Custom Automation Machine Design and Build | Cost Breakdown

Send the same specification document to three suppliers for a custom automation machine and you can easily get quotations that differ by a factor of 1.7. The reason is almost never that one builder charges more for steel than another. It is that none of the quotations says clearly where its scope stops. This article takes a Japanese-owned automotive parts plant in Chonburi, Thailand as a working model, breaks the cost of a purpose-built machine into five layers, and puts real numbers against each one so that you can judge a quotation yourself before you sign anything.

The model plant and the production conditions behind every number

Before talking about money, it is worth fixing exactly which plant and which process we are describing. Change the assumptions and every figure below changes with them. The model that follows was built by TOMAS TECH from the price levels we see in real projects in Thailand. It is not published statistics. Treat it as a measuring stick for reading your own quotation, not as a benchmark to quote back to a supplier.

The figure of 216,000 THB per year per operator deserves an explanation, because it is the number most often understated in an automation business case. It corresponds to roughly 18,000 THB a month, and it is not base pay. It is the total cost carried by the company, including statutory social security contributions, overtime, meal allowance, and transport to and from the plant. The minimum wage in Chonburi in 2026 is 400 THB per day, but even across 250 working days a year that comes to only 100,000 THB. The effective figure of 216,000 THB is more than twice that. Build your savings calculation on the minimum wage alone and you will value the removal of one person at less than half of what it is actually worth. The number that matters for an automation decision is not the legal minimum. It is the total company cost that genuinely disappears from the profit and loss statement when one head is no longer needed on that line.

For background, automation itself has stopped being an exotic choice. According to the International Federation of Robotics, 542,000 industrial robots were installed worldwide in 2024, Asia accounted for 74% of new installations, and the global operational stock reached 4,664,000 units, up 9% year on year. That is useful context for a board that still treats robots as a special project. It is not, however, an argument for automating your particular process. Robots becoming common and your specific station paying back are two entirely separate questions, and only the second one is dealt with below.

One more definition. This article is about a purpose-built machine, sometimes called a special purpose machine or a custom automation machine, designed for this process and no other. It is not about a catalogue machine that you select from a standard range. That distinction drives everything that follows, because a purpose-built machine is not really a purchase of hardware. You are buying the work of pinning down a specification. Every reason that quotations diverge, and every reason that payback is hard to read in advance, comes out of that single fact.

The cost of designing and building an automation machine splits into five layers

The cost of a purpose-built machine falls naturally into five layers. Layer 1 is mechanical. Layer 2 is control and electrical. Layer 3 is safety. Layer 4 is commissioning. Layer 5 is maintenance. Layers 1 to 4 make up the initial capital cost. Layer 5 is the money that keeps leaving your account every year after the machine is running.

The first thing to take from this table is that the visible machine, the metal you can actually put your hand on, is only 36.4% of the initial cost. You can walk a visitor through the plant, point at the cell and say that it cost 8,790,000 THB, but barely more than a third of that number is physically in front of them. The other two thirds went on design work, on programs, on judgement criteria, on the documented basis for the safety case, and on the act of making the thing run in production conditions.

That imbalance is the reason quotations diverge. Physical parts are easy to price and hard to omit. Design, software, safety evidence, and commissioning are easy to omit and hard to price, so different builders quietly draw the line in different places.

Layer 1 Mechanical is only 36.4% of the initial cost

Layer 1 covers the frame, the transfer and positioning mechanism, the feeding equipment, the tooling, the bought-in components, and the mechanical assembly and adjustment work.

The line item that most often falls out of a quotation inside layer 1 is the feeding equipment at 480,000. Whether a vibratory parts feeder will work at all depends on the shape of the part, so a builder who has not yet seen enough samples is genuinely reluctant to commit a price. Leaving it out makes the quotation look cheaper. It also means that when the machine arrives, somebody has to stand there and place parts one at a time, which erases the labor saving that justified the project in the first place.

