Conveyor Design in Thailand — Merge, Divert, Buffer and Real Payback

“We installed the conveyor, but the floor is not as much easier as we expected.” We have heard that sentence again and again at Japanese-owned plants in Thailand. The cause is never the motor speed or the belt material. Flow stops where two lines come together, where one line splits toward different destinations, and in the stretches that should have been holding product back but were never designed to. The first thing you decide in conveyor design is not transfer speed. It is flow design, which means deciding where you merge, where you divert, and where you accumulate.

There is a second wall you will hit in Thailand before you place the order. If you try to justify a conveyor on labour savings alone, the payback runs into double digits. This article uses a 25 m inter-process transfer as the worked example. We break the 1,050,000 baht initial investment down in baht, and we open up every step of the arithmetic behind two answers: 16.1 years when only labour is counted, and 4.9 years when the line time lost waiting for parts is counted as well. When you have read it, rerun the same calculation with your own transfer distance and transfer frequency. The decision axis is not “fast or slow.” It is “have you designed the places where flow stops.”

Conveyor design does not start with speed

The first numbers on a request for quotation are usually transfer speed and transfer distance. “25 m, 20 m per minute,” or something similar. Those two numbers alone are not enough to design a conveyor, because far too much is still undecided.

Transfer speed only becomes meaningful once you know how many pieces flow per unit of time. And pieces per unit of time is set by how often the upstream process releases a piece. If upstream releases one piece every 20 seconds, running the conveyor at 20 m per minute simply means the conveyor carries widely spaced work with gaps between every piece. The reverse case is worse. If upstream discharges 5 pieces at once in a burst, an average of 20 m per minute guarantees nothing. The moment arrivals bunch up, you get congestion at the downstream entry point.

In other words, the real substance of conveyor design is not speed design but flow rate and queue design. Transfer inside a factory stops in almost exactly three kinds of location.

• Merge points — where two or more flows become one. Overlapping arrival timing means collision.
• Divert points — where one flow splits into two or more. You need destination logic, and you need an escape route for when the destination backs up.
• Accumulation zones — where the takt difference between upstream and downstream is absorbed. If this zone is short, a momentary downstream stop propagates all the way upstream.

Order a conveyor by comparing equipment specifications alone, with these three left undesigned, and you get a conveyor that flows beautifully on the day it is commissioned and stops as soon as real production mixes several product variants. The awkward part is that this failure cannot be fixed by adjustment after installation. Merge angle, buffer length and the space available for the divert mechanism are all locked into the layout of the support structure.

One more point specific to Thailand, stated up front. If you put transfer automation into an approval request as a labour cost reduction, the payback will almost certainly run into double digits. The reason is simple. The Thai minimum wage is 337 to 400 baht per day as of 2026, with Bangkok, Chonburi and Phuket among the provinces at 400 baht, so the absolute amount of labour cost you can remove is small. We work through the arithmetic later, but even if you automate the entire 25 m transfer, labour savings alone give a payback of 16.1 years.

Does that mean automation is pointless? No. What creates the payback is not the transfer work itself. It is the line time lost because material does not arrive in time. In most plants this time is never measured. Transfer is classified as indirect work, line stoppage is classified as waiting for parts, so two losses with the same root cause are recorded on two separate forms. The second half of this article is about connecting those two.

Choosing the transfer method — conveyor, AGV, cart or manual carry

By the time a conveyor is being discussed, quite a few projects have already assumed the answer is a conveyor. But there are four ways to move material between processes, and each has clearly different conditions where it wins. Get this wrong and no amount of good conveyor design will make the investment pay back.

There are four decision axes. Stability of transfer volume, frequency of layout change, how much the route is shared, and level changes or floors.

Figure 1 maps the four transfer methods onto two axes, frequency of layout change on the horizontal and transport volume on the vertical. The conveyor wins only in the upper left, meaning transport volume is high and the layout will not be moved for the foreseeable future.

Is the transfer volume stable

A conveyor is fixed equipment. It assumes that a decided volume keeps flowing along the route you laid down. If a process runs 120 transfers a day, every day, a conveyor is strong. But if a process runs 200 transfers a day in peak season and 15 transfers a day in the slow season, then 1,050,000 baht of equipment sits idle in the slow season.

As a rule of thumb, a process whose transfer frequency varies by more than plus or minus 40% over the year is a poor fit for a stand-alone fixed conveyor. In that situation, combining a conveyor with AGVs or carts and limiting the conveyor to the trunk section that always flows gives better return on investment. For a cost comparison against AGVs, see our breakdown of AGV price and implementation cost, which sets the figures out by fleet size.

How many years will the layout stay fixed

This is the axis most often overlooked at Japanese-owned plants in Thailand. One commonly cited conveyor failure is that layout changes became impossible. Once you have installed 25 m of roller conveyor and its support structure, that 25 m straight line is effectively part of the building.

