover, and comparing a checksum per file before and after. The migration tool should record what was not transferred and why. The old environment is only switched off once the owner has confirmed their data is complete.
What happens to the history in an old package?
Often only master data and open items move to the new package, and the history stays in the old one. That old system then has to remain readable, either as an isolated read-only environment or as a complete export you can open without the package. How long that applies is something to settle with your accountant.
Can you convert the data in our old accounts package?
No. Converting data inside the database of a third-party package is done by that software’s vendor or its implementation partner. We handle what is needed around it: the server, the access, the backup, the cutover planning and keeping the old system available read-only.
Further reading
What a data migration connects to.
Shall we map your file server first?
We scan permissions, paths and references and give you an overview per top-level folder to go through with the owners. Then you know what moves, and what breaks if you leave it unresolved.
Practical IT knowledge in your inbox
New guides on management, security and the workplace, written by the people doing the work. No sales talk, and you can unsubscribe in one click.