Visual Production Signals: Define the Response Behind Each Colour

A visible problem still needs someone to answer it

A finishing operator raises a help signal because the rack of approved packing materials is empty. The lamp works. Everyone passing the station can see it. Yet no one knows whether the material handler has taken the call or whether the shift leader must find another person.

After a while, the operator asks a colleague to walk to the store. The lamp stays on. Later, a visitor sees the same light and assumes that help is already coming. The factory has made the problem visible without creating a reliable response.

Visual management in a garment factory should help people decide what to do next. A useful production signal identifies the place, the condition and the person expected to act. It also shows whether the request is still waiting, someone owns it, or the work can resume under the approved method.

The examples below use invented stations, times and local rules. They explain how to design and test a response system. They are not required response times, universal colour meanings or instructions for emergency controls.

Give each display a clear job

A layout drawing helps someone find a station. A work instruction explains how to perform a task. A performance board compares results with a plan. A help signal asks for action on a condition that exists now. These displays may sit close together, but they have different users and update needs.

The Lean Enterprise Institute's explanation of andon describes a visual tool for showing operating status and abnormalities. It can also call attention to a needed action. The response is part of the idea: a station signal should bring the relevant leader or support person to the problem.

That is a more useful design goal than covering a wall with coloured information. A monthly efficiency result does not tell a handler which station needs cartons now. An active material call does not explain whether the line met its weekly plan.

Keep the live response display close to the people who can use it. Put longer-term results where the team reviews patterns and decides improvements. ApparelDiary's factory KPI dashboard guide develops that wider performance-review task.

The two can share data without sharing every field. A help request needs a station and time immediately. A later review may need its cause, duration and repeat history. Asking the operator to write a long failure report before calling for help delays the purpose of the signal.

Start with the response at one finishing station

Consider station F04. Its operator finds that the marked rack has no more of the approved packing material for the current order. The remaining material at the workstation may allow some work to continue, but an unapproved substitute cannot be used simply to avoid a stop.

The proposed signal calls the material handler. The operator identifies F04, the order and the missing material reference. The handler accepts the call, checks the need and brings the correct released supply. The operator confirms that it matches before the service request is closed.

If the material cannot be supplied, the handler reports that fact to the shift leader. The leader then decides the next permitted production action. That decision might involve waiting or moving to other authorised work. It does not change the material specification or release a quality hold.

This sequence reveals what a lamp alone leaves out. Someone has to receive the call. Someone has to decide whether the response is sufficient. The operator needs to know the current position while waiting.

Toyota's own production-system explanation connects a worker's call button with an andon board that informs the leader. Its account also links problem handling with the conditions for restarting work. A garment factory can learn from that connection without copying a car assembly line's stop logic or fixed-position system.

Write the meaning before choosing the colour

For each signal, write a short response rule in ordinary language. Define the trigger precisely enough that two trained workers would raise the same kind of request for the same condition.

“Materials problem” may cover an empty rack, the wrong item, a held lot or a missing label. Those cases do not always need the same responder. In F04's example, the call means that the approved supply is missing at its assigned point. A suspected wrong item follows the separate quality route.

The rule should also explain what the operator may do while waiting. That instruction belongs to the actual work and product controls. It cannot be inferred from a general colour such as amber.

Once the meaning is clear, select colours that fit the site's existing visual language. Use the same label for the same state on cards, screens and station lights. A colour that means “help coming” in one area should not mean “no one has answered” in another.

Do not make the legend carry information that the live signal lacks. If several stations use one lamp, the responder still needs to know which station called. A clear location code and plain state label are often more valuable than another colour.

Separate a reply from a restored service

The word “acknowledged” often causes trouble. A person may press a button to silence a sound without accepting responsibility for the problem. Another team may read the same action as confirmation that the issue is fixed.

Define the event. In this example, a first reply means that a named responder has accepted the request. It does not mean that the person has arrived, the material is correct or production is ready.

A simple service record can show three stages: waiting for a reply, owned and under action, and closed after confirmation. Keep the current owner visible during the middle stage. A name left in an old roster is not proof that someone has seen today's call.

Production state and product state need separate fields where they matter. The station might still be working with its remaining approved supply while a material request is open. It might also have a closed material request but a separate quality hold on one batch.

Turning off the help light must not release that held batch. Product release belongs to the authorised quality decision and its record. Likewise, closing a service call does not approve a new machine setting or a different work method.

This distinction prevents one green symbol from making several promises. A display can say “material request closed” accurately. It should not claim “all work ready” unless the system has checked every condition that statement requires.

