Method guide · 5S and visual management
Andon and jidoka: stop, call, respond, with a response-time example
Short answer
Jidoka is the rule: when something is wrong, stop and fix it rather than pass it on. Andon is the signal that makes the rule work on a line: an operator pulls a cord or presses a button, a light shows where, and a team leader comes within a time the site sets. If the problem is not fixed in that time the station or line stops, and if it is still not fixed the call climbs to the supervisor and then the managers, by the clock. Measure it with the median and 90th percentile time to respond, the time to resolve, pulls per shift and the reasons that repeat, and treat every pull as the system working.
Jidoka is the rule, andon is the signal
Jidoka started with a loom. Sakichi Toyoda's power loom of 1896 had a device that stopped it when the weft thread broke or ran out, so it could not go on weaving bad cloth, and Toyota's own history names these stopping devices as the origin of jidoka in the Toyota Production System. A loom that stops itself does not need someone watching it, so one person could look after several. Jidoka became one of the two pillars of the system, next to just-in-time (see the TPS house template).
On a line the same idea needs people. Machines can stop themselves with a sensor, a counter or a poka-yoke device, but many problems are only seen by the operator: a scratch, a part that does not fit, an empty bin. The andon (the Japanese word for a lamp or lantern) is how the operator stops the work and calls for help without leaving the station. Our jidoka guide covers the four steps of a stop (detect, stop, fix, investigate the cause). This page covers how the stop and the call work on a real line, and how to measure them.
| Form | How it is triggered | Good for |
|---|---|---|
| Cord or button at the station | The operator pulls or presses it | Calls from people: a defect, a missing part, a question |
| Light on the machine (stack light) | The machine's own sensors or controls, or the operator | Showing each machine's state from across the floor |
| Overhead board | Any pull or machine signal on the line; the board lights the station's number | Showing the team leader where to go, and the whole line's state at once |
| Pager, phone or screen | The same pull, sent to the responder's device | Responders who are not always on the line: maintenance, materials, quality |
Most lines use more than one: a cord or button to call, a board or light to show where, and a device for the people who are not on the line.
If you are choosing software for the devices and the board, our andon software comparison covers eight tools.
The three states and the stop rules
| State | What it means | What happens |
|---|---|---|
| Green | Running to standard | Nothing. Nobody has to watch the line to know it is fine. |
| Yellow | Someone has called | The light and the sound go on at the station's number and the team leader goes there at once. On many paced lines the line keeps moving to the end of the cycle. |
| Red | Stopped | The problem was not fixed in time, or it was serious enough to stop at once. The station or line waits until the responder restarts it. |
Some sites add a color for material calls, often blue or white. Keep to as few colors as people can read at a glance.
A line stops in one of two ways, and the site has to choose before the andon goes live. The fixed-position stop is the one Toyota uses on its assembly lines: stopping mid-cycle leaves work half done and makes the restart confusing and less safe, and a line that only stops if the problem outlasts the cycle makes pulling a smaller decision for the operator.
| Design | How it works | Where it fits |
|---|---|---|
| Stop at once | The pull stops the station or machine straight away | Machines and cells where each station runs on its own, and any problem where carrying on makes bad parts |
| Fixed-position stop | The line keeps moving to the end of the work cycle, a set point on the line. If the team leader fixes the problem and resets the call before then, the line never stops; if not, it stops at that point | Paced assembly lines with a team leader close to every station |
- Who may pull: anyone at the station, without asking first. A rule that needs permission is not an andon.
- What the team leader does first: go to the station, ask what the operator saw, then fix and reset the call, or let the line stop and call the next level.
- How fast the team leader must arrive. With a fixed-position stop the team leader has only until the end of the cycle, so the team leader must be close by. Where the station stops at once, the site sets a time in minutes. Write the number down.
- Who may restart: the responder, after the fix and a check that the next part is good.
- What the andon is not: the emergency stop. A danger to people is stopped with the machine's emergency stop first; the andon then calls for help.
- What gets logged: every pull, including the ones fixed in a minute.