At Thai sites it is not unusual for the parent company in Japan to switch the production items every two to three years. Adding one more line for a new model, or rebuilding cells to shift to EV components, immediately turns a fixed conveyor into an obstacle.

Ask “can we state with confidence that this layout will not move for three years?” and get the answer from the plant manager, not from production engineering. If nobody can say it with confidence, either break the conveyor into short sections, for example splitting the 25 m into five modules of 5 m each so that it can be rearranged module by module, or lean toward an AGV. If you are starting from a review of the layout itself, our guide to factory layout improvement sets out how to organise material flow paths.

How many processes share the route

Carts and manual carry can share a route. The same aisle carries carts bound for process A and carts bound for process B. AGVs are the same, with one unit able to cover several routes. A conveyor cannot do this. A conveyor route is dedicated to that route.

So the more a process has only one type of flow on one route, the better the conveyor performs, and in a plant where one aisle is shared by five processes, conveyor investment efficiency drops. For the wider question of how to consolidate plant-wide material flow, see in-plant logistics improvement.

Are there level changes or multiple floors

If the floor is flat and everything is on one level, a cart will do the job. Conversely, transfer that crosses between floors, or climbs a height difference of more than 1 m, sends manual labour hours through the roof and forces an AGV to be integrated with a vertical conveyor. Vertical conveyors, screw conveyors and bucket conveyors exist precisely to solve this level-change problem. Transfer with a level change is the condition where the conveyor’s relative advantage is largest.

Comparing the four methods

What matters in this table is the relationship between the last two rows. A conveyor has a large initial investment and a small variable cost, a cart has zero initial investment and a large variable cost. The break-even between them is therefore set by transfer frequency multiplied by unit rate multiplied by years. Because the unit rate, meaning labour cost, is low in Thailand, the break-even point sits far further to the right than in Japan. Our impression from the projects we have been involved in is that an investment which pays back in a few years in Japan often takes a decade or more in Thailand for the same transfer volume. That is the structural reason behind the remark that conveyors are hard to justify in Thailand.

The five layers of conveyor design — load, capacity, environment, layout and interface

Once the method is settled on a conveyor, the next step is to pin down the specification. The usual selection sequence is described as five steps covering load, capacity, environment, layout and purpose. In this article we replace that final step, purpose, with interface. The reason is that inter-process transfer projects do not fail because the purpose was vague. They fail because nobody decided how each end of the conveyor connects to the existing equipment.

Layer 1 — the load

Six things to decide: dimensions, mass, shape of the underside, centre of gravity, temperature and any material carried on the surface.

The most accident-prone item is the shape of the underside. A roller conveyor assumes that the underside of the load is flat and rigid. A carton with a sagging bottom, a pallet whose stringers drop into the roller pitch, a plastic tray with feet moulded into the base. Under these conditions you either tighten the roller pitch or switch to a belt conveyor. As a design principle, confirm that at least three rollers are in contact within the length of the load along the direction of travel.

The centre of gravity matters too. Tall work or work with an offset centre of gravity tips over during acceleration, deceleration and diverting. Lower the deceleration rate at the divert and it stops tipping, but throughput drops by the same amount. Unless this relationship is quantified at the design stage, it turns after commissioning into the compromise of “we run it slower,” and the original capacity calculation collapses.

What happens if you skip this. The roller pitch does not match, work drops between rollers and jams become frequent. Converting to a belt means starting again from the support structure, and as a rough guide from projects we have been involved in, that adds cost on the order of 30% to 40% of the equipment price.

Layer 2 — capacity

Four things to decide: peak piece count, spacing between pieces, transfer speed and the total mass carried at one time.

The essential point here is to design for the peak, not the average. Even a process rated at 300 pieces per hour may in reality release 80 pieces in one 10-minute window and 20 pieces in the next 10 minutes. A conveyor designed for the average of 5 pieces per minute breaks down the instant 8 pieces per minute arrive. Measure the discharge pattern of the upstream process for at least a single day. This is the cheapest and highest-return design input available, and it costs nothing.

The calculation method for rubber belt conveyors is specified in JIS B 8805, and rollers for belt conveyors are covered by JIS B 8803. Whether each vendor’s calculation basis conforms to these standards is worth confirming at quotation stage.

What happens if you skip this. Motor capacity is insufficient and the conveyor will not start under full load. Or the upstream buffer overflows at peak and you end up stopping the upstream process.

Layer 3 — environment

Things to decide: temperature, humidity, dust, oil, whether washdown occurs, and whether explosion-proof rating is required.

In Thai plants, pay attention to transfer that crosses between an air-conditioned building and one without air conditioning. Where relative humidity swings widely between dry season and rainy season, belt stretch and condensation on drive components become problems. Near painting or washing processes, water and chemicals get into the bearings. Stainless steel construction and IP ratings are decided in this layer.

