Insight 08 · Smart factory
The machine that tells you why it stopped.
For most of its history a factory knew its output at the end of the shift: a tally sheet by the machine, a count at the gate, and a foreman’s memory of why the line had stood. Today each machine reports its state as it changes, a stop is an event with a reason against it, and the shift’s effectiveness is counted from the machines themselves, not claimed afterwards. This page is about how that came about, what is moving in factories now, where the market stands, how a modern line is run, and what it returns.
Every era moved the question from output to cause.
First the question was how many parts the shift made. Then whether the machine was running, then why it had stopped, then which machine held the whole line back. The question each time was the same: where did the time go, and can you show it?
Time study and the punch clock; tally sheets at the machine; the foreman’s shift report.
Output was counted at the end of the shift; why a machine had stood was a memory.
The programmable logic controller in place of relay panels; a stack light on the machine.
The machine’s logic moved into software; its state was a lamp someone had to see.
Operator screens and SCADA on PCs; fieldbuses; manufacturing execution systems; overall equipment effectiveness from total productive maintenance.
The line’s state reached the control room, and OEE gave availability, performance and quality one name.
Ethernet on the plant floor; a common way to read controllers from a PC, leading to OPC UA (2008).
One maker’s machine could be read like another’s, from outside its cabinet.
Small sensors retrofitted to old machines; MQTT and Sparkplug B; dashboards in a browser.
The forty-year-old press joined the line, and its data reached people outside the control room.
Security rules for connected machines and their makers; a right to a machine’s own data; monitoring that reads beside the control.
Every connection is asked what it can change; the plant keeps what it reads.
A connection that can’t make the machine unsafe.
In the European Union, from 14 January 2027, machinery must be designed so that connecting another device to it, directly or remotely, does not lead to a hazardous situation, and it must keep evidence of any intervention in its safety-critical software or its configuration.
EUR-Lex · Regulation (EU) 2023/1230, Annex IIIThe maker reports its flaws.
In the European Union, since 11 September 2026, the maker of a product with digital elements, a controller or a connected machine among them, gives an early warning of an actively exploited vulnerability in it within 24 hours of becoming aware of it; the rest of the rules apply from 11 December 2027.
EUR-Lex · Regulation (EU) 2024/2847, Art. 14The plant is in scope.
Under the European security directive, medium-sized and larger makers of machinery, vehicles, electrical equipment and electronics are among the entities covered; a significant incident is reported with an early warning within 24 hours.
EUR-Lex · Directive (EU) 2022/2555, Art. 23The machine’s data is its user’s too.
In the European Union, connected products placed on the market after 12 September 2026 must be designed so that the data they produce is, by default, easily and securely accessible to their user, free of charge, in a structured, machine-readable format.
EUR-Lex · Regulation (EU) 2023/2854, Art. 3OEE counted, not claimed.
Plants are moving from a figure typed in at the end of the shift to one built from the machines’ own states and counts, with each stop given a reason and each factor shown on its own.
industrial robots in operation in factories worldwide in 2025, a record.
IFR · World Robotics 2026new industrial robots installed in factories worldwide in 2025.
IFR · same releaseindustrial robots installed in the United States in 2025, now the second-largest market.
IFR · same releasegrowth in the world’s manufacturing value added in 2024.
UNIDO · Yearbook 2025of the world’s jobs are in manufacturing, a share that has slipped since 2015.
UNIDO · same reportfor a machine’s maker to give early warning of an actively exploited vulnerability in it, in the European Union.
EUR-Lex · Regulation (EU) 2024/2847, Art. 14A stop seen while it is short.
A machine that leaves its running state is known in seconds, not at the next walk-round, so the stop is cleared before it becomes the shift.
Stops with their reasons.
A record of why each stop happened turns the shift meeting from recollection into a list, longest first.
The line judged by its bottleneck.
Time recovered anywhere else is idle time; seen through its bottleneck, a line shows where an hour is worth an hour.
Three factors, not one number.
Availability, performance and quality shown apart point to the cause: a stop, a slow cycle or a reject, each with its own fix.
Maintenance by reading.
Run hours and cycles from the controller put the service where the wear is, and a recurring fault code is a pattern before it is a breakdown.
Control left where it belongs.
Monitoring that reads and never writes can’t stop a machine or change a recipe, so the plant’s engineers can let it in.
These are the mechanisms; the size of each depends on the plant, its products and its lines. We don’t publish savings figures.
- 01
What the controllers publish
States, counts and fault codes, read from what each machine already exposes. Nothing written.
- 02
States and stops
Every machine’s state over time, so each stop has a start, an end and a length. The first weeks are spent watching.
- 03
Reasons and shifts
A short list of stop reasons, answered at the machine, and OEE per shift once the counts are trusted.
- 04
The bottleneck
The line seen through the machine that holds it back, and the work orders that follow.
Which radio and which device, for each step: the decision table →
Where we come in
Forty minutes, free, about your plant.
You leave with the five questions answered for one site and our view, in writing, on whether an assessment is worth doing. If you ask, you see the platform running; if you don’t, you won’t. Chirp Technologies advises plant managers on this — which signals a machine already publishes, how to read a line without touching its control, how to count OEE so a shift trusts it, the order to go in — and built HOLOIN, which runs on your own servers, reads machines over OPC UA and Sparkplug B and never writes to them, and shows each machine’s states over time, each stop asking for its reason, and OEE per shift by availability, performance and quality, with “not measured” where a factor isn’t.