Who responds, and the escalation ladder
Stop, call, respond, resolve, restart
| Call type | Examples | First responder | Called next if not running again in 10 minutes |
|---|---|---|---|
| Quality | A defect found, a part that does not fit, a gauge that fails | Team leader | Quality technician |
| Material | An empty bin, the wrong part, the wrong packaging | Team leader | Materials handler (water spider) or stores |
| Equipment | A tool fault, a jam, a machine alarm | Team leader | Maintenance technician |
| Help | A question on the method, relief needed | Team leader | Supervisor |
The team leader answers every call first, fixes what can be fixed at once and brings in the support function for the rest.
| Time since the pull | Who is called | Trigger |
|---|---|---|
| 0 minutes | Team leader | The pull |
| 5 minutes | Supervisor | Nobody has arrived yet |
| 10 minutes | Supervisor and the support function for the call type | The station is not running again |
| 30 minutes | Area or production manager | Still not running |
| 60 minutes | Plant manager | Still not running, or at any time for a safety or customer risk |
Sites set their own minutes. Count from the pull, not from when the responder arrived, so a slow response cannot hide a long stop.
Write the ladder into your escalation matrix by call type, post it at the line, and review yesterday's escalations in the tier 2 meeting. The ladder only works if each level can see that it has been called: a light the supervisor cannot see from the office is not an escalation.
What to measure
| Measure | How to work it out | What it tells you |
|---|---|---|
| Time to respond | Arrival time minus pull time, for each pull. Report the median and the 90th percentile, not only the average | Whether help comes. The slow calls are the ones operators remember |
| Time to resolve | Restart time minus pull time, for each pull. Median, 90th percentile and total minutes stopped | How long problems hold the work. A fast response with a slow resolve points to spares, tools or skills |
| Pulls per shift | Count by shift, station and call type | Where problems show up. A sudden fall with no fall in defects means people have stopped pulling |
| Repeat reasons | The same reason at the same station three or more times in a week (an example threshold) | What to fix for good. A repeat is a countermeasure that did not hold |
| Escalations | Pulls that went past the team leader, by level | Whether the ladder is used, and which problems the line cannot fix alone |
Use the median and the 90th percentile because response times are lopsided: most calls are answered in a few minutes and a few wait a long time. One 14-minute wait barely moves the average, but it is the call the operator remembers next time. To find the 90th percentile by hand, sort the times and take the one at rank 0.9 × n, rounded up (the nearest-rank method). Excel's PERCENTILE.INC interpolates between two values and gives a slightly lower figure on a small sample. Pick one method and keep it, so weeks can be compared.
Log each pull on the andon response log, which works out the minutes and flags calls over target. When stops cost the line output, the OEE calculator and the downtime cost calculator turn the minutes into availability and money.
Worked example: one shift, 14 pulls
An illustrative day shift on an eight-station assembly line, using the example ladder above: the team leader within 5 minutes, the supervisor if nobody has arrived by 5 minutes or the station is not running by 10, the area manager at 30 and the plant manager at 60. Each station stops at the pull and restarts when the problem is fixed.
| Pull | Called | Station | Type | Reason | Respond (min) | Resolve (min) | Highest level called |
|---|---|---|---|---|---|---|---|
| 1 | 06:22 | Station 4 | Material | Fastener bin empty at station 4 | 2 | 6 | Team leader |
| 2 | 06:51 | Station 2 | Quality | Scratched housing from the upstream cell | 3 | 9 | Team leader |
| 3 | 07:35 | Station 4 | Material | Fastener bin empty at station 4 | 1 | 5 | Team leader |
| 4 | 08:10 | Station 6 | Equipment | Torque tool fault at station 6 | 2 | 34 | Area manager |
| 5 | 08:18 | Station 4 | Material | Fastener bin empty at station 4 | 14 | 18 | Supervisor |
| 6 | 09:02 | Station 1 | Help | Question on the new label layout | 1 | 3 | Team leader |
| 7 | 09:40 | Station 4 | Material | Fastener bin empty at station 4 | 3 | 7 | Team leader |
| 8 | 10:15 | Station 3 | Quality | Clip missing on two units | 4 | 16 | Supervisor |
| 9 | 10:24 | Station 7 | Material | Wrong carton size delivered | 12 | 19 | Supervisor |
| 10 | 11:05 | Station 6 | Equipment | Torque tool fault at station 6 | 2 | 8 | Team leader |
| 11 | 11:48 | Station 4 | Material | Fastener bin empty at station 4 | 3 | 6 | Team leader |
| 12 | 12:30 | Station 5 | Help | Relief needed for a first-aid visit | 1 | 4 | Team leader |
| 13 | 13:12 | Station 2 | Quality | Scratched housing from the upstream cell | 4 | 9 | Team leader |
| 14 | 13:55 | Station 6 | Equipment | Torque tool fault at station 6 | 5 | 11 | Supervisor |
Respond: from the pull to someone arriving who could help. Resolve: from the pull to the station running again.