Noise is part of the environment too. One cited conveyor failure is that noise made the working environment worse. A section where metal pallets run over metal rollers is louder than you would imagine. Put a metal roller merge point right beside an inspection station where operators are permanently stationed, and you will get complaints after commissioning without fail.

What happens if you skip this. Early failure of drive components, or retrofitted acoustic covers as a noise countermeasure at an additional cost in the tens of thousands of baht.

Layer 4 — layout

Things to decide: route, height, curves, gradient, aisle crossings and maintenance access space.

Aisle crossings cause the most argument. Where 25 m of conveyor crosses a walkway, you need to go over it with a bridge, go under it by raising the conveyor, or break it with a hinged section. All three add cost. And if you choose to break it, that point becomes a permanent stopping point.

Maintenance space is easy to forget. A conveyor set only 300 mm from a wall cannot be accessed from the side opposite the drive. If replacing a single roller requires stopping the whole conveyor, maintenance cost 5 years later goes up sharply.

What happens if you skip this. Interference with fire protection equipment or emergency egress routes, forcing rework after installation. In Thailand, confirmation from the industrial estate management office or the fire authority may be required, and the time lost to rework costs more than the money.

Layer 5 — interface

Things to decide: the handover method from upstream equipment, the handover method to downstream equipment, signal exchange, the scope of emergency stop interlocking, and the procedure for manual intervention.

Projects where the conveyor itself runs but the overall flow does not are almost always caused by this layer. The upstream moulding machine does not output a completion signal. The downstream inspection unit does not return an accept or reject status. An emergency stop on one side does not stop the other. This is controls engineering territory rather than mechanical, and it is frequently excluded from the conveyor equipment quotation.

What happens if you skip this. Additional interlock development is required and commissioning slips by one to two months. Whether this layer is written down explicitly at quotation stage has the single largest influence on schedule risk for the project.

Summary of the five layers

The main conveyor types fall into four broad groups: belt, roller, chain and special types such as screw, bucket and vertical conveyors. However, most of the type selection follows automatically once the five layers above are filled in. Do not start from the type, start from the five layers. That is the sequence that minimises rework.

The three places where flow stops — merge, divert and accumulate

This is the heart of the article. It is actually rare for a factory to stop because conveyor throughput was insufficient. What stops is the merge point, the divert point and the accumulation zone.

Figure 2 is a schematic of a typical inter-process transfer in which two upstream lines merge and the downstream flow splits in two directions. The reason accumulation is needed in two places, ahead of the merge and ahead of the divert, is explained below.

Merging — what happens when arrivals overlap

Where two lines become one, the arrival timing on each upstream branch is independent. If line A releases a piece every 20 seconds and line B every 25 seconds, the totals are 180 pieces per hour and 144 pieces per hour, for 324 in total. If the conveyor after the merge can carry 400 pieces per hour, the average leaves headroom.

But there are moments when arrivals overlap. If a piece from line B arrives 0.5 seconds after a piece from line A has reached the merge point, they collide physically. There are three ways to prevent this.

1. Alternating release control — put sensors and zone control ahead of the merge point, hold one side and let the other through
2. Gap creation by speed difference — raise the speed of the section just before the merge to physically widen the spacing between pieces
3. Shallower merge angle — as a general rule of thumb, bring the flows together at 30 degrees or less so that even on contact the guides align them

Whichever method you use, there must be a section ahead of the merge point where product can be held. If that section is only one piece long, stopping one side immediately propagates the stop to its upstream process. The buffer ahead of the merge should be long enough to absorb at least the amount of time the upstream process cannot tolerate being stopped.

Diverting — where product goes when the destination is blocked

A divert point requires two decisions. “Where does this piece go” and “is the destination in a state to receive it.” Design only the first and forget the second, and the moment the destination fills up a piece stalls at the divert and everything behind it stops.

The countermeasure is securing a reject line. Provide a short section that takes pieces off the main line and holds them temporarily when the destination is full or when the destination could not be read. Whether that escape section exists is what decides whether a divert-point problem becomes a full line stop or a single piece of rework.

Accumulation — what determines buffer length

An accumulation zone absorbs the takt difference between upstream and downstream. The required length is given by the following.

Required buffer length = (upstream processing rate – downstream processing rate) x (duration of the difference) x occupied length per piece

Here is a worked example using illustrative figures to show the approach. Upstream runs one piece every 20 seconds, or 180 pieces per hour, downstream one piece every 24 seconds, or 150 pieces per hour, a difference of 30 pieces per hour. If that difference persists for 30 minutes, accumulation is 15 pieces. At an occupied length of 0.4 m per piece, the required buffer length is 6 m.

Skip that calculation, allow “2 m for now,” and a 10-minute downstream stop propagates the stop upstream. Conversely, if the actual downstream stoppage averages 3 minutes per occurrence, you may not need 6 m at all. Buffer length is calculated backwards from actual downstream stoppage data, not guessed.

