Unattended Night Operation in Thai Factories 2026 — Capacity Over Labor Cost

“We shut the plant down at night, but could we at least keep the machines running?” We are hearing this question more and more often from Japanese manufacturers with plants in Thailand. Unattended night operation is one of the few levers that can genuinely make a capital investment pay back. When you evaluate it in Thailand, however, the Japanese line of reasoning does not carry over. Thai labor law contains no wage premium for night work, so the Japanese premise that “nights are expensive, therefore replace people with machines” collapses before the analysis even starts. This article walks through the questions a plant manager, production engineering lead, or executive in Thailand should ask first, in order.

What Unattended Night Operation Actually Means — Partial, Not Total, Automation

Start by pinning down the definition

Outside Japan, factories that run without people, including unattended night shifts, are described as lights-out manufacturing. Standard Bots, an industrial robotics publication, defines it as a production method in which no operators are present on the floor, machines, robots, and control systems handle everything from material supply through assembly and inspection, and people are involved only in remote monitoring, maintenance, and exception handling.

The important part is what the same article says next. In practice, most companies adopt a hybrid configuration in which robots take on high-repetition volume processes while people retain monitoring and exception handling, and that arrangement remains, in the article’s words, only partially lights-out. In other words, the mainstream pattern is that certain processes or certain machines run unattended at night or on holidays, while engineering and maintenance stay on day shift.

That picture is a long way from the image often invoked in Japanese manufacturing of a pitch-black factory with the lights switched off. Once you actually begin scoping a project, the reason becomes clear. Taking an entire plant unattended founders not on technical difficulty but, well before that, on the sheer volume of preparation required outside the machine itself, including material supply, finished-goods discharge, first-response to abnormalities, quality assurance, and shipping preparation.

So the first question to ask is not “can our plant go unattended?” It is “which of our machines, for how many hours of the night, and under what conditions, can run without a person present?” Simply reframing the question this way sharply raises the resolution of the whole study.

Think in four stages, not in all-or-nothing terms

Unattended operation is not a binary. Treating it as a sequence of stages makes the investment decision far easier. BRICS, in its explainer on unmanned factories, likewise lays out a phased roadmap that moves from partial automation to robot deployment, then to integration with IoT and MES, and on to validation through simulation. Recast for night-shift operation, it looks like this.

For most Japanese-owned plants in Thailand, the realistic goal is Stage 2, or Stage 3 when conditions allow. Aiming straight at Stage 4 sends the investment figure through the roof and destroys the payback argument discussed later in this article.

This is not a new story

Plenty of articles present unattended night operation as the latest trend, but it has in fact been a practical operating model for many years. One perennial reference case is the Philips plant in Drachten, the Netherlands, which makes electric shavers. The Standard Bots article cited above classifies it as a partially lights-out factory and notes that 128 robots handle the bulk of assembly, with nine people responsible for quality assurance.

That case has been written up repeatedly in industry media since the early 2010s. It is not a 2026 development. The reason to cite it here is not to prove that unattended operation has recently become possible; the point is the opposite. The technology has been viable for more than a decade, and if most factories still shut down at night, it is because the non-technical conditions are not in place.

Three Premises That Differ From Japan When You Plan Night Operation in Thailand

This section is the core of the article. Head office in Japan sometimes hands down an instruction to “cut labor cost through unattended night operation,” but the premise behind it does not hold in Thailand. There are three reasons.

Premise 1 — Thai labor law has no night-work premium

Section 61 of Thailand’s Labor Protection Act B.E. 2541 (1998) requires overtime pay of at least 1.5 times the hourly wage of a working day for work beyond normal hours. That provision contains no uplift based on whether the hours fall at night.

What the Act does impose with respect to night hours is a restriction on employment itself, not a wage premium. Section 39 prohibits an employer from having a pregnant employee work between 10 p.m. and 6 a.m., work overtime, or work on holidays. Section 47 prohibits an employer from having an employee under 18 work between 10 p.m. and 6 a.m., subject to exceptions such as a permit from the Director-General or a designated representative. Section 40 provides that where a female employee performs work between 10 p.m. and 6 a.m. and a labor inspector considers that work hazardous to her health or safety, the inspector reports to the Director-General or a designated representative, who then considers the matter and orders the employer to change or reduce her working hours (the order comes from the Director-General’s side, not directly from the inspector). All three are prohibitions or restrictions on employment applied to protected categories of workers. None of them adds anything to wages. A prohibition and a premium are entirely different concepts, and conflating them leads to a badly wrong business case.