One shift: waiting for help, then fixing
Example numbersTime to respond. Sorted, the 14 times are 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 5, 12 and 14 minutes. The median is the mean of the 7th and 8th: (3 + 3) ÷ 2 = 3 minutes, inside the 5-minute target. The 90th percentile is the 13th value (14 × 0.9 = 12.6, rounded up to 13): 12 minutes. PERCENTILE.INC in Excel gives 9.9. The average is 57 ÷ 14 = 4.1 minutes, which looks fine and hides the two pulls that waited 12 and 14 minutes.
Time to resolve. Sorted: 3, 4, 5, 6, 6, 7, 8, 9, 9, 11, 16, 18, 19 and 34 minutes. The median is (8 + 9) ÷ 2 = 8.5 minutes, the 90th percentile 19 minutes, and the stations were stopped for 155 minutes in all. Five of the 14 pulls went past the team leader: four reached the supervisor, and one, the torque tool at 08:10, reached the area manager at 30 minutes.
Pulls by reason. Three reasons make up 10 of the 14 pulls (71%): fasteners at station 4 (5 pulls, 36%), the torque tool at station 6 (3) and scratched housings from the upstream cell (2). By minutes stopped the order changes: the torque tool comes first with 53 of the 155 minutes (34%), then the fasteners with 42 (27%). Look at both: the count shows what interrupts people most often, the minutes show what costs the most time.
Pulls by reason: by count and by minutes stopped
Example numbersPulls (14)
- Fasteners, st. 4: 5, 36% of the total, cumulative 36%.
- Torque tool, st. 6: 3, 21% of the total, cumulative 57%.
- Scratched housing: 2, 14% of the total, cumulative 71%.
- Other: 4, 29% of the total, cumulative 100%.
Minutes stopped (155)
- Torque tool, st. 6: 53, 34% of the total, cumulative 34%.
- Fasteners, st. 4: 42, 27% of the total, cumulative 61%.
- Wrong carton: 19, 12% of the total, cumulative 74%.
- Scratched housing: 18, 12% of the total, cumulative 85%.
- Missing clip: 16, 10% of the total, cumulative 95%.
- Other: 7, 5% of the total, cumulative 100%.
- 1
Fasteners at station 4: fix the replenishment, not the operator
Five pulls for the same empty bin is a repeat reason: the bin runs out between deliveries. Size the bin to the delivery interval with the kanban sizing calculator and check the timing of the water spider route. This one fix removes about a third of the pulls.
- 2
Torque tool at station 6: keep a spare and find the cause
Three faults and the most minutes stopped. The 34-minute stop was long because there was no spare tool in the crib. Keep a calibrated spare at the line and run a breakdown analysis on the tool.
- 3
The two long waits: a rule for a second call
Both came while the team leader was already at another pull: the 08:18 call during the torque tool stop at station 6, and the 10:24 call while the team leader was at station 3. The fix is a rule, not a reminder: when the team leader is busy, the next call goes straight to the supervisor or a second responder.
- 4
Scratched housings: send it back upstream
Two quality pulls for the same defect from the upstream cell. Raise it with that cell the same shift with a quality alert, so the cell stops making it instead of this line sorting it.
- 5
Leave the help pulls alone
The label question and the relief call are the andon doing its job. The label question is worth a one point lesson for the next shift.
Make pulling safe
An andon only works if people pull it. Jeffrey Liker's The Toyota Way (2004) makes this the fifth of its 14 principles: build a culture where people stop to fix problems, so quality is right the first time. On the floor that culture comes down to what happens in the first minute after a pull, every time.
