Andon: the escalation promise behind the light

The call is a commitment made by management

Asking an operator to declare a problem the moment it appears transfers risk upwards. The operator gives up the option of quietly working around it; in exchange, the organisation owes a response. If that response is a supervisor arriving irritated, or arriving much later, or asking why the line was interrupted, the trade collapses within days. Plants that get this right rehearse it deliberately at the start: early calls are answered fast and visibly, even trivial ones, because the first pattern sets the expectation for years. The technology choice is minor beside the question of who is genuinely free to walk over.

Tiers, intervals and the point at which work halts

A workable scheme names what the operator does first, how long they have to recover the situation themselves, who is called next, and at what point production stops. Each tier needs a person, not a role that is permanently in a meeting. The interval should be short enough that the problem is still live and long enough that a competent operator can fix simple things without an audience. Escalation past the team leader must reach somebody with authority to spend money or change the plan, otherwise the higher tiers are ceremonial and the floor learns to stop at the first level.

Silence is the symptom to worry about

Managers are pleased when the board is green and the cords are quiet. Usually it means the calls have gone underground. Operators are practised at absorbing problems: reworking quietly, taking a spare from a drawer, adjusting a setting that was not theirs to adjust. Those behaviours look like resilience and destroy the information the system exists to gather. A falling call rate deserves the same curiosity as a rising one, so go and ask people what they did the last time something went wrong and whether they told anyone. Where the honest answer is that reporting brings trouble, fix that before touching the equipment.

The cost of stopping, and who may decide not to

Stopping has a real price, and pretending otherwise makes the policy unbelievable. It is also usually smaller than the price of a defect travelling onwards, being caught at test or reaching a customer, because recovery cost climbs with every operation the fault survives. What the plant needs is an explicit rule about who may authorise continuing to run with a known problem, recorded at the time. Leaving that decision informal means it gets made by whoever feels most pressure that day, which is how a documented stop policy manages to coexist with a floor that never stops.

The call log is the best problem list in the building

Every call is an observation of something the process could not absorb. Sorted by station and cause, the log tells you which problems recur, which take longest to answer and which escalate past the first tier, and that ranking is generally more honest than any improvement wish list assembled in a meeting room. It only works if calls are recorded with a cause in words the operator would recognise, and if the list is worked through visibly, because a log that accumulates without resolution becomes one more reason not to call.

Frequently asked questions

Our operators almost never raise a call. Is that a good sign?

Rarely. It means either that the process has genuinely stabilised, which shows up as high delivered quality and few unplanned stops, or that calling has become socially costly. Test it by asking a handful of operators what they did the last time material was wrong or a tool failed. If the answers describe private workarounds, the system is reporting silence rather than stability, and the remedy is management behaviour rather than a better signalling device.

Should the line stop automatically when a call is raised?

It depends on how coupled the stations are and whether stopping is recoverable. On a moving assembly line, a fixed-position stop that halts at the end of the cycle if the problem is unresolved gives the responder a window without freezing everything at once. On independent machines or in a process plant, automatic stopping may be unsafe or expensive to restart, so escalation runs on people rather than interlocks. Decide it area by area, and write down what happens if nobody answers.

How does escalation work in a low-volume shop with no line?

The same contract applies without the hardware. The signal can be a flag on the machine, a magnet on a board or a radio call; what matters is that raising it is quick, visible to the person who must respond, and free of consequence for the person raising it. Low-volume shops usually need longer response intervals because responders cover more ground, so state that interval honestly instead of promising an answer that will not arrive.