Nowhere in the Act is there a provision that raises wages for employees generally on the basis that the hours fall late at night.

Under Japan’s Labor Standards Act, by contrast, night work between 10 p.m. and 5 a.m. carries a premium of at least 25 percent, and where night work and overtime overlap the premium is at least 50 percent. When a plant in Japan evaluates unattended night operation, it is entirely natural for that night premium to be the leading item in the payback calculation. In Thailand, that leading item simply does not exist.

The table below normalizes every figure to the uplift added on top of base pay, so the two countries can be compared on the same basis. Thai overtime is often quoted as “150 percent” because that number refers to the total amount paid after the uplift; as an uplift it is 50 percent.

One nuance deserves attention here. Plants often feel, from day-to-day experience, that the night shift costs more in labor. That feeling is not wrong. But the reason it costs more is not that the hours are at night. It is that some or all of those hours fall outside normal working hours and therefore count as overtime or holiday work. Our own article on palletizing robots models night-shift labor at a higher rate than day shift, but that gap does not arise from the time of day as such. It arises from the portion of the shift that is treated as overtime or holiday work because of how the roster is built.

The distinction feeds straight into the investment decision. If you run a three-shift rotation and keep the night shift inside normal working hours, no premium arises at all. Put plainly, in Thailand you cannot book a large saving from the idea that “replacing the night shift with machines frees up night premiums,” because there were none to free up.

Premise 2 — Overtime and holiday work combined are capped at 36 hours per week

In Thailand, what makes unattended night operation worthwhile sits on the capacity side, not the labor cost side.

Section 26 of the Labor Protection Act B.E. 2541 (1998) provides that overtime hours worked under Section 24 and holiday hours worked under Section 25 must not, in aggregate, exceed the hours prescribed by ministerial regulation, and the limit set by that regulation is 36 hours per week. Rivermate, a global HR information service, sets out Thai working hours the same way, noting that overtime is capped at 36 hours per week and that there is no statutory premium for night work as such, with ordinary overtime rates applying to hours beyond the normal schedule. That cap is the practical ceiling for Japanese-owned plants here.

When demand rises, most plants reach first for overtime from the existing workforce. But that headroom has a legal limit. A plant that has consumed its 36 hours cannot produce any further increase through overtime.

The remaining options are moving to three shifts and hiring more people. Both are legitimate, but both take time and money. Consider the training period before a new hire is productive, retention rates, the difficulty of attracting applicants to rotating shifts that include nights, and the need to place supervisors on the night shift as well. None of these resolves quickly. Securing a supervisor able to make decisions on the night shift is, in particular, a bottleneck at many Japanese-owned plants.

Unattended night operation should therefore be positioned as a way to route around that ceiling without adding headcount. It is not an investment to cut labor cost; it is an investment to grow capacity without being bound by a statutory limit. Reframing it this way also changes which numbers matter in the investment case. The lead items become the gross margin on the added output and the avoided opportunity cost of orders you no longer have to turn away, not the size of a cost reduction.

Premise 3 — Labor savings alone rarely produce a workable ROI

The third premise is a conclusion that has emerged consistently across the FA articles we have published to date. At Thai hourly labor rates, an effect built solely on labor savings will not bring the payback period into a practical range.

Our article on collaborative robot deployment models a labor rate of 75 baht per hour and an initial investment of 2.5 million baht. Case 1, which runs two shifts and counts only labor savings, pays back in roughly 11.5 years. Only in Case 2, which adds three hours of unattended night running and brings previously outsourced work in-house, does the figure shorten to roughly 6.3 years. Case 3, replicating the cell on a second machine, comes to roughly 4.5 years, but that number assumes unattended night operation is already running on the first machine and that reusing the design and programs cuts the investment by 28 percent.

The palletizing robot article shows the same pattern. Case 1, with labor savings only, comes to roughly 23.1 years, which is not a viable investment case at all. Case 2, which builds in reduced outsourcing and holiday work through unattended night operation, plus quality improvement and fewer workplace injuries, comes to roughly 7.8 years.

