A migration is complete and the old system is still running a year later. Why does decommissioning fail, and what makes it happen?

Why it fails

• No benefit is visible. The migration's value was captured when the new system took traffic; switching off the old one delivers nothing anyone can point to, so it loses every prioritisation comparison.
• Uncertainty about consumers. Nobody can prove nothing reads it, and proving a negative is uncomfortable. The safe-feeling choice is to leave it running.
• The data question is unresolved. Retention obligations, historical queries and a vague sense that someone might need it produce indefinite deferral.
• No owner. The migration team disbanded, and decommissioning belongs to nobody.

What it actually costs to leave running

Infrastructure, licences, a security patching obligation, an on-call surface, a compliance scope entry, and cognitive load on everyone who encounters it. The security cost is the most serious and least visible: an unmaintained system with production data is an attractive target and it is the one nobody is watching.

It also blocks the next migration, because the estate's picture remains unclear.

What makes it happen

• Plan it as part of the migration, with a date and an owner assigned before the migration starts. Retro- fitting a decommissioning project after the fact rarely gets funded.
• Prove disuse with telemetry rather than by asking. Log every access for a defined period; the ones that appear are the consumers nobody mentioned.
• Turn it off before deleting it. A period where the system is unreachable but recoverable converts the scary irreversible step into a reversible one, and it surfaces the last consumers immediately.
• Resolve the data question separately. Archive to cheap storage with a documented retrieval path, so the retention obligation is satisfied without the system running.
• Count the removal in the migration's success criteria, so the migration is not declared complete while the old system is still live.

The framing that unblocks it

Decommissioning is where the migration's return is actually realised. Until the old system is off, the organisation is paying for two and has reduced neither cost nor risk — it has only added complexity. Saying that plainly is usually what gets the work prioritised.