A 40-minute breakdown gets a reason code, a maintenance call, a line in the shift report and five minutes in the morning meeting. Ninety-six 25-second stops across the same shift add up to the same 40 minutes and get nothing: no code, no call, no line, no meeting.
On many packing lines the second kind loses more time than the first.
That is the whole problem. Micro-stops are the largest loss category on many lines and the only one that has no owner, because nobody has ever seen it as one thing. This article covers what they are, why every dashboard misses them, how to measure them so they become visible, and how to fix them in batches, which is the only way they get fixed.
A micro-stop is any stop short enough that the operator clears it without calling maintenance and without logging it. In practice that means anything under the plant’s logging threshold, usually two to five minutes: a jam at the infeed, a misfed label, a bottle down on the conveyor, a sensor that needs a wipe, a case packer waiting on a flap.
The threshold is a choice, and it matters more than it looks. Wherever the line is drawn, everything below it disappears from the record. A plant with a five-minute threshold has decided, without ever deciding, that a four-minute stop is not a stop.
Total Productive Maintenance (TPM) made a distinction that is useful here. Sporadic losses are large, sudden and visible: a breakdown, a major jam, a quality hold. They get fixed because they cannot be ignored. Chronic losses are small, frequent and accepted as how the line runs. Nobody fixes them because nobody experiences them as a problem; they experience them as Tuesday. Micro-stops are the archetypal chronic loss, and chronic losses are where the capacity is.
Four reasons, and they compound.
They are below the logging threshold. The downtime system was designed to capture stops worth a reason code. A 25-second stop is not worth the time it takes to code it, so the system never sees it.
They are absorbed into performance. When the run signal does not register a stop, the lost cycles show up as a lower count against the ideal speed. In the OEE calculation that is a performance loss, indistinguishable from the line running slow. The OEE audit article covers this as driver four: micro-stops are the main reason performance factors are unexplainable.
They are cleared reflexively. An experienced operator clears a jam in eight seconds without breaking conversation. It does not feel like downtime to anyone on the line, so it never becomes a topic.
They have no single cause and so no single owner. A breakdown belongs to maintenance. A changeover belongs to the crew. A stop that is sometimes the film, sometimes the guide and sometimes the operator belongs to nobody.
The arithmetic is what makes this serious. A 20-second jam every four minutes is 8% of the shift. A line that reports 95% performance against its target speed may be losing that 8% to stops it has never counted, and the shift report will say the line ran all day.
Micro-stops cluster at the stations where product changes state or orientation: infeed and accumulation tables, labelers, cappers and closers, case packers and erectors, and the handoff from bulk to pack. On almost every line one or two stations carry most of the count, and they are rarely the stations the plant would have guessed.
This is the first practical point. Micro-stops are a station problem before they are a line problem. A line-level OEE, however honest, cannot see the station; it sees a performance factor. Finding micro-stops means measuring at the station, which is a different sensor layout from measuring at the line.
The instinct is to measure micro-stops the way breakdowns are measured, in minutes lost. It does not work. The durations are small and noisy, and a chart of minutes lost to micro-stops looks like static.
The metric that works is stops per hour, by station. Frequency is large, stable and responds to a fix within a single shift. A labeler running at 14 stops an hour that drops to 5 after a guide adjustment is a result anyone can see by the end of the day. The duration can be added back later to value the loss; the count is what finds it.
How to capture the count, in order of preference:
One rule above all: operators do not log micro-stops, and no training program will change that. The system has to capture them, or they do not exist.
Once every stop is a record, the cause is usually visible in one of five cuts.
By station. The Pareto. Two stations typically carry 60 to 70% of the count. Start there and ignore the rest for now.
By SKU. A labeler that stops three times more often on one bottle shape than on the others is a format problem, not a labeler problem. The fix is in the change parts or the guide settings for that SKU, and it will not show up if the data is averaged across the mix.
By shift or crew. A station whose rate differs by crew has an intervention-habit cause: how the operator feeds, where they stand, when they clear. The fast crew’s habit becomes the standard.
By time since changeover. A rate that climbs steadily for two hours after every changeover and then stabilizes is settings drift or a warm-up effect, not a crew effect. A rate that spikes in the first twenty minutes and then drops is a changeover-quality problem: the station was not set correctly and the operator tuned it live.
By material lot or supplier. A stop rate that doubles the day a new film supplier’s roll goes on, or when a different bottle lot arrives, is a procurement conversation. Without the data it is a shift-floor argument about who is doing something wrong.
The cuts are cheap once the records exist. The expensive part was capturing the records, and most plants have never done it.
Micro-stops are not fixed one at a time. There are too many, each is too small, and fixing one does not move the number. They are fixed in batches.
Take the ten highest-frequency station-and-SKU combinations. Give each one an owner and a hypothesis from the cuts above. Make the change. Re-measure the rate on that station within a week. Keep what worked, revise what did not, and take the next ten.
The usual causes, roughly in order of frequency: material variation (film, labels, bottles, cases), worn or wrong change parts, sensor alignment and guide settings, operator intervention habits, and buffer sizing between stations. Most of the fixes cost nothing or almost nothing. This is the targeted-fixes bucket from the hidden capacity method, and it is why micro-stops are usually the fastest capacity a plant recovers: the data points at the cause, the fix is a guide adjustment, and the rate drops the same shift.
