Retiring a mainframe nobody could map
They knew what it did. They did not know who was still using it.
The program
| Sector | Government |
|---|---|
| System age | 31 years |
| Scope | Consumer discovery, archive, retirement |
| Duration | 9 months, of which 11 weeks was discovery |
| Status | Complete |
| Records | Retained under a schedule approved before extraction |
The agency could describe the mainframe's function precisely. What nobody could produce was a list of who still read from it, and without that list there is no such thing as a safe switch-off date.
The distribution list we were handed had two hundred names on it. Most were roles that no longer existed, some were people who had retired, and several were shared mailboxes nobody had opened in years. Asking that list who depended on the system would have produced forty confident answers, most of them wrong in the same direction — people say yes to that question because saying no carries all the risk.
So we measured instead. Eleven weeks of watching what actually connected, at what time, carrying what. Three real consumers. One was a reporting job nobody had run deliberately since 2019 but which had never been switched off. One was a downstream agency that had built against it without telling anybody. The third was the only one on the original list.
The eleven weeks were the whole engagement. Everything after — the archive, the retention schedule, the actual switch-off — took six months and produced no surprises, because the surprises had all already happened.
Eleven weeks to find three consumers. If we had asked instead of measured, we would have found forty and switched nothing off.
>
— Priya Vasterling, Principal, Estate Assessment
A clean rectangle on the floor
The console came out on a Tuesday. What it left was an unfaded rectangle, four bolt holes, and a hall that is measurably brighter empty than it was in use.
