The factory losses hiding in plain sight
A 90-second stop every 15 minutes does not look like a crisis. The operator deals with it, production resumes and there is little reason for anyone else to notice. Because the interruption is brief, it may never be recorded as a maintenance issue or discussed at shift handover. Repeated across a day, however, those lost minutes begin to erode output. Over time, the bigger problem is that people stop seeing the interruption as abnormal and start treating it as simply how the line runs.
John Davagian, CEO of L2L, a connected operations platform for frontline manufacturing, says the productivity opportunity often sits in the gap between what a plant formally records and what operators routinely absorb during a shift. “Major failures are pretty easy to see because they are visible,” he says. “It is the resource coordination, the material delays, the line running at reduced speed, the repeated adjustments and micro stoppages that are really hidden. They add up and compound, and operators tend not to report them because they become part of normal life. At that point, the abnormal becomes normal.”
The phrase captures a weakness in many continuous improvement programs. Plants measure downtime, OEE and output, but the events that erode those numbers are often reconstructed later from incomplete records. A quick calibration, short material delay or repeated reset can disappear into the working day, even when the same problem is stealing production time shift after shift.
When the workaround becomes the process
Davagian argues that the important measurement is not simply how often an abnormality occurs, but the time between recognizing it and taking effective action. In asset-intensive discrete manufacturing, a line that repeatedly loses 90 seconds can miss its daily target without ever suffering a major breakdown. The lost time includes the event itself, the delay in finding the right person and the time spent diagnosing a problem that may already have happened before.
“If a machine is out of material or there has been a failure, the call might still happen by radio, by pager or by somebody walking across the plant to find help,” he says. “That waiting time until somebody gets to the unit and starts diagnosing and fixing it is lost productivity. If the event is not documented, the next shift may go through exactly the same thing again. Our customers come to us because they want to shorten that gap from identifying the abnormality to fixing it and eliminate the recurring events.”
The response delay also erodes the evidence needed to stop recurrence. “Memory does not preserve all the details,” Davagian says. “When you capture the configuration, the product being run, who was operating the machine and what was happening at the point of failure, you have the information needed for continuous improvement. Then you can move into decision and execution mode and make sure the same event does not keep recurring.”
Building that history allows manufacturers to find patterns that would otherwise remain invisible. The objective is not a more detailed post-mortem but recognizing that the quick reset an operator performs several times each shift is a recurring production constraint worth eliminating.
Digital recording alone does not close the response gap. Replacing a paper checklist with a tablet may make information easier to capture, but performance changes only when an abnormal entry leads to action while there is still time to influence the shift.
From visibility to execution
Manufacturers have spent years connecting machines and building dashboards, yet visibility on its own has not delivered the productivity improvement many expected. Davagian sees the missing step as execution: deciding what should happen next, assigning responsibility and confirming that the intervention worked.
“A dashboard tells you what happened. It does not assign a person, put the problem in priority or guide somebody to fix it,” he says. “Digitizing what is happening does not by itself help the plant improve. The value comes from processing what is happening in the moment and dispatching the right resources, whether that means people, work orders or production decisions.”
L2L structures its platform around data, decision and execution layers. Signals from people, systems and machines are placed in operational context; software and increasingly AI can help prioritize a response; the resulting action is coordinated through frontline workflows or passed into existing CMMS, MES or ERP systems. The aim is to shorten the route between seeing an abnormality and doing something that changes its production impact.
JELD-WEN, a manufacturer of doors, windows and other building products, provides a useful example of what can happen when more of that work becomes connected. The company had relied on Excel and printed paperwork before implementing L2L. It subsequently reported a 53 percent improvement in overall performance, a 90 percent reduction in administrative paperwork, up to two hours saved per employee per day on administration and 96,000 additional door skins produced per week eight months after implementation.
The gain extends beyond maintenance. JELD-WEN also uses the platform for training, onboarding, checklists and reporting, showing how time lost to finding information, recording work and coordinating responses can become part of the same productivity problem as machine downtime itself.
AI should remove noise rather than add to it
Adding AI to this operating loop raises a harder question: whether it reduces the work of responding or simply increases the volume of information competing for attention. “AI creates noise when it simply creates another dashboard, another alert or another report,” Davagian says. “It is most productive when it detects the problem, helps decide what to do and carries out the response by assigning a person and giving them the context to fix it. AI is very good at identifying abnormalities, prioritizing them and showing likely causes, but if all we do is spit out more content to workers on the shop floor, they still have to process all of that information.”
Historical production data can also reveal when a line is performing well rather than concentrating exclusively on failure. Davagian points to periods during a shift when equipment settings, materials and operating conditions align and the line approaches its best performance. Comparing those periods with weaker ones can help manufacturers understand what needs to be reproduced, rather than using analytics only to diagnose what went wrong.
AI becomes more consequential once it starts influencing frontline decisions rather than simply identifying abnormalities. At that point, adoption depends as much on how the technology is introduced into everyday work as on the quality of the model itself.
“The worker has to be in charge,” Davagian says. “Leadership needs to put guardrails around what AI can recommend, and the worker needs the ability to accept or reject it. You use that feedback as a learning loop. If leadership deploys AI on the assumption that AI is always right, that is where trust starts to fail, because the person on the front line knows when a recommendation does not reflect how the process works.”
Better operational information can also change the relationship between production and maintenance. Operators often feel ignored when they request help without knowing whether anybody has received or prioritized the call, while maintenance may see a different version of the problem. Making ownership, status and history visible shifts the conversation away from who caused the downtime and towards preventing its recurrence.
The bigger opportunity is capacity. Davagian argues that manufacturers should identify the production time hidden inside recurring friction before assuming that growth requires another line, more equipment or additional floor space. Industry 4.0 connected machines and generated vast quantities of data, but information gathered for retrospective analysis cannot recover a shift after it has finished.
For factories where the abnormal has quietly become normal, the most valuable improvement may not begin with a major transformation project. It may begin with noticing the 90 seconds everybody stopped noticing years ago.
