By the time a supervisor notices a line has slowed, the damage is done. A station that ran slow for an hour has already cost that hour of output, and nothing gets it back. The only lever left is how soon someone finds out. This post covers what that delay costs, what a monitoring system has to do to shrink it, and how to use the time it gives you.
What an hour costs, and why the numbers disagree
Published figures vary widely. Aberdeen Research, as reported by Reliamag, puts the average at $260,000 an hour across all manufacturing sectors and says an hour costs about 50 percent more than it did in 2019. Fluke’s 2025 survey of more than 600 decision-makers and maintenance professionals in the US, UK, and Germany found that 61 percent of manufacturers had unplanned downtime in the past year, at an average of $1.7 million an hour. Siemens’ True Cost of Downtime 2024 puts automotive at $2.3 million an hour and estimates that the world’s 500 largest companies lose $1.4 trillion a year, 11 percent of their revenue.
The figures differ because the samples differ. Aberdeen averages every sector. Fluke surveyed food and beverage, oil and gas, life sciences and automotive. Siemens looked at the largest companies in the world. None of them is your plant. Treat them as a range, then work out your own number: lost margin per hour, plus the labor you still pay, plus overtime to catch up, plus any expedite or penalty costs.
The cost people miss is the delay.
The direct costs are easy to see: lost output, paid labor with nothing to build, overtime. The indirect costs pile up quietly. A missed delivery window can trigger a penalty. A rushed recovery can introduce quality escapes, because teams under pressure skip verification steps. And the longer a problem runs before anyone sees it, the more of those costs it collects.
Most problems do not arrive as a sudden stop. A station slows by a few seconds per cycle, then a few more, until the loss is too big to ignore or the station stops. Walkthroughs, hourly checks, and end-of-shift reports all share one flaw: they find the problem after the interval they cover. A dashboard that refreshes every 15 minutes leaves a 15-minute blind spot on every shift. Ask any vendor for the actual delay between an event and an alert, in minutes.
What continuous monitoring changes
By continuous, we mean three specific things. Our cameras capture 100 percent of cycles at every station they cover. You set the alerts for the conditions you care about, including line stoppages and andon pulls. And each alert links to the footage of the cycle behind it, so your team sees what happened instead of reconstructing it. The same data feeds a production management view of cycle time, throughput, and wait time, linked to video.
Every alert points to a station and a condition, never to a person. A stopped station is a process problem: a missing part, a fixture, a changeover that ran long. The footage helps your team find that cause and fix it, instead of guessing.
One plant manager using our system estimates that every minute of downtime prevented saves their plant about $1,000. That figure reflects their product and volume, so work out your own before you rely on it.
Here is an illustration, not a customer result. Say a final assembly line loses ground on second shift every time it runs a new variant. A shift report shows low output and little else. A station-level view shows one station waiting on a sub-assembly at the start of each cycle, and the footage shows the part arriving late from a staging area. The fix is a change to material flow, found within a day instead of a month of arguing about it. Manual stations like that one produce no machine signal, which is why a camera finds what a PLC cannot.
Make alerts worth acting on
A monitoring system only prevents downtime if people act on what it flags. Four habits make the difference.
Baseline the real cycle times: Use what each station does today, not the target on paper.
Set thresholds for meaningful drift: Alerts that fire on every small wobble train people to ignore them.
Give every alert an owner: Decide who responds, what they check first, and when the problem escalates.
Review the pattern weekly: Recurring issues deserve a permanent fix, not another quick response.
Trends catch what alerts miss.
Single alerts tell you about today. Trends tell you what to fix for good. A station that slows on the same variant every Tuesday, or a changeover that runs long every week, points to a cause an alert alone would never show. This is where your continuous improvement team earns the most from the data: they work from weeks of station history instead of one day’s complaint.
Know the limits, too. Cycle-time trends and recurring stoppages can show which stations need attention before they stop. Vision will not tell you a bearing is wearing out. Vibration and temperature sensors do that job, so use both where equipment health matters.
Downtime and strain often start at the same station.
A station with an awkward layout or a slow fixture can lose minutes and strain the person working it at the same time. The cycle data shows the lost time, and our safety and ergonomics findings flag the strain, so one change to the station can fix both. That helps the floor directly: less strain, fewer rushed recoveries, and a line that does not depend on people working around a problem.
Where to start
Pick the one line where downtime hurts most. Baseline its cycle times, choose a handful of alerts, name an owner for each, and review the results every week for a month. If the data changes what your team does on that line, expand. If it does not, find out why before you add more cameras.
If a line at your plant keeps losing time that nobody can fully explain, Invisible AI can show you the stations behind it. We will walk through your own cycle data and the footage behind it, so your team sees what it has been missing. Request a demo.