Our article on assembly automation robots reports roughly 12.02 years for the baseline configuration and roughly 9.63 years for a lighter configuration with a narrowed specification. Adding the effect of avoiding the need for a third shift as demand grows brings it to roughly 6.36 years.

What all three articles share is that the opening case, counting labor savings only, lands at or beyond ten years, and that the number collapses the moment a condition that increases operating hours is added. That structure is exactly why unattended night operation keeps reappearing as the decisive lever in FA investment.

What Can and Cannot Be Left Unattended at Night

Conditions that make a process a good candidate

A process is well suited to unattended night running when it meets the following conditions.

• It can run for long stretches on a single setup, with no part-number changeover falling at night
• Material for the whole night can be staged in advance, such as coil stock, bar stock, or palletized blanks
• Finished parts can be discharged and accumulated without a person, using a conveyor and stocker, or palletizing
• Good-or-bad judgment can be automated, or the line can be stopped in a way that prevents defects reaching downstream
• An abnormality can bring the machine to a safe stop, and leaving it in that state until morning does not make the damage worse

That last condition is easy to overlook and decisive. The essential risk of unattended night operation is not that something breaks. It is that the machine keeps producing all night after it breaks. Eight hours of scrap wipes out the day’s added output in an instant.

The changeover wall in variable-mix, variable-volume production

The single biggest reason unattended night operation gets difficult at Japanese-owned plants in Thailand is variable-mix, variable-volume production. A plant that runs many part numbers in small lots goes through changeover several times a day. Changeover means swapping fixtures, swapping robot end effectors, switching programs, and confirming the first piece through inspection. None of these is easy to make work without a person.

There are two realistic responses. One is to fix a single part number for night running and confirm that its monthly requirement can be absorbed within continuous night operation. The other is to install automatic changing mechanisms for end effectors and fixtures so several part numbers can be handled, but that sends the investment figure sharply higher and belongs to Stage 3 and beyond.

Put differently, the plants where unattended night operation is most worth evaluating are those with a mainstay part number whose volume is predictable and unlikely to change for the foreseeable future. Run that part at night and give day-shift hours, with people present, to the low-volume parts whose demand swings. Whether you can make that split is what decides applicability. On how to select the process itself, our article on factory automation consulting sets out five evaluation axes, namely actual labor hours, defect occurrence, stability of demand, safety and working environment, and suitability for 24-hour operation. It is worth reading alongside this one.

Technical approaches to variable-mix production

Research and development is also moving toward reducing the burden of changeover itself. Yaskawa Electric’s Technical Report 2026 No. 5, published on 7 August 2026, presents three application cases of AI-equipped robots and human-collaborative robots for variable-mix, variable-volume production in food manufacturing. The industry differs from ours, but the underlying question of how to automate a process with frequent changeover applies directly to metalworking and assembly floors.

In the topping process case, MOTOMAN NEXT automates sauce filling on hamburger-steak bento boxes using AI recognition that compares a reference photograph against the actual image. In the produce loading case, AI Cube’s AlliomWorks VegeFruPutter recognizes the features of an apple to align its orientation, re-shoots from a different angle when it cannot recognize the item, and retries automatically when a grip fails. In the palletizing case, CoboPal3 is built on the MOTOMAN-HC35 human-collaborative robot with a payload of 35 kg and a maximum reach of 2,030 mm, and a newly developed torque sensor is reported to improve collision response time and strengthen safety.

The direction the report sets out is to combine short deployment time (Easy to Setup) with operability that the shop floor can handle (Easy to Use). That is instructive for unattended night operation as well. Equipment that takes a long time to bring up means calling in a specialist engineer every time the part number changes, and the hours you can actually run unattended never increase.

Even with such technology in place, however, the scope you can leave unattended shrinks as long as changeover falls within the night window. You have to separate what technology can ease from what the production plan has to avoid in the first place.

Why poka-yoke devices are a precondition for night operation

On a manned day-shift line, an operator notices an abnormality and stops the process. At night, with no one there, that function disappears entirely. Everything a person would have caught during the day therefore has to be caught by the equipment. That is why poka-yoke devices are a precondition for unattended night operation rather than a nice-to-have.