There are two accumulation approaches: conventional accumulation, in which each piece pushes the one in front, and zero-pressure accumulation, which applies no contact force. If the product deforms easily, or if surface scratches make it defective, you need zero-pressure accumulation with zones that stop independently. Naturally, sensors and controls increase with the number of zones, so the cost goes up.

Belt tracking and jams, the chronic losses

There is one more loss that rarely shows up in the numbers. Belt mistracking and jams.

Mistracking is caused by the parallelism of the tail pulley, belt tension and uneven placement of the load. Keep running with the belt off-track and the belt edge wears until it has to be replaced. Jams concentrate at merge points and divert points, and even if each one takes only a few minutes to clear, 10 a day adds up to time you cannot ignore over a year.

The important part is that neither of these is recorded as a failure. Operators fix them on the spot, so they appear in neither the maintenance record nor the downtime record. As a result, when the investment is being evaluated, this loss never enters the cost-benefit calculation. When you assess the current state before implementation, we recommend asking operators to record the number of times they cleared a jam for just one week.

Where sorting automation starts to pay

After inter-process transfer, the next most common enquiry is sorting. The request is to automate the process of separating product by ship-to destination, by model, or into good and defective.

How to think about sorter types

Sorting mechanisms can be classified by how they act on the product.

On throughput, the published figures for each type vary enormously with conditions. Even within sliding shoe sorters, effective throughput can differ by nearly a factor of two depending on product length, destination pitch and required accuracy, so we will not assert specific numbers here. As a general guide, pusher and pop-up types are described as operating in the range of tens of pieces per minute, with sliding shoe and cross-belt types an order of magnitude higher, but always obtain the real machine throughput from the vendor by presenting your own product conditions. Catalogue values are best-case values.

The number of destinations drives the cost

What is easily missed is that cost is driven not by throughput but by the number of destinations. Every additional destination adds one divert mechanism, one chute, one full detection device and one control circuit. Sorting to 3 destinations and sorting to 12 destinations can differ by several times in total cost even with an identical main structure.

So the first design question is not “how many pieces per hour” but “how many destinations do we genuinely need.” Even when it looks like 12, if the top 3 destinations account for 80% of the volume, automating only those 3 and keeping the rest manual gives better investment efficiency. Build a Pareto chart of actual volume by destination before you start designing.

How to handle missorts

Automated sorting will always produce missorts. Barcode read failures, label orientation, overlapping product. What you must decide at design stage is not “how to reduce missorts to zero” but what happens when a missort occurs.

• Divert unread product to a reject line and reintroduce it manually
• Where downstream do you detect product that went to the wrong destination
• If it is not detected, at which pre-shipment process does it get stopped

Sorting automation without those three decided will not work in operation even at 99% accuracy, because nobody has worked out how many pieces the remaining 1% is per month, or who handles those pieces and how.

The break-even boundary

Whether sorting automation pays is judged by the following comparison.

Annual hours currently spent on manual sorting x hourly rate against investment divided by expected years of use + annual running cost

At Thai labour cost levels, the cases where this comparison alone justifies automation are limited. Sorting automation investment runs into the same structural problem as conveyors. What creates the payback is usually not the labour hours of sorting, but the cost of returns and reshipment caused by shipping errors and the opportunity loss from shipments delayed while waiting for sorting. Put a value on those two before you compare. If you are looking at case packing and sorting together, our article on the decision axes for case packing robots is also useful.

Cost structure in a Thai plant — breaking the quotation into five layers

Now to the money. The most common mistake when comparing conveyor quotations is comparing only the equipment price. The real total splits into five layers, and the equipment itself is just under half of it.

This article uses the following model case.

• An assembly plant in Thailand, two-shift operation, 300 operating days per year
• A 25 m inter-process transfer, currently done by cart
• The equipment installed is 25 m of roller conveyor plus the drive unit

Figure 3 breaks the 1,050,000 baht initial investment into the five layers. The equipment accounts for 45.7%, while electrical work and the control panel take 20.0% and design, on-site attendance and commissioning take 11.4%, so everything other than the equipment adds up to 54.3%.

The five layers of the initial investment

Here is what drives the price difference in each layer.

Equipment, 480,000 baht — varies with roller diameter, pitch, frame material and drive type. There is a large gap in equipment price between fabrication in Thailand and import from Japan. As discussed later, however, choosing import creates a problem with the lead time for spare parts.

Structure and installation, 150,000 baht — varies with the condition of the floor. Whether the existing anchors can be reused, how far out of level the floor is, and whether night or weekend work is required all change the labour hours. As a rough guide from projects we have been involved in, you should assume that retrofitting onto a running line makes this layer 1.5 to 2 times larger.