Give the responder enough time and support to act

The HSE's alarm-management guidance says that an alarm needs a useful, defined response and enough time for the operator to carry it out. Its main setting is plant alarm design. The same questions are useful when reviewing an ordinary production help signal, although a help board is not a safety alarm system.

Choose a response target after understanding the work. A handler who serves several floors cannot promise the same arrival time as someone beside the station. Walking distance, current duties and the need to collect the correct item all affect the answer.

For the worked log later in this article, the fictional team chooses a first-reply target of three minutes. This is a local trial rule, not an industry benchmark. It asks someone to take ownership within that time; it does not promise to solve every request in three minutes.

The backup route matters just as much. State who receives a call when the primary handler is on a break or already handling another issue. If two urgent calls arrive together, a leader may need to set the order and explain the delay to the waiting station.

Escalation should add help or a decision. Sending the same message to more people without naming an owner can make responsibility less clear. The current owner should retain the case until another person explicitly accepts it.

Make the message readable without colour alone

On a digital board, a coloured dot with no label leaves meaning dependent on colour perception. W3C's explanation of WCAG 2.2 use of colour calls for another visual way to convey the information, such as text or shape. This guidance concerns web content; it does not certify a physical factory sign.

For the factory's own design, the same basic idea gives a useful practical test. Show “HELP WAITING” with the station code, or “OWNED” with the responder. Keep the wording visible when the display is viewed in greyscale or a colour is hard to distinguish.

Use the language that the intended users understand. A translated label should retain the same action and state. Adding more languages in tiny print can make every version difficult to read, so test the layout from the actual viewing position.

Lighting, glare, height and obstructions also matter. A label that looks clear on a desk may be hidden behind hanging garments or a trolley at the station. Ask workers to read and interpret the display during normal work, including the other shift.

A text label helps only people who can read that label. Where users need other forms of communication, design them with those users. The test is whether they can find the station, understand the condition and identify the next action, rather than merely recognise a colour.

Keep ordinary help signals distinct from safety communication

The factory's safety signs, emergency alarms and equipment controls have their own purposes and applicable rules. An ordinary production board should not overwrite their meanings or delay the required response to a hazard.

The HSE's safety-signs guidance, written for its UK regulatory context, stresses clear, legible signs and warns that too many signs can confuse people. It is a useful reminder to review the complete workplace display, rather than adding a new set in isolation.

Before release, have the relevant site owner check the proposed labels, locations and colours against existing safety communication. Follow the applicable local requirements for that safety system. This article does not supply a replacement set of warning symbols or an emergency-response procedure.

The F04 material request is a routine support call. If a worker identifies a safety concern, the site's safety response applies immediately. The person should not have to wait for the three-minute trial target or the next performance meeting.

Use an event log that preserves the waiting time

The simplest useful log records when the request was raised, when a named person first replied, when the responder arrived and when service was restored. Add the station, issue reference, owner and closure evidence to the underlying record.

Use one agreed clock and a clear date. If a request crosses midnight or a shift boundary, retain the original start time. Replacing it with the next shift's start makes an old problem appear new.

The following invented log covers eight new requests. All times are on the same date. The first seven are restored and closed after confirmation. H remains unanswered at the noon cutoff. The clock intervals below are elapsed minutes, with no excluded breaks.

The seven reply times are 2, 3, 4, 1, 5, 2 and 9 minutes. They total 26 minutes:

Mean first-reply time for answered requests = 26 ÷ 7 = 3.71 minutes.

Sorted, the times are 1, 2, 2, 3, 4, 5 and 9 minutes. The middle value is 3 minutes, so the median is 3. The longest completed reply time is 9 minutes. Together, these figures say more than the mean alone.

At noon, H has already waited 12 minutes. It does not have a completed reply time. Entering zero for H and dividing by eight would produce 26 ÷ 8 = 3.25 minutes, an apparently faster result caused by an unanswered call.

Report the open request beside the completed-time summary. Its age is a live fact. Its final reply time is still unknown.

Check the denominator behind the response target

Four answered calls met the three-minute target: A, B, D and F. Within the seven answered calls, that is 4 ÷ 7 × 100 = 57.1%.

However, all eight new requests had reached their three-minute deadline by noon. Only four met it. For the full set of requests due by the cutoff, the result is 4 ÷ 8 × 100 = 50.0%.

Neither percentage should appear without its denominator. The first describes answered calls only. The second includes the unanswered call whose target has already expired. A recently raised request whose deadline had not yet arrived would need separate treatment, not an automatic failure.