The second thing worth noticing is that concept design and detailed mechanical design together come to 600,000, from 180,000 plus 420,000. Drawing work happens before anyone cuts steel. If the specification moves after the drawings are done, that 600,000 has to be redone, and everything downstream of it moves with it. This is the classic entry point for cost overrun on a purpose-built machine, and it is why the specification freeze phase deserves proper time. If you are outsourcing tooling on its own rather than a whole machine, we set out the equivalent breakdown in our guide to outsourcing jig design and manufacturing.

Why layer 2 Control and electrical is the largest at 41.5%

Layer 2 is the biggest single layer, at 3,650,000, or 41.5% of the initial cost.

There are three reasons layer 2 ends up on top.

The first reason is that vision inspection is expensive. Cameras, lighting, and the processing unit come to 740,000, and tuning the judgement thresholds adds another 420,000, for a combined 1,160,000. That is more than a third of the entire mechanical layer. It helps to stop thinking of it as the price of cameras and start thinking of it as the price of deciding where a defect begins. The work involves laying out limit samples, converting the boundary between good and bad into numbers, varying the lighting, and confirming that the result repeats. It is patient, iterative engineering, and it does not get cheaper because the camera did.

The second reason is that a robot does not run on its own hardware alone. The unit costs 680,000, and teaching plus application creation costs 260,000. Out of a combined 940,000, roughly 28% is the labor of making the arm do something useful. Anyone weighing up whether a robot belongs in the cell at all will find our criteria in the guide to choosing a robot system integrator.

The third reason is that software is a long-lived asset that quotations treat as an afterthought. The PLC program at 380,000 and the HMI screens at 180,000 will be used every single day the machine runs, and touched again every time the process changes. A quotation that prices this work cheaply usually turns out not to include handover of source code and configuration files. For the panel side of the equation, see our guide to ordering control panel design and manufacturing in Thailand, and for the software side, our guide to outsourcing PLC program development.

Layer 3 Safety cannot be trimmed

Layer 3 is 730,000, or 8.3% of the initial cost. It is the smallest layer by value, and it is the one you cannot cut.

Look closely at two of those lines. Risk assessment and documentation is 150,000, and safety circuit validation and records is 90,000. Together that is 240,000, roughly a third of layer 3, spent on things that are not objects. Neither of them affects whether the machine runs. Both of them are exactly what an auditor or an insurer will ask to see after an incident.

This is why comparing the price of the fence between two quotations is misleading. Two fences at the same price mean completely different things depending on whether a risk assessment document arrives with them. If it does not, your own manufacturing engineering team will end up writing it later, and those hours appear in nobody’s quotation. They are real hours all the same, and they usually land on the person who has the least time to spare.

Layer 4 Commissioning is what converts “it was built” into “it works”

Layer 4 is 1,210,000, or 13.8% of the initial cost.

Layer 4 is the money that moves the machine from a state of having been built to a state of being usable. The largest line is witness testing and the production trial run at 340,000. This is the first point at which real parts run through the machine and someone confirms that the required cycle time of 7.5 seconds is actually achieved and that the inspection judgement holds steady over a full batch rather than over ten hand-picked samples.

The line that gets overlooked most often at a plant in Thailand is operation and changeover training in Thai at 120,000. Hand over manuals in Japanese or English only, and the operators and line leaders who are standing at the machine at two in the morning cannot recover it when it stops. A machine that needs an engineer called out for every changeover does not just lose availability. It quietly turns back into a machine with a person attached to it, which is precisely what you paid 8,790,000 to avoid. Specify the language of the training at the request-for-quotation stage, not after the machine arrives.

Layer 5 Maintenance comes to 1,135,000 over five years

Layer 5 is what happens after handover. Over five years it comes to 1,135,000.

The largest item in layer 5 is neither inspection visits nor consumables. It is re-tuning of the inspection conditions at 400,000, running at 80,000 a year for five years. This cost is specific to machines that carry vision inspection, and it exists because the boundary between good and bad drifts. Material lots change, moulds wear, and lighting dims with age. A vision inspection system is not something you install and forget. It is equipment that needs looking after every year. Plenty of builders leave this line out of their quotation. Leaving it out does not stop the cost arising, it only means it arrives without a budget line to sit against.