Electrical work and control panel, 210,000 baht — the second largest layer after the equipment. This layer swells because of the connection to existing equipment. You have to open the panel of the existing machine to pick up its completion signal. Or the existing machine cannot be modified because of the manufacturer’s warranty, so an external sensor is used instead. From what we have seen on projects we have been involved in, it is not unusual for this layer to grow to around 1.3 times the original quotation. Layer 5, the interface, was emphasised earlier precisely because it is what makes this cost unpredictable.

Safety measures, 90,000 baht — the layer you must not cut. Details follow later.

Design, attendance and commissioning, 120,000 baht — the layer most likely to attract “can the design fee be free” negotiation. Cut it and the design accuracy of all five layers drops, and you end up rebuilding the support structure at several times the cost. In Thailand in particular, reconciling the parent company’s design standards in Japan with local installation practice takes real hours.

What to check when comparing quotations

Lining up the totals from three quotations tells you nothing. Put which of the five layers is included and which is excluded into a single comparison table. Cheap quotations almost always list electrical work and design and commissioning as separately chargeable.

Payback calculation — 16.1 years on labour cost alone

Now the arithmetic. The assumptions are the same model case as the previous section.

Current transfer workload

• Transfer frequency is 120 times per day, both shifts combined
• Each transfer takes 3.5 minutes including the round trip plus loading and unloading
• 120 times x 3.5 minutes = 420 minutes per day = 7.0 hours per day
• At 300 operating days a year, 7.0 hours x 300 days = 2,100 hours per year

Now the operator’s hourly rate. We assume a base monthly figure excluding overtime of 14,000 baht, meaning base pay plus allowances, with no overtime included, and 173 scheduled working hours per month. The key is that the numerator and the denominator are both on an excluding-overtime basis.

• 14,000 baht divided by 173 hours = approximately 81 baht per hour
• Scheduled annual hours, 173 hours x 12 months = 2,076 hours
• Annual labour cost, 14,000 baht x 12 months = 168,000 baht

For reference, the Thai minimum wage as of 2026 is 337 to 400 baht per day, with Bangkok, Chonburi and Phuket among the provinces at 400 baht. Setting the base monthly figure at 14,000 baht assumes a base monthly amount that includes allowances on top of that daily level, and it excludes overtime pay.

Do not use this hourly rate as it stands in a process where overtime on transfer work is routine. In that case, rebuild the rate with total labour cost including overtime in the numerator and total working hours including overtime in the denominator. Adding overtime pay to the numerator while leaving scheduled hours in the denominator overstates the saving. Whichever approach you take, redo the calculation with your own actual figures.

After the conveyor is installed

Movement over the 25 m section is automated and only loading and unloading remains.

• Remaining work per transfer is 1.2 minutes
• 120 times x 1.2 minutes = 144 minutes per day = 2.4 hours per day
• Per year, 2.4 hours x 300 days = 720 hours per year

Now the hours saved.

• 2,100 hours – 720 hours = 1,380 hours per year
• In headcount terms, 1,380 hours divided by 2,076 hours = approximately 0.66 of a person
• In money terms, 1,380 hours x 81 baht = approximately 111,780 baht per year

The hourly rate of 81 baht is 14,000 divided by 173, or 80.92 and so on, rounded. So that readers following along on a calculator get the same figures, this article uses the rounded 81 baht per hour as the rate from here on, and takes 111,780 baht as the base amount for the calculations that follow.

Annual running cost

• Power, 0.75 kW x 16 hours per day x 300 days = 3,600 kWh. 3,600 kWh x 4.2 baht = 15,120 baht per year
• Maintenance, parts and inspection, at 3% of the initial investment = 1,050,000 x 0.03 = 31,500 baht per year
• Total, 15,120 + 31,500 = 46,620 baht per year

Case A — labour cost only

• Annual net benefit = 111,780 – 46,620 = 65,160 baht per year
• Simple payback = 1,050,000 divided by 65,160 = 16.1 years

16.1 years means it does not pay back. Even if the physical life of a conveyor is, as a general guide, somewhere around 15 to 20 years, the production items will change and the layout will change well before that. An approval request grounded only on labour cost reduction should be rejected at this point.

Many proposals try to shrink the number here by understating maintenance cost or increasing the number of operating days, but that only moves figures on paper. It does not change the cash flow on the floor by a single baht.

Case B — a line that also loses time waiting for parts

Now look at the same process from another angle. On this line, stoppages occur because parts do not reach the operator, classified as waiting for supply. Assume that line stoppage from waiting for supply, summed across all causes regardless of what they are, ran at 18 minutes per day across both shifts. The causes include the operator who does the cart transfer being pulled away to help another process so that parts arrive late, but also defects in the previous process, delayed die changes, and parts shortages themselves.

