Output often fades in the last 30-60 minutes of a shift because crews start cleaning, finish paperwork, avoid starting a new job, or stop once the target is met. To see it, plot good units in 15-minute blocks for the last 90 minutes over at least 20 shifts. Fix it by scheduling end-of-shift tasks, setting a rule for late job starts, and handing over at the running line.
Nobody decides to lose the end of a shift. It happens through small and reasonable choices: wipe down early, write up the log, do not begin a changeover that the next crew will finish.
Each choice makes sense for the person making it. Added up, the line gives away a block of output every shift, and often hands the next crew a slow start too.
Use 15-minute blocks
An hourly chart can blur this pattern, because it often lives in the last 30 minutes. So use shorter blocks, and go back at least 90 minutes so that you have a steady reference.
Example: a line with an ideal rate of 90 units a minute, or 1,350 per 15-minute block. Here are the median good units for each block over 20 shifts.
Against the steady 1,080, the last four blocks fall short by 30, 120, 300 and 630 units. That is 1,080 units in total, or 1,080 / 90 = 12 minutes at the ideal rate.
So on this hypothetical line, the fade costs exactly one steady block of output each shift. On a 480-minute shift those 12 minutes are 2.5% of the schedule.
First, check the data
End-of-shift counts can give a false picture. Before you talk to anyone, rule these out.
- Counts entered late. With manual counts, a final batch typed in at the bell looks like a spike in the last block. Compare with a machine counter or a pallet tally.
- Rounding to the target. If shift totals land too neatly on the plan, the log may be shaped to fit it. See how to check downtime data.
- The clock. Check that the system's shift boundary matches the real bell. A 10-minute offset moves output from one shift to another.
- The schedule. A planned clean or a planned stop at the end of the shift is not a fade.
Read the shape
Different shapes point to different causes. Find yours in the table, then go and confirm it at the line.
| Shape | Likely meaning | How to confirm |
|---|---|---|
| Slow fade over 45-60 min | People pulled away for tasks, material not topped up, lower urgency | Watch who leaves the line, and when |
| Cliff in the last 10-15 min | Line stopped early for cleaning, paperwork or leaving | Time of the last unit against the bell |
| Spike near the end | Rush to hit a target, or counts entered late | Scrap rate in that block, machine counter |
| Idle gap after a job ends | Nobody starts a new job or changeover this late | Job end times against stops with no reason |
Our example is a slow fade that ends in a cliff: people drift to end-of-shift tasks, then stop.
Measure it the same way for all crews
Track one number per shift: minutes from the last good unit to the bell. It takes seconds to read, and you can put it straight on a board.
A second useful number is the output of the last 30 minutes as a share of a steady 30 minutes. In the example, that is (780 + 450) / (1,080 + 1,080) = 57%.
If all crews look the same, the cause lies in the routine. If one crew is far better, ask them how they do it, then use why OEE differs between shifts to rule out product mix.
Why it matters beyond the 12 minutes
The end of one shift is the start of the next. A line stopped early, left dirty and not stocked is the reason the next crew starts late.
Plot the last hour and the next crew's first hour as a single picture. The real loss at each shift change is the two together, and one fix can help both.
How to fix it
Cleaning, logs, counts, tool returns. Time each one.
Give each task a time, an owner and a code, and stagger them so someone stays on the line.
Decide when a job or changeover still starts before the bell.
The line keeps running through the bell where it safely can.
Compare 15-minute blocks before and after.
Give cleaning its own slot
If a line must be cleaned, then plan and log it under its own code, instead of letting it creep in early. Where it belongs in OEE is covered in sanitation downtime.
Do not skip safety or hygiene tasks to protect output. The goal is to stop unplanned drift, not to run until the last second at any cost.
Write the rule for late starts
Example: start a changeover if at least half of it fits before the bell, and pass it on with notes if not. Starting matters, since an idle line in the last 40 minutes helps nobody.
This only works if crews get credit for shared work. If the shift that starts a changeover looks worse than the shift that finishes it, nobody will start one, so read how to count changeovers in OEE.
Stop shift targets from ending the shift
Fixed shift targets can create the pattern. Once the number is met, there is little reason to push, and a rule that "over target does not count" makes it worse.
An hour-by-hour board spreads the target through the shift. It also shows a falling count while there is time to act, which is the point of short interval control.
Shorten the paperwork
If logs get written up in a rush at the end, they pull someone off the line and are less reliable too. Enter stops as they happen, with simple tracking, and the end-of-shift write-up shrinks.
The handover itself should take minutes. The shift handover checklist keeps it to what the next crew really needs.
What to avoid
- Blaming the crew. People respond to how the shift is set up. Ask what they are required to finish before they leave.
- Banning all early stops. You will get the same behavior with a tidier log.
- Judging by one week. Use at least 20 shifts, and medians, so that one breakdown does not bend the picture.
- Measuring only output. A rush can raise the last block and also raise scrap. Track both.
Sizing the win
Example: the line above wins back 8 of its 12 minutes on each of 3 shifts. That is 24 minutes a day, or 2,160 units at the ideal rate, with no new equipment.
Use the downtime cost calculator to turn your own minutes into money. Then decide if a short overlap between crews would pay for itself.
The pattern needs counts in short blocks and stops with a reason. Fabrico takes stops, speed and counts live from your machines, operators tag the cause at the line, and you see OEE per line and shift. See how Fabrico tracks stops per shift.
What people ask next
Why does production drop before shift change?
Usually a mix of early cleaning, paperwork, not starting new jobs and less urgency once a target is met. The shape in 15-minute counts tells you which is biggest. Use the table above to read it.
Should end of shift cleaning count against OEE?
Planned cleaning normally stays in scheduled time unless your plant has a written rule that excludes it. Pick a rule, give it its own code and apply it to all crews. See sanitation downtime.
How long should a shift handover take?
We found no standard figure that fits all plants, so aim for just long enough to pass on status and open problems. Keep the line running during it where that is safe. The shift handover checklist gives a short format.
Is it better to run through the shift change?
Where it is safe and the line can be staffed, yes, since a line that never stops avoids restart losses too. Not every process allows it, so compare the cost of a staffing overlap with the output won. First hour of shift losses helps you size both ends.
Why do manual counts jump at the end of the shift?
Often the jump is only the moment the count gets typed in. Check it against a machine counter or a finished goods tally before you act on it. OEE data collection methods compares the options.
Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.
Related guides: OEE on 12-Hour Shifts: Comparing Crews Fairly