Worked example. On the packing line used across this series, micro-stops were 9 OEE points, nearly 11 hours a week, and invisible in the reported 72%. Station-level capture showed the labeler carrying 41% of all stops, the case packer 29%, the filler 14% and the rest scattered. Labeler stops concentrated on two of six SKUs and climbed for ninety minutes after every changeover into those formats. Two guide-setting standards and one replaced change part brought the labeler rate down 60% by the second week. The case packer turned out to be a case-blank supplier issue. Three of the nine points were recovered inside six weeks, with no capital, which is the figure the targeted-fixes bucket in the capacity article assumes. The numbers are illustrative, but the shape is the one we see on most lines: one station, two SKUs, a settings problem, a material problem.
Chronic stops are a leading indicator. A station whose stop rate rises week on week with no change in SKU or material is wearing or drifting: a bearing, a belt, a sensor losing alignment, a pneumatic that is slowing. The breakdown that eventually gets a reason code and a maintenance call was visible in the micro-stop trend for weeks before it happened.
That makes the stop-rate trend a condition-based maintenance trigger. When the OEE system and the maintenance system refer to the same machine by the same code, a rising rate on one station can raise a preventive work order before the sporadic loss arrives. It is one of the few places where operations data directly shortens the maintenance backlog.
The caveat is practical. This only works when both systems name the machines the same way. In plants where the OEE stations and the CMMS assets were set up by different teams years apart, matching them is the first job, and it is a data-hygiene project rather than an analytics one. It is worth doing; the alternative is two systems that both know about the labeler and cannot tell each other.
Everything above can be done with a PLC signal, a spreadsheet and a determined engineer. Three things are hard to do that way, and they are what Fabrico’s manufacturing performance platform (MES, OEE, CMMS & AI) is built around: seeing the cause of each stop, valuing it, and changing the plan because of it.
Seeing the stop, not just the signal. A run signal tells you the station stopped. It does not tell you whether the film jammed, the guide blocked, or the operator reached in. Fabrico’s computer vision module watches the station, not the people at it. It suggests whether the event was a jam, a blockage, a micro-stop or a manual intervention, and separates changeover stages the same way. The operator confirms the suggested cause with a tap, and the footage is kept as evidence, so every stop record carries a cause without anyone typing one. On stations with no stop signal at all, the camera is the sensor.
Ranking the list by money, not by count. The ten-stop list above is ordered by frequency, which is the right place to start and the wrong place to finish. Ten stops an hour on a station running a high-margin SKU are worth more than twenty on a low-margin one. The financial impact module makes that calculation possible with very little effort from the plant: enter the selling price and the margin for each SKU once (contribution margin, for capacity decisions), and from then on every lost hour on the line is valued at the margin of whatever was running at the time. On that basis, Fabrico’s AI actionable insights rank the recommended fixes by financial impact: each chronic loss is converted to recoverable hours, then to recoverable units on the SKUs that actually run there, then to dollars. The result is a list that says “fix the labeler guide on SKU E first; it is worth $31,000 a year”, with an owner and a measured result attached. The insight proposes; the plant decides and executes.
Scheduling around the stops you have not fixed yet. A micro-stop pattern is also a planning input. The production scheduling module plans each line at its demonstrated rate per SKU rather than the standard, so a format that stops the labeler three times more often is scheduled at the output it really delivers. Until the fix lands, it routes that format to the line where it runs cleanly, groups it to limit the post-changeover climb in stop rate, and when a station’s rate spikes mid-shift, the replanning engine re-sequences the remaining orders across the other lines while the financial impact module shows what the spike cost and what the replan recovers. The fix removes the loss; the schedule stops it from hitting the high-margin orders in the meantime.
The effect is that the micro-stop program stops being an engineering hobby and becomes a ranked backlog with dollar values, fed into a schedule that already accounts for it, which a plant manager can defend in the capacity review.
What counts as a micro-stop? Any stop short enough that the operator clears it without maintenance and without logging it, in practice anything under the plant’s logging threshold of two to five minutes. The threshold is a choice, and everything below it disappears from the downtime record.
Why don’t micro-stops show up in OEE? Because they are below the logging threshold and the lost cycles are absorbed into the performance factor, where they look like the line running slow. A line can report 95% performance while losing 8% of the shift to stops nobody counted.
How do you measure micro-stops without operator logging? Automatically: run-signal transitions from the PLC or a sensor, count-rate gaps from a photo-eye, or computer vision at the station. Operators add reason codes to longer stops; they never estimate micro-stop durations. Measure by stops per hour per station rather than minutes lost.
What is a normal micro-stop rate? There is no universal figure; it depends on the station, the format and the material. A useful internal benchmark is the best station-and-SKU combination on the same line: if the labeler runs at 4 stops an hour on one format and 14 on another, the 4 is the target.
Are micro-stops a maintenance or an operations problem? Both, which is why they have no owner. Most are operations: settings, change parts, material, habits. A rising rate on one station with nothing else changing is maintenance: wear or drift that will become a breakdown. The data tells you which.
If your performance factor has a gap you cannot explain, it is micro-stops. The only way to confirm it is to capture every stop at the station for a few weeks and look at the count.
The fixed-scope pilot does exactly that: one representative line, six weeks, an industrial sensor and hub installed with your team, and computer vision at the station where the stops are, with every stop recorded from day one. The loss Pareto in week three is usually the first time the plant sees its micro-stops as one number, and the three highest-value interventions in the readout are usually on this list. The fee is fixed and credited in full against a first-year subscription if you roll out. The pilot runs on Fabrico’s manufacturing performance platform (MES, OEE, CMMS & AI), which connects machine data, OEE and loss analysis, production scheduling, SKU-level output value and maintenance in one system, so every stop it captures, however short, ends up ranked, owned and re-measured.
Request a demo or read how computer vision classifies stops at the station.
Login
Blog