The poka-yoke functions that matter most for night running are detection of missing parts, detection of wrong part orientation or wrong part type, judgment of fastening torque and press-fit load, automatic dimensional measurement after machining, and a mechanism that reliably ejects any piece judged NG so it never reaches the next process. As our article on assembly automation robots notes, what decides success or failure in assembly is not the accuracy of the robot itself so much as the tolerance and variation of the parts being fed in. At night there is no one to absorb that variation, so the requirement is stricter still.

Technical Requirements Behind Unattended Night Operation

Designing anomaly detection and alerting

The first thing to design for an unattended window is how an abnormality is detected and how, and to whom, it is communicated. Newji, in its explainer on rolling out factory automation and unmanned night operation, likewise treats risk management as especially important once people are removed, and points to remote monitoring systems and alert functions as the way to respond quickly to equipment failure and unexpected trouble.

In practice, building detection in three layers leaves fewer gaps.

• Alarm signals from the equipment itself, such as cycle abnormalities, torque abnormalities, tool breakage, and homing failure
• Continuous monitoring of production results, such as cycle-time deviation, zero output over a set period, and sudden changes in defect rate
• Environmental monitoring, such as camera-based motion detection, acoustic anomaly detection, and deviation in hydraulic pressure, air pressure, or temperature

The third layer, cameras and sound, has become a common configuration for unattended windows in recent years. Ceiling-mounted cameras can track the movement of people and AGVs to flag contact risk, and motion detection can be combined with acoustic anomaly detection to isolate and notify only those events that occurred during the unattended window. Recording equipment error signals together with the linked video footage also cuts the time spent on root-cause analysis the next morning by a wide margin. The single biggest time sink in unattended operation is arriving in the morning and having to start the investigation from a position of not knowing why the machine stopped.

Agreeing who is notified and what first response looks like

Plenty of plants build the mechanism that sends the alert and never decide what happens after it arrives. Before you start unattended running, the following need to be settled in writing.

• Who the primary contact is, and whether that person receives the alert in Thai or in Japanese
• After how many minutes without a response the alert escalates to a secondary contact
• Where the line falls between an event that warrants calling someone in and an event that can be left stopped until morning
• Entry and exit logging for anyone coming into the plant, and whether working alone is permitted

If you end up with a call-out every single night, labor cost does not fall and the person on the receiving end does not stay in the job. Whether you can clearly define the abnormalities that may be left stopped until morning without calling anyone is what determines whether unattended operation survives.

Tool life management and consumables

When machining equipment runs unattended at night, tool life management becomes the practical constraint. A plant where an operator judged when to change a tool has to begin by turning that judgment into numbers. The usual sequence is count-based management by number of parts machined, then spindle load monitoring, then preparation for automatic tool change.

Methods Machine Tools, in its material on unattended night running, describes configurations that deliver consistent high-precision work by pairing a FANUC RoboDrill with an LR Mate robot, along with equipment requirements such as automating a five-axis table and a tool changer capable of handling 28 tools. It is fair to treat spare tool capacity as directly convertible into the number of hours you can run unattended.

Beyond tooling, you also need to confirm that cutting fluid, air, dryer drain capacity, dust collector filters, sealants, labels, and similar consumables will last the night. It is unglamorous, but unattended operation fails on exactly this kind of detail.

Material supply and finished-goods discharge

As our factory automation consulting article points out, the simple equation that “running 24 hours improves capital efficiency” does not hold on its own. It holds only once the elements outside the machine are in place, namely material supply, finished-goods discharge, abnormality notification, quality assurance, the maintenance regime, and automatic recording of production results. Those additional costs should be built into the plan from the outset.

Running eight unattended hours at night requires eight hours of material within reach of the machine and eight hours of space to put the finished parts. This is where additional investment in stockers, conveyors, and AGVs appears, and it is not unusual for that spend to match the cost of the machine itself. Leaving it out at the quotation stage means going back for a supplementary budget after the capital request has already been approved.

The Investment Case — Built on Added Output, Not Labor Savings

What our published models show

Lining up the models we have published across our FA articles makes the pattern clear.

The assembly automation robot entry in the middle column is constructed differently from the others. That 6.36-year figure is not the baseline configuration with longer operating hours. It comes from switching to a lighter, narrowed specification at roughly 9.63 years and then adding the effect of avoiding the need for a third shift as demand grows. The other two rows add effects against the same investment amount, so although the figures sit in the same column, the underlying structure differs and the table should be read with that in mind.

