How to decommission a system in a controlled way
Decommissioning an old system requires three things once the replacement is live: an inventory of dependencies and integrations before you shut it down, a decision on what to archive versus delete under accounting and data protection rules, and cancellation of contracts and licenses that would otherwise keep costing money. Skip one and you get aftershocks.
When a new system finally goes live, everyone celebrates, and the old one is often left to its fate. But a half-decommissioned system isn’t gone – it’s a lingering risk with old personal data, unpaid licenses, and dependencies nobody remembers anymore. The final phase deserves the same care as the launch. Here are the three parts you can’t skip.
Inventory dependencies before you shut it down
The most dangerous assumption in a decommissioning is that an old system stands alone. In practice, it’s often wired into other things in ways nobody has full visibility into anymore.
An aging system might feed data to other services, receive files from an integration, or be the silent source behind a report some department relies on every month. Pull the plug without knowing that, and you can knock out something entirely different – and the fault only surfaces weeks later, when the connection is hard to trace.
So run an inventory before shutdown: which integrations go in and out, which systems pull data from here, and who actually uses it? A safe approach is to first run the old system in read-only mode for a while – it stays accessible for reading but stops accepting new data. If anyone’s missing it, it shows up then, while the damage is easy to repair.
Archive or delete: a legal fork in the road
Once the system can actually be shut down, the question of the data remains. And it’s not a question of keeping everything or throwing everything away, but two separate and deliberate decisions.
Archiving is preserving data in a readable format for the future – because the law requires it, or because it might be needed. Deletion is deliberately erasing data that may no longer, or no longer needs to, be kept. Just letting the system go dark is neither, and can breach rules in both directions.
The two requirements often pull in different directions, too:
- Accounting rules require that financial records be kept for a statutory period in Sweden. You can’t delete that early, even if the system is going away.
- Data protection (GDPR) conversely requires that personal data not be kept longer than necessary. Archiving the entire old database “just in case,” out of convenience, can therefore breach data minimization.
An orderly decommissioning therefore goes through which data is which: what must be preserved, what must be deleted, and in what format the archive should sit so it stays readable even once the system itself is gone.
| Data type | What drives the decision |
|---|---|
| Financial records | Must be archived for the statutory retention period |
| Personal data | Should be deleted once the purpose has ended – must not be kept unnecessarily |
| Operational data with no requirement | Preserve if it has value, otherwise delete |
Cancel the contracts that keep ticking otherwise
The third part is the most straightforwardly financial, and yet the one most often missed. Ceasing to use a system doesn’t stop its costs.
Licenses, cloud accounts, support agreements, domains, and integration services keep running until someone actively ends them – and many have notice periods that mean the last invoice arrives long after the system has gone dark. A decommissioned system that nobody has formally closed out can quietly cost money for years, a line item nobody thinks to question because “we don’t even use that system anymore.”
So go through every linked contract as part of the shutdown: what are we paying for, what commitment and notice periods apply, and when can each be ended? This is often where the entire decommissioning pays for itself.
A scenario: the system that never died
A company switches business systems and launches the new one with fanfare. The old one is left running “for now.” Nobody inventoried dependencies, so a nightly file feed to the payroll system quietly stops arriving – discovered only at the next payroll run. Nobody made a decision on the data, so the old customer database sits there with personal data long after it should have been deleted. And nobody cancelled the contracts, so the license keeps getting charged every quarter for two years before a new finance manager asks what the line item is.
None of this was inevitable. It was just a final phase that never happened.
How to use this
Treat decommissioning as its own small project with three questions: Do we know what depends on the system before we shut it down? Have we decided what to archive and what to delete? And have we cancelled all contracts and licenses? If you can answer yes to all three, the decommissioning is controlled instead of a lingering risk.
At Weapp, we’re glad to build an orderly decommissioning into the work when we help replace a system, so the old one closes as carefully as the new one opens. See our services or get in touch and we’ll walk through the plan.
Frequently asked questions
When is it safe to shut down an old system?
Only once the replacement is in stable operation and you know nothing else depends on the old one. Many systems quietly feed data to others through integrations, and a hasty shutdown can knock out something entirely different. It's safer to first run the old system in read-only mode for a while, in parallel, until you're sure nothing is missing it.
What's the difference between archiving and deleting data?
Archiving is keeping data in a readable format for the future, for example because the law requires it. Deletion is deliberately erasing data that may no longer, or no longer needs to, be kept. Both are active decisions. Just pulling the plug is neither, and can breach both accounting and data protection rules.
How long must data be kept?
It depends on the data type. Accounting records have a statutory retention period in Sweden, while personal data, conversely, must not be kept longer than necessary under GDPR. The two requirements can pull in different directions, which is why a decommissioning needs to go through which data is which – not treat it all as one pile to either keep or throw away.
What happens to contracts and licenses if we just stop using the system?
They usually keep costing money. Licenses, cloud accounts, support agreements, and integration services keep running until someone actively cancels them, and many have notice periods. A decommissioned system that nobody has formally closed out can keep ticking away in the budget for years. Go through every linked contract as part of the shutdown.
Why does decommissioning count as its own phase?
Because it's almost always forgotten. The project gets celebrated when the new system goes live, and the old one is left to its fate. But a half-decommissioned system is a lingering risk: old personal data, unpaid licenses, and dependencies nobody keeps track of anymore. An orderly decommissioning closes those risks instead of letting them sit and fester.