A quotation gap of 1.7 times is a difference of scope, not a discount

This is the core of the article. Ask several builders to quote the same process and the numbers will come back far apart. In most cases that gap is not a discount. It is a difference in scope.

5,210,000 against 8,790,000. A factor of 1.7. And here is the uncomfortable part. The cheaper quotation is not wrong. For the scope it describes, the price is correct and probably competitive. The problem is that the 3,580,000 of work that was left out does not stop existing. Whatever is excluded stays with you. With no feeding equipment, a person lines up the parts. With no vision inspection, a person checks them by eye. With no safety documentation, your manufacturing engineering team writes it. With no commissioning package, your maintenance supervisor organises the transport, the crane, and the trial run in between their normal job.

Before you put quotations side by side, check what each supplier has written about the five boundary questions below. For each one, look for an explicit “included” or “not included” — and treat silence as exclusion. Items that nobody wrote down are almost never in the price.

Boundary 1 Who measures the variation in the part

A purpose-built machine is designed against the physical part, not against the drawing. That means nobody can design it without knowing how much the part actually varies. For a moulded resin component, shrinkage, lot-to-lot difference, and the way flash forms all matter, and all of them differ from the nominal drawing.

The question to settle is who measures that variation, when, and across how many pieces. Does the customer measure and hand over the data, or does the builder take samples away and measure them? If the builder measures, are those hours inside the quoted price? Skip this step and the design starts on assumptions. The machine then gets built, a different lot arrives, the parts jam, and the tooling has to be reworked at your expense or in an argument about whose expense it is. This is exactly why the specification freeze phase is given a full 4 weeks in the schedule later in this article.

Boundary 2 How much of the feeding problem belongs to the machine

Whether the 480,000 for two parts feeders is included has a direct effect on the labor saving that justifies the whole investment.

Are bulk parts going to be oriented by a feeder and presented to the machine, or will a person lay them out in a tray for the machine to pick from? The second option produces a cheaper machine and a person standing next to it. You end up with a plant where the machine was bought cheaply and one operator is still standing in front of it, and that is exactly what happens when feeding was never made the machine’s responsibility.

Write one line in the request for quotation describing the state in which parts arrive at the machine. Bulk in a bin, oriented in a tray, or in a magazine. Without that line, each builder will assume something different, and the quotations you receive will not be comparable at all, no matter how carefully you read them.

Boundary 3 Who decides whether a part passes inspection

This is about whether the vision inspection set at 740,000 and the inspection algorithm tuning at 420,000, together 1,160,000, are in scope.

Even when they are, there is a second question about who owns the judgement. Who produces the limit samples? Who signs off the pass line? If the machine calls a part defective, can a human overrule it, and if so, who is authorised to do that? Go into witness testing with this unresolved and the trial run turns into a debate about whether the machine is too strict or the quality standard was always that strict, and production start slips by weeks while two departments talk past each other.

The practical split that works is this. The customer produces the limit samples and sets the pass line. The builder converts that line into numbers and makes the machine reproduce it. Putting that one sentence into the request for quotation removes a large share of the arguments that normally surface at handover.

Boundary 4 Is the existing fence really enough

This is whether layer 3 at 730,000 is included. You will sometimes be told that safety is not required because the machine is going inside an existing guarded line. Do not accept that at face value.

Check three things. First, whether adding the machine has invalidated the existing risk assessment, because it usually has. Second, whether changeover or cleaning creates a new route by which a person approaches moving equipment. Third, whether pressing the emergency stop brings the new machine and the existing line to a coordinated halt rather than leaving one of them powered. If any of those three fails, the existing fence does not remove the need for a fresh safety case.

Then make the supplier state, in writing, whose deliverable the risk assessment document at 150,000 and the safety circuit validation records at 90,000 will be. The party that designed the machine is the natural author of both.

Boundary 5 Are commissioning and handover documents included

This is layer 4 at 1,210,000. Three points deserve particular attention.

First, the acceptance criteria for witness testing. How many parts, at what cycle time, at what defect rate, constitutes a pass? A quotation that does not state these as numbers has left acceptance to subjective judgement, and subjective judgement always favours whoever is more determined to declare the machine finished.

