A night shift that scraps more than days is rarely a story about people working less carefully in the dark. In most plants the gap traces back to structural causes: scrap gets booked to the shift that finds it rather than the shift that made it, the night schedule inherits the jobs most likely to go wrong, the people who correct a drifting process have gone home, and the building itself changes overnight. Effort and attitude come a distant last, and plants that start there spend a year on toolbox talks without moving the number.
Each cause leaves a different signature in the data and needs a different fix. Here is how to tell them apart.
Is the gap real, or an artefact of how scrap gets booked?
Test this first, because it costs nothing and it is wrong more often than anyone expects.
Scrap is usually recorded at the moment somebody decides a part is unsaleable — which is often not the moment, the machine, or the shift that produced the defect. Three booking habits create a phantom night-shift problem:
- Disposition happens on one shift. If the quality technician or the material review board only works days, a night-shift operator sets a doubtful part aside and days books it — while the day shift's own doubtful parts get dispositioned the same day and quietly reworked. Nights never get the same benefit of the doubt, because nobody who can grant it is there.
- The last shift on the job absorbs the batch. Where scrap is counted against a work order at close-out rather than at the operation, the shift that finishes a run inherits everything the run produced. If nights typically finish what days started, nights carry the paperwork.
- Sorting is scheduled at night. Quarantine sorts and rework get pushed to the quieter shift, and every reject found during a sort lands on the sorting shift's ledger.
The check: take one month of scrap tickets and compare who wrote the ticket against which shift ran the operation that made the part. If those columns disagree often, fix the booking rules before interpreting anything else — a scrap system that cannot attribute a defect to the operation that caused it will mislead every investigation you run through it. The wider principle, measuring the process step rather than the person nearest the bin, is in our guide to cutting scrap and rework.
Why the night schedule inherits the jobs most likely to go wrong
Assume the gap survives the booking check. The next question is not who was working, but what they were working on.
Scrap is not spread evenly through a run. It concentrates at the start: the first pieces after a changeover, after a new material lot goes on, after a tool change, after a machine has stood cold for hours. Steady-state running produces comparatively little. So any shift that gets a disproportionate share of beginnings produces a disproportionate share of scrap, with identical people and identical discipline.
Beginnings drift to nights for entirely rational reasons. Planners protect the day shift's long, high-value runs and schedule changeovers into the gap. Material deliveries arrive during the day, so a new lot goes on late. Maintenance takes machines down in the evening, so restarts land at night. Trials get placed where they disturb the least output.
The check: count changeovers, tool changes, material-lot changes and cold starts per shift, then express scrap as parts scrapped per changeover and parts scrapped per running hour rather than per shift. If the per-changeover figures match across shifts and only the totals differ, you do not have a night-shift problem — you have a scheduling pattern. The fix is then either to redistribute the beginnings across shifts or to cut what each beginning costs — a setup discipline problem, not a shift problem.
What happens to quality when the support functions go home
This is the cause that survives every other correction, and the strongest argument for treating a shift gap as a design issue rather than a behaviour issue.
A process rarely fails cleanly. It drifts: a fixture loosens, a coolant concentration falls, a sensor creeps out of calibration, a die begins to pick up. On days that drift is caught early, because the plant is full of people who catch things — a process engineer walking the line, a toolmaker who can regrind at short notice, a quality technician who can measure a doubtful feature in ten minutes.
At night the same drift starts the same way. The difference is what happens next. There is nobody to confirm the reading, nobody authorised to stop the job, and often no realistic alternative to running on and flagging it in the morning. The defect rate is not higher because the drift is worse; it is higher because the drift runs longer before anyone acts on it. A fault caught in twenty minutes on days is a handful of parts. The same fault caught at handover is several hours of production.
The signature in the data is distinctive: night scrap is lumpy, not uniformly elevated. It arrives as long runs of the same defect code on the same machine rather than as a slightly worse background rate. If your night-shift excess disappears when you remove two or three bad nights from the month, this is your cause, and no amount of training will address it.
What does address it:
- Give the shift authority to stop. A supervisor allowed to halt a job and take the schedule hit is worth more than an escalation number nobody rings.