The assumptions behind each article also differ, so this table is not intended for a direct numerical comparison across rows. What you should take from it is the shared structure. In every article, labor savings alone fail to produce a viable investment case, and only when a condition that increases operating hours is added does the payback period become realistic.

Change how you book the benefits

Because Thailand has no night premium, you cannot book the benefit of unattended night operation as a reduction in night-shift allowances. The items to book instead are these.

• Contribution margin on the added output, meaning profit after variable costs on the additional volume sold
• Reduced outsourcing cost from bringing previously subcontracted machining in-house
• Reduction in the volume that was previously covered by holiday working
• Avoided opportunity cost from shorter lead times, meaning orders you no longer have to decline
• Freeing up daytime machine capacity so that day-shift hours can be given to the parts with volatile demand

The most important item on that list is the first one, and the question attached to it is whether the added output will actually sell. Unattended night operation is an investment that increases capacity, so with no demand behind it the benefit is zero. Before you write the capital request, confirm three things with the sales side. Is there real inquiry volume for the additional output? Is that part number unlikely to change for the foreseeable future? And is the higher volume premised on a lower unit price? Increase capacity without checking, and all that remains is the depreciation charge.

Budget for the extras from the start

Japanese-market explainers put the initial investment for unattended operation somewhere between several million and several tens of millions of yen. Actual figures in Thailand vary greatly with the configuration, so what matters is less the headline number than making sure nothing outside the machine itself is left out of the scope. The BRICS article lists four implementation challenges, namely the scale of the initial investment, a shortage of specialists in areas such as PLC control and robot programming, securing installation space and safety clearances, and the small number of FA companies able to cover both hardware and software.

At Japanese-owned plants in Thailand, the third and fourth of those bite hardest. Forcing equipment into a gap between existing lines leaves insufficient safety clearance, which in turn produces an operating pattern with a person stationed outside the fence, and the unattended part never materializes. On the second point, equipment makers will deliver the machine, but few of them take responsibility all the way through upper-level system integration and the abnormality notification scheme.

Starting Small

Where to begin

The Newji article on rolling out factory automation and unmanned night operation sets out a sequence of steps covering assessment of the current state, target setting, cost estimation, technology research, pilot deployment, employee training, and continuous improvement. Made concrete for a Japanese-owned plant in Thailand, the order looks like this.

Step 1 is a continuous night-running test on your existing equipment. Before spending anything new, run the machines you already have continuously through the night with a person present, and record how many hours pass before they stop and what causes the stop. Once you know which stoppage causes recur, the target for investment narrows itself down automatically. At most plants this step alone surfaces concrete issues such as an inability to stage material in advance or the absence of automatic ejection for defects.

Step 2 is partial unattended running on one machine, one part number, for a few hours. Not the whole line. Take the single most suitable machine and the single most predictable part number, and start with just the two to three hours after the end of the working day. Use this step to put the alert routing and first-response criteria into live operation, and to standardize the next-morning check routine.

Step 3 is extending the unattended window. Extend the unattended hours in proportion to the track record of running without stopping. This is where additional investment in tool life management and material stockers becomes necessary, but because you have the evidence from Steps 1 and 2, both the target of the investment and the justification for the amount are clear.

Step 4 is replicating across machines and processes. Just as the second unit in the collaborative robot article came in 28 percent cheaper, capital efficiency on horizontal replication of the same configuration is far better than on the first installation. Positioning the first unit as the proof case for replication lets you avoid judging the whole program on the payback period of that first unit alone.

How not to do it

The approaches that fail are equally clear. Opening with a plan to take the entire line unattended at once. Building the payback case on reduced night-shift allowances. Choosing a part number with volatile demand as the night-running target. Building the alert notification without deciding the first-response criteria. All four either stall partway or break down operationally after go-live.

Frequently Asked Questions

What is unattended night operation?

It refers to combining robots, machining equipment, and transfer devices so that production continues on the equipment alone during hours when no operators are on the floor. Outside Japan it is called lights-out manufacturing. Most real deployments are not whole-plant but partial, with only certain machines or certain processes running unattended at night or on holidays. Monitoring is done remotely or by periodic patrol, and maintenance and engineering typically remain on day shift.