The log also gives arrival waits of 5, 4, 5, 3, 7, 5 and 15 minutes for A through G. Their mean is 44 ÷ 7 = 6.29 minutes. The time to restore service totals 111 minutes, or 15.86 minutes per completed request.

Those are different measures of the same events. They are not three pools of lost production time to add together. Reply and arrival occur within the longer period before restoration. Nor is every waiting minute a stopped production minute: some stations may have continued approved work while support was on its way.

Reconcile the open work as well as the completed calls

Suppose two older requests were still open when this log began. Eight new requests arrived, and the seven completed cases in the table were closed. No other closures or transfers occurred.

Closing open requests = 2 opening + 8 new − 7 closed = 3.

The three are the two older requests and H. A board filtered to today's arrivals could hide the older two even while its arithmetic looks tidy. Keep the open-case view available across dates and shifts.

Repeated button presses also need a clear rule. Several presses about the same unresolved shortage should not create several independent material jobs. Retain the notifications if they help explain the experience, but link them to one issue record. If a genuinely new problem occurs after a confirmed closure, link the new event to the earlier case without deleting either history.

When a request is withdrawn or merged, record why. Do not use deletion to make the queue smaller. A disappearing row gives the operator no evidence that anyone understood or resolved the need.

Keep the original reason alongside any later diagnosis. An operator may report missing material, while the handler finds that the supply was placed at the wrong station. Both facts help: the first explains the call, and the second points toward a change in delivery practice. Replacing the first entry can hide what the operator actually faced.

Test stale information and silent failures

A quiet board can mean that work is stable. It can also mean that the connection failed, the light lost power or workers stopped using the system. Test these possibilities before trusting an empty display.

A digital board should make its update status visible. In a separate example, a board viewed at 14:00 last received data at 13:42. Its information is eighteen minutes old. That gap should be apparent; a calm green display should not imply a fresh check of every station.

Choose the allowed update delay for the actual response need. A daily report and a live help board need different rules. Define how people call for help if the display or network is unavailable, and make the fallback known before the outage occurs.

For a physical card or lamp, test the whole path. Can the worker raise the signal? Can the intended responder see it from the normal work area? Does the correct station appear? Can someone else clear it by mistake?

A test call should be labelled and excluded from live performance counts. Its purpose is to check the route, including the backup responder. A working bulb alone does not prove that the response system works.

Hand over the open case, not just the board

At a shift change, the outgoing owner should explain each unresolved case, the actions already taken and the next check. The incoming person confirms the current condition and accepts responsibility. The original request time stays with the case.

HSE's shift-handover guidance describes preparation, exchange of relevant information and an incoming-team cross-check. Applied here, that means the display supports a conversation rather than replacing it.

For F04, “material issue open” gives too little information. The next person needs the order, missing reference, supply check already made and any permitted work arrangement. They also need to know whether the goods have a separate hold that the service call cannot clear.

If a temporary solution restored service, record its limits and review point. A closed call may still leave a longer-term action, such as correcting a repeated supply error. Keep that action in the improvement record while the live board reflects the true current condition.

Review whether the signal changed the work

Begin with a small area and observe complete cases. Ask operators what they expected after raising a call, and compare that with what actually happened. A short demonstration should include a missing responder, an open case at handover and an attempted early reset.

Look for signals that remain on after the condition is resolved, calls cleared before help arrives, and requests made outside the system. These patterns reveal different design problems. More training will not repair a missing backup role or a display nobody can see.

Review repeat causes as well as reply times. Faster replies can improve support while the same shortage continues every day. Conversely, fewer calls are not proof of improvement if workers have stopped expecting help.

Compare like periods before claiming a lasting gain. A quieter shift, extra support or an easier mix of requests can shorten the mean without changing the response process. Keep those conditions with the results. For a longer review, ApparelDiary's statistical process control guide explains why a few favourable figures do not establish stable improvement.

The useful result is a shared understanding of the condition and the next action. Each live signal should lead to an owner, a response and a clear close. Its colour helps people notice the message; the working arrangement behind it makes the message worth raising.

References

• Lean Enterprise Institute: andon and the response to an abnormal condition.

• Toyota Motor Corporation: production system, jidoka and the andon call.

• Health and Safety Executive: alarm-management principles.

• W3C: understanding WCAG 2.2 success criterion 1.4.1, use of colour.

• Health and Safety Executive: safety signs and signals, L64 guidance.

• Health and Safety Executive: shift handover.