Context

At 4,200 m.a.s.l., a 48-hour scheduled shutdown becomes a non-stop negotiation with production. The question isn't whether we'll cut — it's where.

The problem

The mill PLC migration ran in parallel with a pending FAT and an expired lift permit. Adding optimistic hours, we didn't fit the window.

"If it adds hours to the window, it doesn't fit." — the instrumentation supervisor.

What changed

• We moved the FAT to the night before and kept only SAT on site.
• We compressed LOTO with a pre-signed digital permit.
• We coordinated the lift with the incoming shift, not the outgoing one.

The result

Original window:  48 h
Executed window:  36 h
Margin:           12 h ahead of target

What I learned

1. The shutdown schedule is not won at the end — it's won at the first turn.
2. A badly placed FAT in the Gantt is worth more than three review meetings.

Next step

We're measuring the next shutdown with the same pattern. If it works twice, we move it into the playbook.