Do factories in Thailand have to pay a night-shift allowance?

Section 61 of Thailand’s Labor Protection Act B.E. 2541 (1998) sets overtime pay at not less than 1.5 times the hourly wage of a working day, and that rate is flat regardless of whether the hours fall at night. The Act contains no provision applying a night premium to employees generally. Restrictions on night employment for pregnant employees and employees under 18 do exist separately, but those are prohibitions and restrictions on working, not wage premium provisions. The Japanese-style payback structure of “automating the night shift to eliminate night allowances” therefore does not work in Thailand. Where a night shift does extend beyond normal working hours, ordinary overtime premiums apply, so actual labor cost depends on how the roster is built.

Can unattended night operation work with variable-mix, variable-volume production?

Yes, conditionally. It works within the scope where changeover does not fall at night, meaning within the length of time a single part number can run continuously. The deciding factor is whether your part mix includes a mainstay item with predictable volume, and whether you can assign that item to the night window while leaving the volatile low-volume items to manned daytime hours. Running several part numbers overnight requires automatic changing of fixtures and end effectors, which raises the investment substantially.

If we run 24 hours, will the investment pay back quickly?

Not necessarily. In the models published in our earlier articles, every case that counted only labor savings landed at or beyond ten years, and only when a condition that increases operating hours was added did the figure come down to the six-to-eight-year range. The benefit of longer operating hours also assumes the added output actually sells. Increase capacity without confirming the demand behind it and all that remains is the depreciation charge.

Is it wrong to automate in order to reduce labor cost?

The practical conclusion in Thailand is that a case resting on labor cost reduction alone struggles to win approval. Because the capital investment is large relative to prevailing hourly labor rates, the savings do not accumulate fast enough to repay it. The realistic approach is to book labor savings as one component of the benefit while evaluating it alongside added output, in-housing of subcontracted work, defect reduction, and lower workplace-injury risk.

What exactly should we do first to start small?

Before any new investment, run your existing equipment continuously through the night with a person present and record how many hours pass before it stops and what caused the stop. With that record in hand, both the target of the investment and the justification for the amount become clear. Next, take one well-suited machine and one part number and run just the two to three hours after the end of the working day unattended, putting the alert routing and first-response criteria into live operation.

Summary

When you evaluate unattended night operation for a plant in Thailand, the first thing to discard is the Japanese instinct that nights are expensive in labor terms and should therefore be given to machines. Since Thai labor law contains no night premium provision, that formula does not compute.

Two premises should take its place. The first is that overtime and holiday work combined are capped at 36 hours per week by law, so growth through overtime eventually hits a ceiling. Unattended night operation is a way to route around that ceiling without adding headcount. The second is that, as our FA articles consistently show, labor savings alone do not produce a viable payback, and the numbers become realistic only once operating hours increase.

Make the rollout phased. Rather than taking the whole plant unattended, start with partial unattended running on one machine, one part number, for a few hours, and extend the unattended window as you eliminate stoppage causes. Sequenced that way, both the investment figure and the justification in your capital request are built on evidence.

Unattended night operation is not something you achieve by buying equipment. It starts working only when the preparation outside the machine is in place, including material supply, anomaly detection, first-response criteria, and tool life management. The encouraging flip side is that much of that preparation can begin today without any significant spend.

If you would like help mapping out which machine, which part number, and how many hours you could start with at your own plant, or you want to talk through how to structure a continuous night-running test on your existing equipment, please get in touch through our contact page. TOMAS TECH is based in Bangkok and supports Japanese manufacturers in Thailand with FA and production management system deployment. Even at the stage before any specific equipment discussion, we can talk through where to start based on your current production data and part mix.

References

• Standard Bots, “Lights-out manufacturing in 2026”
• Methods Machine Tools, “Lights-Out Manufacturing with Methods Night Crew”
• Yaskawa Electric, Technical Report 2026 No. 5
• JETRO Bangkok’s Japanese-language translation of Thailand’s Labor Protection Act B.E. 2541 (1998)
• Rivermate, “Thailand Working Hours”
• Newji, “Steps to Implement Factory Automation and Nighttime Factory Unmanned Project”
• BRICS, “What is an unmanned factory”