• 18 minutes per day x 300 days = 5,400 minutes = 90 hours per year
• At a line contribution margin of 2,800 baht per hour, 90 hours x 2,800 baht = 252,000 baht per year
• Of that, the portion caused by transfer, meaning the portion that disappears when the conveyor is installed, is 60%, so 252,000 x 0.6 = 151,200 baht per year

Let us be explicit about what the 60% means. It is the share of the 18 minutes, across all causes, that is attributable to transfer. The remaining 40% is due to causes other than transfer, such as defects in the previous process or parts shortages, and installing a conveyor does not remove it. This ratio differs from plant to plant. When you run the calculation for your own site, total up the cause field on your stoppage records for just one month and measure the transfer-related share.

There is also an assumption behind this valuation. We value the lost time at contribution margin on the assumption that the lost operating time cannot be recovered, meaning demand exceeds capacity. In a plant that recovers the time through overtime or weekend working, what is lost is not contribution margin but the actual cost of recovering it. In that case, replace the figure with the overtime premium, utilities and other amounts actually paid to recover those 90 hours. The valuation comes out lower, but it matches your own reality better.

• Total benefit = 111,780 + 151,200 – 46,620 = 216,360 baht per year
• Simple payback = 1,050,000 divided by 216,360 = 4.9 years

Comparing the two cases

Note. In both cases, the 111,780 baht labour reduction is an amount that is realised in cash only as headcount avoidance. Unless you actually remove a person, the labour cost in the books does not fall. Do not book it as a cost reduction. This is covered in detail in the next section.

Same equipment, same investment, same running cost. The only difference is what was counted as a benefit. The roughly 11 years between 16.1 years and 4.9 years comes from a loss that is already occurring every day but that nobody records as an amount of money.

A note on avoiding double counting

The thing to be most careful about in this calculation is double counting the benefit. Let us be explicit.

There is only one baseline, one counterfactual. It is the world in which no conveyor is installed and cart transfer continues. Case A and Case B are both compared against that same single world. Case B does not introduce a different world. It simply counts an additional, different kind of loss occurring in the same world.

On that basis, confirm that the two benefits do not overlap. The 1,380 hours in Case A are people’s hours. They are the hours operators spend on transfer, and when they are saved the operators can be redeployed to other work. The 90 hours in Case B are line hours. They are hours the line stands still because parts have not arrived, and they occur regardless of whether anyone is moving.

The key point is that what Case B monetises is not all 90 hours but only the 60% attributable to transfer, equivalent to 54 hours. That transfer-related stoppage occurs while the transfer operator has been pulled to another process and is therefore not doing transfer. Because it happens during time when transfer is not being performed, it is not included in the 1,380 hours of transfer work counted in Case A. One is the time a person spends on transfer, the other is the time a line spends waiting for parts. The subject of the time is different, so they do not overlap. The remaining 40% has nothing to do with transfer, so it was never counted as a conveyor benefit in the first place.

We should also be explicit about how the 111,780 baht of labour cost is realised in cash. This amount is realised in cash only as headcount avoidance. It frees up 0.66 of a person’s worth of hours, but unless you actually remove one person, the labour cost in the books does not fall by a single baht. So in an approval request, the accurate wording is not “111,780 baht of cost reduction” but “0.66 of a person’s additional headcount will not be needed when production increases.” A proposal that leaves this distinction vague will be sent back by the finance department every time.

Whether there is room to increase production is itself an assumption in the decision. Average capacity utilisation in Thai manufacturing in the second quarter of 2026 was 57.47%, and the MPI for the same quarter was -1.79% year on year. Utilisation sitting below 60% is a fact that indicates spare capacity remains on the equipment side. However, this is an average across all industries, and it does not imply a causal claim that spare capacity means you can increase production. Judge it against your own order outlook.

It is worth checking the short-term movement as well. Thailand’s Manufacturing Production Index, the MPI, was -3.1% year on year in June 2026, and automotive production was -7.55%. Note that the MPI is published by the Office of Industrial Economics, the OIE, whereas automotive production volume is a statistic published by the Federation of Thai Industries, the FTI. They are indicators from different sources, so attribute them separately when quoting them in internal documents. A decision to add fixed equipment while production is contracting should be made cautiously, which is a further reason to take the 16.1 years of Case A as a reason not to proceed.

BOI incentives and what is easy to miss in the 2026 scheme

One way to shorten payback is through incentives from the BOI, the Thailand Board of Investment. However, the 2026 scheme is clearly limited in scope, and because so many approval requests misread this, we will set it out precisely.

What BOI Announcement No. 4/2569 contains

Here are the key points of BOI Announcement No. 4/2569, published in the Royal Gazette on 31 March 2026.

Easy to miss 1 — the scope is limited to automotive and HEV / PHEV

This announcement applies to automotive manufacturing, category 3.6, and HEV / PHEV manufacturing, category 3.8. It is not the case that all manufacturing is eligible. Plants in other sectors such as electrical and electronics, food, plastic moulding or metalworking cannot apply this announcement as it stands. For other sectors you need to check the BOI’s framework for automation and robotics separately and ask your BOI contact whether there is a scheme available under your own business category.