Second, the list of documents to be handed over. Manuals, mechanical drawings, electrical schematics, bill of materials, PLC program, HMI screen data, robot programs, and the inspection judgement conditions. The PLC program and the inspection judgement conditions are the two that genuinely do get withheld in practice. Without them, every future modification has to go back to the original builder, at whatever rate they choose to quote in three years’ time.

Third, the language and audience of the training. Is it delivered in Thai, to operators, line leaders, and maintenance as three distinct groups? Or does it amount to a briefing for the Japanese management team and nothing more?

The seven items to settle before you order

Take the five boundary questions above, add cycle time and maintenance, and you have seven items to write into the request for quotation before you send it out. A request for quotation that contains these seven puts every supplier on the same footing. Put the other way round, a set of quotations produced without them cannot be compared at all, and comparing them anyway is how plants end up choosing the supplier with the narrowest scope.

Item 4 carries more weight than its position in the list suggests. Writing that you want to move from 0.35% to 0.08% tells the builder to propose the camera, the lighting, and the judgement method capable of closing that specific gap. Writing “please include vision inspection” tells them nothing, and each supplier will pick a different level of capability and a different price to match. Numbers create comparability. Adjectives do not.

Labor saving alone does not pay for this machine

This section is deliberately blunt, because the alternative is a business case that falls apart in year three. In this model project, labor saving alone does not repay the investment within five years.

Follow the arithmetic. Six people become two, so labor cost falls by 864,000 per year, or 4,320,000 over five years. The five-year total cost is 9,925,000. That means the labor saving covers only 43.5% of what the machine costs across the same period. Build the internal approval paper on headcount reduction alone and you are asking the board to approve an investment that returns less than half its cost in five years. Boards notice.

What makes the case work is quality. Cutting the escaped appearance defect rate from 0.35% to 0.08% removes the 0.27% difference. Against annual output of 1,800,000 pieces, that is 4,860 pieces a year that no longer leave the plant as defects. At 180 THB per escape, that is 874,800 per year. Add that to the 864,000 of labor saving and you get 1,738,800 per year. Divide the five-year total of 9,925,000 by 1,738,800 and the simple payback is 5.7 years.

The 180 THB per escaped defect is not the cost of the part. It is the total cost carried by the company, covering sorting, re-inspection, freight, the paperwork at the customer’s site, and the internal hours spent handling the complaint. That number is different for every process and every customer, which is why you must derive your own rather than borrow this one. Until that figure is replaced with your own plant’s number, a purpose-built machine business case cannot get past the objection that labor saving alone does not cover it.

One more thing needs stating plainly. The increase in production capacity is not counted as a benefit here. The manual cycle time of 9.2 seconds becomes 7.5 seconds on the machine, so in principle the same hours produce more parts. But extra output with no customer attached is not cash. Capacity gain belongs in the benefit table only when the orders already exist and current capacity cannot serve them. Adding speculative volume shortens the payback period dramatically on paper, which is exactly why it is so tempting, and exactly why it should be resisted. It is a number that gets approvals signed, not a number that comes back. For the wider labor cost picture in Thailand, see our analysis of responding to rising labor costs in Thailand, and for worked examples from other processes, see labor saving case studies and their return on investment.

How far BOI production efficiency incentives reduce the effective cost

Thailand offers investment promotion for spending that improves the production efficiency of an existing factory. Automation equipment can fall within that measure.

On the numbers alone this looks decisive. Use the 50% cap in full and the effective cost drops to 5,530,000, which against an annual benefit of 1,738,800 gives a payback of 3.2 years. Reach the 100% cap and the effective cost is 1,135,000, which is nothing more than the five-year maintenance bill.

There is one condition that has to be understood before any of that goes into an approval paper. What is exempted is tax, not a grant. No cash arrives. Corporate income tax that you would otherwise have paid is simply not paid. It follows that if there is no corporate income tax to pay, there is no benefit to collect.

At a corporate income tax rate of 20%, using the 4,395,000 cap in full requires 21,975,000 of taxable income over three years, which is 7,325,000 a year. Using the full 8,790,000 cap requires 43,950,000 over three years, or 14,650,000 a year.