- Thank the person who pulled, before asking anything else.
- Never count pulls against a person or a team. A pull is a problem found at the station instead of at the customer.
- Judge the response, not the call: the leaders' response times go on the board, the operators' pull counts do not.
- Close the loop: tell the operator what was done about the cause, at the station or in the next huddle.
- Watch for silence. Fewer pulls with the same defect rate means problems are being hidden, not solved.
Common mistakes
| Mistake | What it looks like | Fix |
|---|---|---|
| An andon nobody answers | The light is on and nobody comes, or the team leader is away in meetings | A response target, a back-up responder, and leader standard work that keeps the team leader on the line |
| Too many stops from poor signals | Sensors that trip on nothing, one sound for every call type, a board nobody can read from the line | Fix false triggers first, give each call type its own signal, and check the board can be read from where the responder stands |
| Stops hidden by makeshift fixes | Spare fasteners in a pocket, a part reworked on the bench, a tool held together with tape. No pull, so no record | Ask about workarounds on the gemba walk and treat each one as a pull that was not made |
| Only the average is reported | An average response of 4 minutes, while two calls a shift wait over 10 | Report the median and the 90th percentile, and review every call over target |
| Restarting without a check | The line restarts as soon as the responder walks away | Restart only after the fix and a good next part |
| The log is never read | Hundreds of pulls logged and the same reasons every week | Review yesterday's pulls in the daily meeting and give each repeat reason an action with an owner |
Go deeper
This page is the short version. These guides cover each part in full.
- Andon Systems: Visual Quality Alerts and Problem Notification in ManufacturingWhat an andon is, the five-step response cycle, light towers, boards and digital andon, and the data an andon produces.
- Jidoka in Lean Manufacturing: The Complete GuideThe four-step jidoka response, the tool groups that make it work, and how jidoka and just-in-time depend on each other.
- Quality at the Source: Building Quality Into the Production ProcessWhy each step checks its own work, the principle the andon puts into practice.
- Tier Boards: Cascading Daily Management ExplainedHow problems the line cannot fix move up the tiers, the meetings where escalations are reviewed.
- The 3-Step Framework to End Daily Issue FirefightingTurning repeat problems from the floor into causes that get fixed.
All our articles on this subject are in the 5s and visual management topic.
Free templates and tools
- Andon response logEvery andon call with response and fix times, flags against your targets, and counts by type and station.
- Escalation matrix templateWho is called for each type of issue and after how long, with severity rules, contacts and an escalation log.
- Tier 2 meeting agenda and escalation sheetA 15-minute tier 2 agenda in ten timed blocks, fixed roles and rules, escalation triggers with time limits, and a log that ages every escalation and sends the answer back down.
- Jidoka worksheet (stop at abnormality)For one line or cell, every abnormality that should stop the work: how it is detected, how the work stops, who is called, the immediate fix and the root cause step, plus a log of real stops.
- Water spider route plan (mizusumashi)A water spider's fixed route: stops with arrival times, what is delivered and collected, route time against the cycle, cart load and a trip log.
- OEE calculatorOverall equipment effectiveness from shift time, downtime, ideal cycle time and part counts.
- Downtime cost calculatorWhat an hour, an event and a year of unplanned downtime cost, from lost output, idle labour, scrap and overtime.
Doing this in LeanSuite
Daily Issue Management
Gives frontline teams one place to flag, track and resolve issues in real time, so the same fires stop coming back every week.
See Daily Issue ManagementTiered Meetings
Puts daily huddles on live, cascaded dashboards, so the shop floor and the top floor see the same numbers and act on them at every tier.
See Tiered MeetingsQuality Improvement
Log quality issues from the floor, trace each one to its root cause and follow the corrective action through to closure.
See Quality Improvement
See it on your own lines
Bring the problem you are working on to a 45-minute demo. Or look around with sample data first, or pilot LeanSuite on your own lines for 90 days, with the setup done by our team.
FAQ
Questions about andon and jidoka
Rather see it on your own problem?
Bring the problem you are stuck on, and we will show you on a call how LeanSuite handles it.