Even a first-tier supplier making automotive components may be outside the scope if its registered business category is not 3.6. Before you build an incentive into an approval request, check the business category stated on your own BOI certificate.

Easy to miss 2 — the 1,000,000 baht threshold is closer than it looks

What follows applies only if your company falls under business category 3.6 or 3.8. On that basis, the minimum investment condition of 1,000,000 baht, excluding land and working capital, is well within range for conveyor projects. The initial investment in this article’s model case is 1,050,000 baht, slightly above the threshold. However, the model case is a general assembly plant and is therefore not in an industry covered by this announcement in the first place. Read it only as an indication that conveyor projects land in the region of the threshold.

The point to watch here is what can be counted toward the investment amount. The equipment alone, at 480,000 baht, does not reach the threshold. Whether the structure and installation, electrical work and control panel, safety measures and design and commissioning layers can be included changes whether an application is viable. That line is drawn case by case, so the reliable approach is to confirm it with the BOI with the quotation already broken into the five layers. That is another practical reason to break the quotation into five layers.

Easy to miss 3 — the 30% locally made condition

This too is a condition that applies only if you fall under business category 3.6 or 3.8. Where 30% or more of the machinery value is tied to automation machinery made in Thailand, the exemption rate rises from 50% to 100%. A conveyor is a type of equipment that satisfies this condition relatively easily. Fabricate the support structure and the conveyor itself in Thailand and the locally made share of the machinery value will be high.

However, proving “made in Thailand” requires documentation. Registration as a domestic manufacturer, origin of the materials and proof of the assembly location are all required. At the stage of selecting a supplier, confirm that the supplier can issue the documentation proving domestic manufacture. Assembling that paperwork retrospectively after installation is difficult.

Easy to miss 4 — the application deadline is the end of 2027

Because the application deadline is the end of 2027, a project starting its concept phase in the second half of 2026 appears to have time in hand. In practice, though, the sequence from freezing the specification through quotation, approval, ordering, fabrication, installation and commissioning takes 6 to 12 months. Include the current-state survey that precedes design and you need to start even earlier. We recommend drawing up a schedule backwards from the deadline while you are still in the concept phase.

Safety and maintenance — entanglement prevention and scheduled inspection

Of the five cost layers, the one you must not cut is the 90,000 baht for safety measures. Here is why.

Entanglement happens in a predictable set of places

The mechanically hazardous points on a conveyor are almost always the same ones.

• The nip between the drive pulley and the belt — where the most serious accidents occur
• The gap between rollers and the frame — gloves and sleeves get drawn in
• The mesh between chain and sprocket — the guard tends to come off
• Moving parts at merge and divert points — hands go in here frequently when clearing jams
• The area where product can fall — under any section routed overhead

Of these, the most common circumstance in actual accidents is clearing a jam. Reaching into a running conveyor to clear a blockage is routine behaviour in every factory. So the essential point of safety design is not only fitting guards but building a structure that removes any need to reach in while the conveyor is running. If you know where jams occur, design in a mechanism at that location that can be released without putting a hand in.

Under Japan’s Industrial Safety and Health Regulations, fitting entanglement prevention devices and carrying out scheduled inspection are obligations of the employer. Different legislation applies at plants in Thailand, so the requirements under the relevant Ministry of Labour regulations must be confirmed locally on a case-by-case basis. As a practical matter, however, designing to meet the parent company’s equipment safety standards in Japan errs on the safe side, including for later audits.

Design the scope of the emergency stop

An emergency stop button is not something that merely has to be present. What you must decide in design is what stops when it is pressed.

When someone presses the emergency stop in the middle of the 25 m conveyor, does the upstream moulding machine stop or not. If it does not, where does the work it discharges go. If it does, how many minutes does it take to restart the moulding machine. When this design is vague, the floor drifts into the practice of not pressing the emergency stop because restarting is a nuisance. That is the most dangerous state of all.

As a principle, make the scope of the emergency stop match the area the person pressing it can see, and place it somewhere easy to reach. Take 10 m or less as a starting point for spacing as a general rule of thumb, and set the final spacing according to the applicable legislation and the results of your risk assessment.

Inspection items and frequency

Maintenance was set at 3% of the initial investment, or 31,500 baht per year, but that figure only means something if the inspections actually happen. Here are the minimum items.

In Thai plants, producing the inspection sheet in Thai makes a large difference to whether it works in practice. An inspection sheet only in Japanese or English gets ticked as a formality without any substance behind it. When selecting a vendor at implementation, make a Thai-language maintenance manual and the availability of scheduled visits by local engineers part of your conditions.

How to think about spare parts

The longest conveyor stoppages are the ones spent waiting for parts. A single roller, a single belt or a single sensor is missing and the line is down for days. For a conveyor imported from Japan, parts procurement can take two to four weeks.