- Make in-process checks decisive. A check whose result the shift cannot act on is paperwork. Control limits with a stated action — adjust, stop, quarantine — turn a measurement into a decision, which is the practical purpose of statistical process control described in our quality control guide.
- Name a person who can be called, and mean it. An on-call rota with a real expectation of answering costs far less than the parts it saves.
- Shorten the approval loop. If first-off approval needs a daytime signature, the night shift is running unapproved work by design.
Does the building itself change overnight?
Sometimes the gap is environmental, and it is easy to miss because nobody is standing in the plant at three in the morning to notice.
Ambient temperature and humidity swing overnight, and several processes care: adhesives, coatings, paint booths, curing, moulding, and anything with a tolerance tight enough to see thermal expansion. Compressed air, chilled water and site electrical load all behave differently against a cooler, quieter night. Lighting is artificial, which matters for any visual inspection involving colour, gloss or fine surface defects.
The check: plot the defect code against clock time rather than against shift. Environmental causes produce a curve — a rise through the coldest hours, a recovery near dawn — while human causes produce a step at the shift boundary. That single plot separates two explanations that look identical in a shift summary, and it is the cheapest experiment here.
Inspection deserves its own mention. If night inspectors work under different light or to a different sampling plan, you may be measuring inspection rather than production — and parity there is a precondition for comparing shifts at all.
When it really is staffing and experience
Occasionally, after all of the above, a genuine capability gap remains — and it is worth naming honestly rather than pretending every difference is structural.
The pattern is usually not carelessness. It is experience distribution: newer operators, agency labour and less cross-trained teams concentrate on the least popular shifts, because plants staff days first and fill nights with whoever is left. Add longer gaps between refresher training and fewer chances to watch an experienced hand solve a problem, and the difference shows up in the parts.
Two fixes carry weight. The first is deliberate experience mixing — spreading experienced operators across shifts instead of concentrating them where the visible output is. The second is a structured handover: not a chat at the door, but a short written record of what is running, what has been adjusted, what is suspect, and what the last measurement said. Most repeat defects that survive a shift change survive it because that information did not.
How to measure the gap so the answer is usable
Shift totals cannot diagnose anything. Build the table one level down:
- Rows: defect code.
- Columns: shift.
- Cells: scrap per running hour, and scrap per changeover.
- Filter: one part number and one machine at a time.
Almost every real shift gap collapses into a small number of cells — one defect code, on one machine, on one shift. That is a specific problem with a specific mechanism behind it, and the point at which structured root cause analysis starts to pay. A plant-wide "night shift quality initiative" is what gets launched when nobody has built the table.
FAQ
Should I compare shifts on scrap rate or scrap quantity?
Rate, normalised to running hours rather than shift length. Shifts differ in downtime, changeovers and product mix, so a quantity comparison silently blames whichever shift ran the most parts — or the most beginnings.
How much shift-to-shift variation is normal?
There is no universal figure worth quoting, and any specific number you read is a claim about somebody else's plant. Judge variation against your own history instead: a gap that is stable month after month is structural, and one that comes and goes is event-driven.
Is fatigue a real cause of defects on night shifts?
Alertness genuinely varies across the circadian cycle, and it is a legitimate factor in any task depending on sustained visual attention. It is also the least actionable explanation on the list, which is why it belongs after the structural causes rather than instead of them.
Our night shift also has less downtime. Does that change the analysis?
Yes, and it is a common finding: fewer interruptions from meetings, deliveries and visitors means more running hours, so more parts per shift and a higher scrap quantity at an unchanged rate. Another reason to normalise first.
What if the day shift is the one scrapping more?
The same framework applies with the causes reordered. Days typically carry more changeovers in plants that run production through the night, more trials, more disposition decisions and more interruptions. The diagnostic — book to the operation, normalise per running hour, plot against clock time — does not care which shift you are defending.
A shift gap is a measurement question before it is a people question. Fix how scrap is attributed, normalise for changeovers and running hours, plot the defect against the clock, and the explanation usually names itself. For more vendor-neutral guidance on running a cleaner, more predictable shop floor, visit Manufax.