A site running on thin margins, or with a loss year in the period, will not reach the cap. Before anyone writes “effectively half price” into an approval document on the strength of the headline cap, check the site’s taxable income for the last three years. If there is no realistic prospect of 7,325,000 in taxable income each year, the effective cost is not going to be 5,530,000. It is also worth noting that import duty exemption on machinery is granted separately and under different conditions, so a site that cannot use the income tax exemption may still benefit there. Eligibility and category differ project by project, so confirm both with BOI before the numbers are committed to a business case.

Where to draw the line between building in-house and outsourcing

The question of whether the in-house maintenance team could simply build the machine comes up in almost every project. It deserves a numerical answer rather than a cultural one.

5,060,000 is 58% of the outsourced 8,790,000, an apparent saving of 3,730,000. On that figure alone, building it yourself looks like the obvious answer.

But what stops an in-house build is not the mechanical work. It is control, safety, and inspection. Welding a frame, assembling a conveyor, and mounting pneumatic cylinders are a natural extension of what a maintenance team already does well. The wall comes later. Multi-axis servo synchronisation, robot teaching, the lighting and threshold work behind vision inspection, and the design and validation of a safety circuit are four areas in which day-to-day maintenance work simply does not accumulate experience. A team can be excellent at keeping equipment running and still have never once designed a safety circuit from scratch, because nothing in their normal year requires it.

There are also three things that the 5,060,000 does not include.

The first is that two maintenance engineers are away from their real job for eight months. The labor cost of 16 person-months at 55,000, giving 880,000, is in the table. What is not in the table is the loss caused by slower response to breakdowns on existing equipment across those eight months. At a site where line stoppages are frequent, that opportunity cost eats the 3,730,000 difference without much difficulty, and it does so invisibly because nobody books it anywhere.

The second is that drawings, safety documentation, and inspection judgement criteria do not end up in the company. They exist in the head of whoever built the machine. With an outsourced build, those documents are defined as deliverables and priced explicitly, at 150,000 and 90,000 in layer 3 and 90,000 in layer 4. In an in-house build, the time to write them is not booked against anyone, so in practice they never get written. The day that engineer transfers to another site, the machine becomes something nobody dares to touch.

The third is that there is nobody outside to call when it breaks. An external service engineer cannot support a machine they have never seen, built to drawings that do not exist.

The realistic dividing line is this. Mechanical structure and transfer suit in-house work. Control, safety, and inspection suit outsourcing. Rather than building or buying the whole machine, there are genuine cases where the best value comes from fabricating the frame, base, and transfer section internally while outsourcing the control panel, the programs, the inspection, and the safety system. That split only works on one condition. The interface has to be documented first, meaning who is responsible for what, and the exact specification of every mechanical and electrical connection point between the two halves. Split the work without settling that, and commissioning becomes two parties waiting for each other while the production start date passes.

Building in Thailand or building in Japan

A Japanese-owned plant in Thailand usually has a genuine choice between ordering the machine from a machine builder in Japan and importing it, or ordering from a builder inside Thailand.

Building in Japan has real advantages. You get access to the established methods of a builder with a long track record in that specific process, and the parent company’s engineering department in Japan can review the specification easily and in its own language. The disadvantage is that distance converts directly into both cost and time. Look again at the layer 4 breakdown. Disassembly, transport, and delivery is 180,000, and site installation with piping and wiring connection is 220,000. Longer transport pushes both of those up. More significantly, every problem that surfaces after production starts becomes a problem that requires someone to get on a plane.

Building in Thailand has the advantage of proximity after commissioning, and the ability to produce training and documentation in Thai. The 120,000 for operation and changeover training in layer 4 only creates value when it is delivered in the language the operators actually work in. The same applies to the layer 5 items. Annual inspection at 60,000 per year and re-tuning of inspection conditions at 80,000 per year are far more likely to actually happen when the company responsible is an hour away rather than a flight away. Whether the 1,135,000 of five-year maintenance actually gets performed is largely determined by distance.

