Walk onto most automated production lines that have been running for a few years and you will find people doing jobs nobody ever specified. Someone standing at a transfer point because product jams there. Someone clearing the same conveyor blockage they cleared yesterday, and the day before that. A rework station halfway down the line, and another at the end, both of which exist because an upstream operation does not do what it was supposed to do.
None of that was in the original scope. It was added quietly, one person at a time, after the integrator left and the line failed to reach the numbers it was bought to deliver. And once it has been there for six months it stops being treated as a fault. It becomes the way the line runs.
A line that needs people to keep it running is not a line that needs more people. It is a line with faults that were never closed out.
Why this gets missed
Downtime gets measured. Scrap gets measured. Labour added to compensate for a technical fault usually does not, because by the time it appears on the shift rota nobody connects it back to the installation.
It also hides well in OEE. If an operator clears a jam in twenty seconds and the line never stops, availability barely moves. The fault is real, it happens forty times a shift, and it is completely invisible in the metric most plants use to judge line health. The only place it shows up is in the headcount, and headcount tends to be reviewed as a staffing question rather than an engineering one.
Key point: Manual intervention that prevents a stoppage does not register as downtime. A line can show strong availability figures and still be dependent on people to achieve them.
The faults that create these roles
In our experience the same handful of root causes account for most of it. All of them are fixable, and none of them require replacing the equipment.
1. Transfer points that were never properly proven
Product jamming at a transfer is rarely a mystery. It is usually a guide rail set for a product variant that is no longer the main runner, a gap or height difference between conveyor sections, or a transfer designed around a nominal pack dimension rather than the real tolerance range. When a plant runs several formats through one line, the transfer often works for the format used during commissioning and fails on the others.
2. Conveyor speeds that were never balanced
Accumulation only works if each section runs faster than the one feeding it, with enough buffer to absorb a stop downstream. Where speeds were set during startup and never revisited, product backs up in the wrong place. The usual sign is a person permanently stationed at one point, spacing product by hand so the next machine can index cleanly.
3. Sensor positions that detect the wrong thing
A photoelectric sensor mounted where it sees the pack label rather than the pack body will read inconsistently on a change of artwork. A reflective sensor aimed across a shiny stainless guard will pick up false triggers. These faults produce intermittent faults that are hard to reproduce and easy to work around, so a person gets put there instead.
4. Sequences written to pass a FAT
A factory acceptance test runs on good product, in controlled conditions, usually on a single format. A sequence written to satisfy that test may have no fault recovery, no handling for a part arriving slightly out of position, and no graceful behaviour when a downstream machine stops mid-cycle. It passes, it gets signed off, and then it meets real production.
5. Diagnostics that do not tell anyone anything
Where the HMI reports a generic fault code with no indication of which station or which sensor caused it, people learn to diagnose by walking the line and looking. That habit becomes a role. Good diagnostics are usually the cheapest single improvement available on an existing line, and they are almost always a software change rather than a hardware one.
Putting a number on it
The argument for fixing any of this only lands when it has a number attached. The method is simple, and you can do it in an afternoon.
Walk the line during normal production and list every person whose task exists because something does not work as intended. Be strict about the test: if the fault were fixed tomorrow, would this role still be needed? Then multiply by shift pattern and by what that cover actually costs you per hour, including agency margin.
| Role | Underlying fault | People | Illustrative annual cost |
|---|---|---|---|
| Clearing jams at infeed transfer | Guide rails set for one format only | 2 per shift, 2 shifts | £138,000 |
| Hand-spacing product before case packer | Conveyor speeds never balanced | 2 per shift, 2 shifts | £138,000 |
| Mid-line rework station | Upstream labeller misapplies on format change | 2 per shift, 2 shifts | £138,000 |
| Total, one line | £414,000 |
The figures above are based on an agency rate of £18 per hour, and on two people covering each issue on every shift, which is what we typically find where a fault has been absorbed into the rota rather than fixed. An eight hour shift is £144 a head. Two people across two shifts is £576 a day, and over a five day week and forty eight working weeks that comes to roughly £138,000 a year for a single unresolved fault. Three of them, on a single line, is over £400,000.
Use your own rate and your own running pattern. The point is not the precise total but the order of magnitude, because agency cover of this kind sits in an operations budget rather than a capital one, renews quietly every year, and almost never appears in any business case for engineering work.
Now multiply it by your number of lines
Everything above is a single line. Where a plant runs two, three or four similar lines, the same faults usually exist on all of them. They were installed by the same integrator, to the same specification, and commissioned the same way, so they fail the same way.
The repetition cuts both ways: if the fault is common across the lines, so is the fix. Work proven on one line normally transfers to the rest at a fraction of the original engineering effort. A multi-line estate is the strongest case for doing this properly, not the most daunting.
Set that against what it costs to correct the underlying faults. Rebalancing conveyor speeds, repositioning sensors, rewriting a transfer sequence and adding meaningful fault reporting is measured in engineering days, not capital expenditure. In most cases the work pays for itself inside a quarter.
Not sure what your line is really costing you?
We walk the line with you, identify the faults behind the manual intervention, and tell you what it would take to close them out.
Book a Free Line Review →What fixing it actually involves
The work is unglamorous and mostly software. Typical scope on a line like this includes:
- Re-proving transfers across every product format currently run, not just the primary one
- Recalculating and resetting conveyor speeds and accumulation logic through the affected zones
- Repositioning or replacing sensors where the detection principle is wrong for the application
- Rewriting sequence steps to handle misposition, partial cycles and downstream stops without operator intervention
- Adding station-level and device-level fault reporting to the HMI so faults can be diagnosed from the panel
Where a line is busy and cannot be released for trials, changes can be proven offline first. Virtual commissioning lets sequence changes be tested against a simulated model of the line before they go anywhere near live equipment, which removes most of the risk of making things worse while trying to make them better.
The question worth asking
Every plant has a number for what its lines cost to buy. Very few have a number for what they cost to keep running by hand.
If your line needs people to babysit it, count the roles that only exist because of a fault, and put an annual figure against them. It is normally bigger than anyone expects, and once it is written down the conversation about fixing the underlying faults becomes a great deal easier to have.