Key takeaways
- An andon is a visual signal that makes a problem visible the instant it happens, so the right help arrives fast instead of the line quietly limping on.
- The usual colours are green for running, amber for a developing issue or a call for help, and red for a stop; the scheme only works if everyone reads it the same way.
- Andon is how jidoka (build in quality, stop on abnormality) becomes visible on the floor: surface the problem, call for help, contain it before it spreads.
- The number that matters is response time. Cutting the wait between a signal and someone acting on it is where andon pays for itself.
An andon system is a visual signalling method that makes a problem on the line visible the moment it happens, so the right person can respond before the issue turns into scrap or lost output. In its simplest form it is a coloured light on a machine; in its fullest form it is a networked system that times every response, records why the line stopped, and escalates automatically when no one answers. This guide explains where the term comes from, how andon works, the common signal colours and types, the stop-and-fix rule behind it, a worked example of why response time is the number that matters, and the mistakes that stop andon from delivering.
Where andon comes from and what it does
Andon is a Japanese word for a paper lantern, borrowed into the Toyota Production System to describe the light that shows the status of a workstation or line at a glance. The idea is deliberately low-tech in spirit: a problem you cannot see is a problem no one fixes, so you give every abnormality a signal that is impossible to miss. When an operator hits a button or pulls a cord, a light changes and the whole area knows that this station needs attention now.
The point is speed of response. On a line without andon, a developing fault (a jam clearing itself for now, a part measuring near the limit, materials running low) tends to stay in the operator's head until it becomes a full stop or a batch of defects. Andon shortens that gap to seconds. It turns a private worry into a public, timed call for help, which is the difference between a two-minute recovery and a twenty-minute one.
The signals: colours and what they mean
Most plants use a small, fixed colour set so the meaning is unambiguous from across the floor. The exact scheme is a local standard, but a common convention looks like this:
- Green: running normally, at rate, no action needed.
- Amber / yellow: a developing problem or a call for help that is not yet a stop, for example low materials, a minor quality concern, or a request for a supervisor or technician.
- Red: the station or line is stopped, or a fault needs immediate attention; work does not resume until it is resolved.
- Blue (optional): often reserved for a quality issue or a specific call such as a material handler or changeover team.
The colours are only half of it. A signal with no defined response is just a light. Each colour needs a named owner, an expected response time, and a clear rule for what happens if the call is not answered. That discipline, not the hardware, is what makes andon work.
Types of andon
Andon spans a wide range of sophistication, and starting simple is usually the right call:
- Stack lights (light towers): the classic three-colour tower on a machine, driven by the PLC or a button. Cheap, instant, and readable across the bay.
- Andon cord or button: the operator-triggered signal made famous on assembly lines, where anyone can call for help or stop the line without asking permission.
- Andon board: an overhead display showing the status of every station on a line, so supervisors see the whole area at once.
- Digital / software andon: signals raised from PLCs, sensors or a screen that log the event, time the response, capture the stop reason, and escalate up a chain if no one answers within the target.
The value of the digital tier is that it does not just show the problem, it measures the response and remembers it. That turns a wall of lights into data you can act on, which is where andon connects to the six big losses and your OEE.
Stop and fix: the rule behind the light
Andon is the visible face of jidoka, the principle of building in the ability to detect an abnormality and stop rather than pass a defect downstream. The cultural core of a strong andon system is that pulling the cord is expected, even rewarded, not punished. An operator who signals a small problem early is doing exactly the right thing; a plant where people fear stopping the line will get a wall of green lights and a stream of defects.
In practice the rule is contain first, then solve. The signal buys immediate help to stop the problem spreading, and the recurring calls become the input to root cause analysis so the same andon does not keep lighting up. Many of those recurring calls are the small, self-clearing stops that never reach a log, the ones covered in micro-stops and the hidden factory.
Why response time is the number that matters
The return on andon is almost entirely in how fast a signal turns into action. A worked example makes it concrete. Suppose a packaging line raises about 15 andon calls per shift, and today the average wait between the signal and someone starting to act on it is 5 minutes. Tighten the process, staff the response, and add escalation, and you cut that wait to 90 seconds:
- Time saved per call: 5 min - 1.5 min = 3.5 minutes.
- Per shift: 3.5 min x 15 calls = 52.5 minutes of recovered running time.
- Over 240 operating days (single shift): 52.5 min x 240 = 12,600 minutes, about 210 hours a year of downtime removed from one line.
That recovered time is availability you were already paying for. To put a currency figure on it for your own line, drop the minutes into the downtime cost calculator, and track whether faster response actually lifts availability using the OEE calculator. The lesson is blunt: a beautiful andon board that no one answers within target saves nothing. Measure the response, not the lights.
Common mistakes, and how to keep andon honest
Andon fails in predictable ways. The most common is no response standard: the light goes amber and nothing is defined about who comes, how fast, or what happens if they do not. The second is a punishing culture, where operators learn that stopping the line brings trouble, so they stop signalling. The third is alarm overload, where everything triggers a light and the important calls drown in noise; a good scheme keeps the signal set small and meaningful.
The fix in every case is the same: pair each signal with a named owner, a target response time, and an escalation path, then watch the response data. If most calls are answered late, you have a staffing or layout problem, not a hardware one. If the same station lights up every shift for the same reason, that is a root cause to kill, not a light to admire. Andon is a feedback loop; judge it by whether stops get shorter and rarer over time.
From lights to data: choosing an andon or monitoring system
If you are moving beyond a stack light and a button, the question stops being "which light" and becomes "which system captures and acts on the signal". When you evaluate andon or machine-monitoring software, weigh response timing, automatic stop-reason capture, escalation when a call goes unanswered, and whether the events roll up into downtime and OEE you can trust. We keep a vendor-neutral rundown in the guide to the best andon system software.
The hardest part is usually not the signal but the reason behind it: on high-speed lines the calls that matter most are the brief micro-stops that clear before anyone can note why. The platform we recommend for that is Fabrico, which reads stops directly from the machine and uses computer vision to show the true cause on video, then routes the fix as a work order in a closed loop. It is EU-built, so your production data stays in EU jurisdiction (ISO 27001 / 20000-1 / 9001, supports audit-readiness). If you are weighing software to turn andon signals into acted-on downtime, it is worth a look.
Book a walkthrough of how Fabrico surfaces the true cause behind each stop and closes the loop to a work order.
FAQ
What does andon mean?
Andon is a Japanese word for a paper lantern, adopted in the Toyota Production System for the visual signal that shows the status of a workstation or line. In practice an andon is any signal, a stack light, a board, a screen or a triggered alert, that makes a problem visible the instant it occurs so help can respond quickly.
What do the andon light colours mean?
The common convention is green for running normally, amber or yellow for a developing problem or a call for help such as low materials, and red for a stop or a fault that needs immediate attention. Some plants add a blue light for a quality issue or a specific colour for changeover. The exact scheme is a local standard; what matters is that everyone reads it the same way at a glance.
Is andon the same as jidoka?
They are closely linked but not the same. Jidoka is the principle of building in the ability to detect an abnormality and stop, so defects are not passed on. Andon is the visual signalling system that surfaces the abnormality and calls for help. Andon is how jidoka becomes visible and actionable on the floor.
Does an andon system need software?
No. A simple andon can be a stack light and a button, and many effective systems start there. Software helps when you want to time every response, log every stop reason automatically, escalate if no one answers, and turn andon events into downtime and OEE data. If you are choosing a platform, evaluate response timing, automatic reason capture and escalation rather than just the light on the wall.
Related: the six big losses · micro-stops & the hidden factory · root cause analysis · PLC / controls troubleshooting · best andon system software