Here is a usable rule of thumb. A machine whose construction is standardised and which will rarely be touched after commissioning can reasonably be built in Japan. A machine with heavy process-specific engineering, whose judgement conditions have to be adjusted as material lots vary, is usually cheaper in total when built in Thailand. The model project in this article is firmly in the second category. A machine that needs 400,000 of inspection re-tuning over five years cannot be supported by a structure that requires international travel every time the threshold drifts.

Thirty-two weeks from enquiry to production

Here is a realistic schedule. The total is 32 weeks, roughly 7.4 months.

The most important thing in this table is that design takes 8 weeks and fabrication and assembly takes 8 weeks. They are the same length. The lead time on a purpose-built machine is not driven by how long it takes to cut metal. Deciding takes exactly as long as making.

Do not try to compress the 4 weeks of specification freeze. Squeeze it to 2 and the items that were never settled get carried into the design phase, where 8 weeks of design becomes 10. Worse, when the specification changes mid-design, parts that have already been ordered become unusable and the 8 weeks of fabrication stretches too. Every week saved on specification freeze comes back as two to three weeks of delay downstream. This is the single most reliable pattern in custom machine projects, and it is also the one most consistently ignored under schedule pressure.

If the machine is going in to support the launch of a new product, count 32 weeks back from the production start date. But note that these 32 weeks are measured from the enquiry. They do not include the internal budget request and approval process, and they do not include the time needed to evaluate and file a BOI application. In practice, starting the conversation 10 to 12 months before the production start date is the realistic figure.

Frequently asked questions

What is the difference between an automation machine, a special purpose machine, and labor saving equipment

These are differences of terminology rather than firm categories. Broadly, any equipment that replaces human work is called an automation machine. Within that, equipment designed for one specific process and no other is usually called a special purpose machine or a bespoke automation machine, while equipment that mechanises only part of a task to reduce the physical burden on the operator is usually called labor saving equipment.

What matters in practice is not the name but how much of the five layers the equipment includes. Labor saving equipment is not cheaper because of what it is called. It is cheaper because it does not include feeding, vision inspection, or a full safety case. When you read a quotation, ignore the category name on the cover page and check the scope across layers 1 to 5.

How long does it take to design and build an automation machine

For the model project in this article, 32 weeks, roughly 7.4 months from enquiry to production. That breaks down into 2 weeks for concept and budgetary quotation, 4 weeks for specification freeze, 8 weeks for design, 8 weeks for fabrication and assembly, 4 weeks for in-house debugging and witness testing, 2 weeks for transport and delivery, and 4 weeks for site installation through to the production trial.

That figure excludes your own internal budget request and approval process. If the production start date is already fixed, begin the process 10 to 12 months ahead of it. And note that cutting the 4 weeks of specification freeze does not actually shorten anything, because it returns as 2 to 3 weeks of delay further downstream.

What information should be handed over first when requesting a quotation

Seven items. Part conditions, meaning drawings, physical samples, and measured dimensional variation. Cycle time and operating conditions, meaning the required cycle time of 7.5 seconds, 3,750 operating hours per year, and annual output of 1,800,000 pieces. The feeding assumption. Inspection criteria, meaning the current escape rate of 0.35% and the target of 0.08%. Safety requirements. Commissioning and handover documents. Maintenance assumptions.

The ones that make the biggest difference are the items you can express as numbers. Cycle time, operating hours, annual output, and defect rate. Handing over those four as figures is enough to align every supplier on the same assumptions and make the quotations genuinely comparable. Without them, you cannot tell whether a price difference reflects scope or capability, and you will probably guess wrong.

Can a smaller factory make an automation machine pay back

Yes, under conditions. In the model used here, labor saving alone delivers only 864,000 per year, or 4,320,000 over five years, which is 43.5% of the five-year total of 9,925,000. What makes the case work is adding the reduction in escaped defects at 874,800 per year, giving 1,738,800 per year and a simple payback of 5.7 years.

So the deciding factor is not the size of the factory. It is how much you are currently paying for escaped defects and rework. A process with a large figure there will pay back even at modest volume. A process where quality is already stable and labor saving is the only available benefit will struggle to pay back even at high volume. Where BOI production efficiency incentives can be used, the effective cost can fall to 5,530,000 with a payback of 3.2 years, but reaching the 4,395,000 cap requires 21,975,000 of taxable income over three years, so a site with thin profits will not get there.

