The business case for unattended hours is written in stoppages

The usual business case for extending unattended hours is simple arithmetic. The machines can run, the night is currently idle or thinly staffed, and capacity multiplied by hours gives a number. The number is large.

It is also the number least likely to materialise. What decides the return on unattended operation is not how long the machines can run. It is how often the operation stops with nobody there to restart it, and for how long.

Why the capacity sum misleads

A halt at night lasts until morning. On a staffed shift a minor deviation costs minutes, because somebody who has seen it before deals with it. Unattended, the same deviation stops the line until the first person arrives. One unexplained stop at midnight can erase most of the night.

The stops are not random. They come from situations nobody wrote a response for: a material that behaves oddly, a machine that drifts after a changeover, a reading that is normal in one state and a problem in another. The staffed shift resolves these from experience. The system has no experience to draw on.

Halting is the right behaviour. An unattended system that guesses in an unknown situation is worse than one that stops, because a guess can produce a shift of scrap. So the fix is not a braver system. It is making the unknown situations rarer.

Restarts cost time too. A line that stopped overnight may need checks and a warm-up before it produces good parts again. That time belongs in the case as well, and it rarely appears in the capacity sum.

What the business case should contain

Before anybody promises a return, five things are worth establishing, and most of them can be worked out from records the plant already keeps.

Infographic

Building the case for unattended hours

1. 01Count today's stopsEvery halt and intervention on the shifts meant to go unstaffed.
2. 02Group the causesThe exception types the operation actually sees.
3. 03Mark known responsesThe types somebody could write a response for today.
4. 04Count real hoursOnly the hours that unhandled exceptions will not consume.
5. 05Set the scopeWhich lines and products run unattended, and what stays staffed.

The stoppage record. Every halt and intervention on the shifts in question, including the ones that took half a minute because somebody happened to be standing there. Those are exactly the ones that turn into hours when nobody is.

The exception types. The stops grouped by cause. The grouping usually shows whether the problem is a few recurring situations or a long tail, and that distinction decides what is achievable.

Which responses are known. For each type, whether somebody can state today what should happen: continue, hold, stop or escalate. Where the response lives only in two operators’ heads, writing it down is the actual work.

The hours that are genuinely recoverable. Unattended capacity minus the time the still-unhandled exceptions will consume. It is a smaller number than the capacity sum, and the only one worth putting in front of a board.

What stays staffed. Some lines, products or changeovers should not run unattended at all. Naming them is part of the case, not a weakness in it.

What it costs to find out

Mostly time from the people who work nights, and a few weeks. A record of stops usually exists somewhere — maintenance logs, MES events, shift handover notes — and the people who handled them can explain the causes better than any report. It does not need new systems or sensors. It needs time from the people who remember what happened.

An honest readiness assessment usually narrows the ambition. That is its value. It replaces a capacity estimate with an answer to three questions: which parts of the operation can run unattended now, which need instrumentation or captured knowledge first, and which should stay staffed.

The narrower answer is also the one that survives. A plan built on the capacity sum tends to meet its first difficult night and lose the board’s confidence with it; a plan built on the stoppage record knows in advance which nights will be difficult.

What a good outcome looks like

A case the board can hold the team to. It states the unattended hours on a named set of lines, the exception types that are handled and the ones that are not, and how many nights are still expected to end in a halt.

From the first unattended night, a record of how the real nights compare. Where they diverge, the stoppage record says why, and the next round of work is chosen from that record rather than from the original ambition.

How we approach it

Dark factory context development starts with exactly that assessment, typically three to four weeks. Where it makes sense, it continues with a context model for one line, built with the shift teams and run under supervision before any human confirmation is removed.