An andon is a visual signal system that lets any operator flag a problem the moment it appears — with a cord, a button, or a digital trigger — summoning help to the line immediately and escalating to a line stop if the problem isn't resolved. The word is Japanese for a traditional paper lantern.
The signal is the easy part. What separates a working andon system from a decorative light is everything behind it: who responds, how fast, what happens when they can't fix it, and whether anyone ever looks at the pattern of calls. This guide covers the meaning, the history, the escalation path, and the cord-versus-digital question.
What does andon mean in manufacturing?
In manufacturing, andon means a signal — typically a light, a board, or an alert — that shows the status of production at a glance and calls for help when something goes wrong. The name comes from the Japanese word for a paper-enclosed lantern; in the plant, the andon board plays the lantern's role, making status visible across the floor.
A full andon system has three parts. The trigger: the cord, button, or automatic sensor an operator uses to raise the signal. The display: the light stack or board that tells everyone which station needs help and why. And the response process: the defined chain of people who come to the line, with time limits and escalation rules. Plants often buy the first two and skip the third — which is why so many andon lights blink at rooms where nobody moves.
The principle underneath is jidoka, one of the two pillars of the Toyota Production System and lean manufacturing: build quality in by stopping work when a defect or abnormality appears, fixing it at the source, and never passing a known problem downstream.
Where did the andon cord come from?
The andon cord comes from Toyota, and the idea behind it is older than the company's car business. In 1924, Sakichi Toyoda — founder of the Toyoda loom works that became Toyota — introduced the Type-G automatic loom, which stopped itself the moment a warp thread broke. One worker could tend many machines, and no machine could keep weaving defective cloth (Toyota 75-year history: the birth of jidoka).
Toyota's assembly plants extended that logic to people: a cord along the line that any worker could pull on detecting a problem, lighting the andon board and summoning a team leader — with the line stopping at a fixed point if the issue wasn't resolved. Toyota describes andon as a core element of how the production system surfaces problems for immediate attention (Toyota UK: Andon in the Toyota Production System; Toyota Motor Corporation: TPS).
The cultural point is the part most plants underestimate: the cord only works where pulling it is a protected act. If operators get grief for stopping the line, they stop pulling, problems flow downstream, and the andon becomes scenery. Management's response to the first dozen pulls teaches the floor whether the system is real.
How should andon escalation work?
Andon escalation should follow a defined path — operator to line lead to supervisor to specialist — with a time threshold at each step, so a signal always either gets resolved or gets more help. Here is a generic recommended pattern; tune the thresholds to your line speed, takt, and staffing:
- Operator raises the signal (minute 0). The operator detects the abnormality — quality issue, material shortage, machine behavior — and triggers the andon. The station keeps running if it safely can; the signal is a call for help, not necessarily a stop.
- Line lead responds at the station (within ~2 minutes). The first responder is the nearest team lead, at the station, looking at the problem with the operator. Most calls end here: a quick fix, a confirmed workaround, or a decision to escalate.
- Supervisor plus specialist engaged (within ~10 minutes). If the line lead can't resolve it, the area supervisor is engaged and pulls in the right specialist — maintenance for equipment, quality for product, materials for shortages. The signal's reason code should have already told them who's needed.
- Stop-or-run decision made (within ~15 minutes). If the problem still isn't contained, someone with authority makes an explicit decision: stop the line, or run with a documented containment. Never let the line drift on with a known, undecided problem — that is exactly what jidoka exists to prevent.
- Close the loop (same shift). Every escalated andon call gets a reason, a resolution, and — for repeat offenders — a root cause analysis. An andon call that ends with no record is a lesson the plant paid for and didn't keep.
The thresholds above are a starting pattern, not a standard — a high-speed bottling line and a machine shop will set very different clocks. What matters is that the thresholds exist, everyone knows them, and missed thresholds escalate automatically instead of relying on someone remembering to make a call.
Andon cord vs digital andon: what changes?
A digital andon keeps the cord's principle — anyone can summon help instantly — and adds routing, time-stamping, and memory. The physical cord (or button) is beautifully simple, and its weaknesses only show at the system level: it alerts whoever happens to be nearby, it doesn't track response times, and yesterday's pulls vanish unless someone logged them by hand.
| Physical cord / button | Digital andon | |
|---|---|---|
| Raising the signal | Pull or press at the station | Tap on a tablet or station screen; sensors can also trigger automatically |
| Who gets alerted | Whoever sees the board or hears the tone | The specific responder for that station and reason — with automatic escalation if they don't respond in time |
| Reason capture | Verbal, or written later (often never) | Reason code captured at trigger time |
| Response time tracking | None, unless timed by hand | Time-stamped at every step automatically |
| History and patterns | Lives in memory and anecdotes | Queryable record of every call: station, reason, duration, resolution |
| Failure mode | Signals ignored; no accountability trail | Alert fatigue if thresholds are set lazily; garbage reason codes if the UI is slow |
The honest trade-off: digital andon fails differently, not never. If every minor blip pages a supervisor, alerts get ignored within a month; if logging a reason takes six taps, operators pick the top of the list every time. A good digital andon is designed like the cord — one motion to raise, one glance to understand — with the routing and record-keeping invisible underneath. This is a core piece of connected worker technology: the operator's signal, the responder's alert, and the record all ride the same system.
What does andon data tell you over time?
Over time, andon data becomes a map of where your process is weakest: which stations call most, for what reasons, at what times, and how long help takes to arrive. A single andon call is an event; a thousand are a diagnosis.
Frequency by station and reason shows you where to aim improvement effort — the station that calls three times a shift for material shortages has a materials process problem, not an operator problem. Response-time trends tell you whether your escalation path actually works or just exists on a poster. Duration and resolution patterns separate chronic small stops — which rarely make it into a paper downtime log but quietly eat availability — from the big breakdowns everyone already knows about.
This is where the digital version pays for itself: the calls become data you can rank, trend, and act on. That's the pattern Harmony builds toward with plants — operator signals, downtime events, and quality checks captured once at the station, visible in real time, and mined for the recurring causes worth fixing permanently. Chattanooga Labeling Systems saw the general version of this shift when production logging moved off paper: problems that used to surface in the next morning's report now get attention during the shift they happen.
Start with the response process, not the hardware. Define who responds, how fast, and what happens on timeout; protect the operator who raises the signal; and keep the record. The lantern is the easy part.