Can an automation machine be added to existing equipment later

It can, with three things to check. The first is safety. Adding equipment inside an existing guarded area can invalidate the existing risk assessment. You also need to confirm that pressing the emergency stop brings the new machine and the existing line to a coordinated halt.

The second is cycle time alignment. Installing equipment faster than the processes either side of it does not change the cycle time of the line. Measure where the bottleneck actually is before choosing what to automate, because automating a non-bottleneck station produces a very well documented zero.

The third is the access route for delivery. Placing a machine inside an existing line usually means disassembling it, carrying it in, and rebuilding it on site. The 180,000 for disassembly and transport and the 220,000 for site installation in layer 4 both rise with difficult access. Tell the builder about ceiling height, aisle width, and crane availability at the request-for-quotation stage rather than on delivery day.

Summary

A purpose-built machine is not a purchase of hardware. It is a purchase of the work required to pin down a specification.

That is why the visible machine accounts for only 3,200,000, or 36.4%, of the 8,790,000 initial cost. The largest layer is control and electrical at 3,650,000, or 41.5%. Safety is 730,000, or 8.3%. Commissioning is 1,210,000, or 13.8%. Maintenance adds 1,135,000 over five years, giving a five-year total of 9,925,000.

A quotation gap between 5,210,000 and 8,790,000, a factor of 1.7, is a difference of scope and not a discount. Whatever is excluded stays with you. Feeding equipment at 480,000, vision inspection at 740,000, inspection tuning at 420,000, safety at 730,000, and commissioning at 1,210,000 are the five items to check, one boundary question at a time.

On benefits, reducing headcount from 6 to 2 saves 864,000 per year, or 4,320,000 over five years. That is only 43.5% of the five-year total and does not repay the investment on its own. What closes the gap is the reduction in escaped defects at 874,800 per year, giving 1,738,800 per year and a simple payback of 5.7 years. The capacity gain from 9.2 seconds to 7.5 seconds is deliberately excluded, because output with no buyer is not cash.

BOI production efficiency incentives can bring the effective cost down to 5,530,000, a payback of 3.2 years, but what is exempted is tax, not a grant. Using the 4,395,000 cap requires 21,975,000 of taxable income over three years. An in-house build looks like 5,060,000, or 58% of the outsourced price, but that figure excludes eight months of two engineers being away from their real work, and it excludes the fact that drawings, safety documents, and judgement criteria never end up in the company at all.

TOMAS TECH is happy to help with nothing more than the scoping stage. Which of the five layers the quotation on your desk actually covers, how to put a defensible number on what escaped defects currently cost you, and where to draw the line between in-house work and outsourcing. If you can share those three things, we can hand back the material to judge by, well before anything goes near an approval paper. You do not need to have decided to invest, and an early-stage conversation is perfectly welcome, so please get in touch through our contact form whenever it is useful.

References

• Thailand Board of Investment “Measure for Improvement of Production Efficiency” https://www.boi.go.th/upload/menu/production_increase_en_5a5712e3861a0.pdf
• Thailand Board of Investment “Investment Promotion Policy for Automation and Robotics Industries in Thailand” https://www.boi.go.th/upload/content/BOI%20Thailand%20Investment%20Promotion%20Policy%20for%20Automation%20Robotics%20Industries%20Japanese_619b411183526.pdf
• JETRO “Incentives for foreign investment in Thailand” https://www.jetro.go.jp/world/asia/th/invest_03.html
• International Federation of Robotics “Global Robot Demand in Factories Doubles Over 10 Years” (25 September 2025) https://ifr.org/ifr-press-releases/news/global-robot-demand-in-factories-doubles-over-10-years
• Thai Law Online “Minimum Wage in Thailand” https://www.thailawonline.com/minimum-wage-in-thailand/
• THAIBIZ “Upgrading Thai manufacturing through factory automation” https://th-biz.com/features_202102/
• Maruka Machinery (Thailand) “Automation and labor saving machinery” https://www.smri.asia/jp/maruka/products/3096