We recommend holding drive belts, rollers in several pieces of each size, sensors and the main consumables for the motor as spare stock from the time of installation. This cost is frequently not included in the five layers of the initial investment, so confirm it separately at quotation stage.

Pre-order checklist

Here is everything above in a form you can use directly inside your company. Fill in these 11 items before you issue a specification. Take quotations with blanks still in it and additional costs will appear later without fail.

1. Measured transfer frequency — both peak and average. One day of data is enough, but measure it. Do not use estimates
2. Underside shape and centre of gravity of the load — is the base flat, does it have feet, is the centre of gravity offset. Confirm from a drawing or a photograph of the actual item
3. Discharge pattern of the upstream process — even or in batches. Record arrival times for one day
4. Actual downstream stoppage data — average stoppage duration per occurrence and frequency. Used to calculate buffer length backwards
5. Number of merge and divert points — at each one, has an escape route been designed for when it backs up
6. How long the layout is fixed — can the plant manager state with confidence that this layout will not move for three years
7. Aisle crossings and how they are handled — over, under or broken. Has interference with fire protection and egress routes been checked
8. Signal specification with upstream and downstream — completion signal, accept or reject status, scope of emergency stop interlocking. Also confirm whether the existing equipment can be modified
9. Five-layer breakdown of the quotation — are equipment, structure and installation, electrical work and control panel, safety measures, and design and commissioning shown separately
10. Maintainability check — does replacing a single roller require stopping the whole line. Is there enough access width for maintenance
11. BOI business category check — have you checked the business category on your BOI certificate and asked your contact whether any incentive applies

Of this checklist, items 1, 3 and 4 can only be measured by you. And those three have the largest influence on design quality. They are the parts you cannot hand over to the vendor.

FAQ

How much does conveyor design cost?

In this article’s model case, a 25 m section, roller conveyor, two-shift operation, the total initial investment is 1,050,000 baht. The breakdown is equipment 480,000, structure and installation 150,000, electrical work and control panel 210,000, safety measures 90,000, and design, attendance and commissioning 120,000, with the equipment at 45.7% of the total.

This amount varies widely with distance, the load, the number of sorting destinations and the complexity of the connection to existing equipment. Even so, the structure in which everything other than the equipment accounts for more than half the total is common to most projects. A quotation compared on equipment price alone will always grow later.

Which is cheaper, a conveyor or an AGV?

The structure of initial investment and variable cost is different, so the answer reverses depending on the conditions. A conveyor is fixed equipment with a large initial investment and a small variable cost. An AGV costs more as the fleet grows but handles route changes well.

The decision turns not on the absolute cost but on three points. First, will the layout stay fixed for three years. Second, will one route be shared by several processes. Third, does annual transfer volume vary by more than plus or minus 40%. If the first is yes and the second and third are no, a conveyor wins. The reverse favours an AGV. The AGV cost structure is broken down by fleet size in our article on AGV price and implementation cost.

Can it be added to an existing line later?

It can, but as a rough guide from projects we have been involved in, expect structure and installation cost to be 1.5 to 2 times higher. On a running line, anchor drilling and installation work have to be done at night or at weekends, which increases the hours.

The bigger problem is the electrical work and control panel layer. Picking up the completion signal from the upstream equipment means opening the panel of an existing machine, and where the manufacturer’s warranty prevents modification, an external sensor has to be substituted. Projects that have not thought this through at quotation stage see commissioning slip by one to two months. If you are considering a retrofit, confirm the interface specification first.

Is it better to fabricate in Thailand or import from Japan?

There are three decision axes.

The first is lead time for spare parts. With equipment imported from Japan, replacing a single roller can take two to four weeks. Unless you are working on the assumption that spares are stocked at the time of installation, this difference bites during operation.

The second is BOI incentives. Under BOI Announcement No. 4/2569, where 30% or more of the machinery value is tied to automation machinery made in Thailand, the corporate income tax exemption rate rises from 50% to 100%. However, the scope is limited to automotive manufacturing, category 3.6, and HEV / PHEV manufacturing, category 3.8, so checking your own business category comes first.

The third is accuracy requirements. For sections that demand positioning accuracy of the load, or sections with unusual environmental conditions, the priority is choosing a fabricator with a track record. In practice many projects end up with a combination, with the straight trunk section fabricated in Thailand and the divert and interface points that need accuracy imported.

How many units does sorting automation need before it pays?

Framing the question as “how many units” is itself what leads to the wrong decision, because cost is driven by the number of destinations rather than throughput, and payback is not driven by the labour hours of sorting.

At Thai labour cost levels, the cases where the saving in manual sorting hours alone justifies automation are limited. What creates the payback is usually two things, the cost of returns and reshipment caused by shipping errors, and the opportunity loss from shipments delayed while waiting for sorting. Put a value on those two first.

Then build a Pareto chart of actual volume by destination. If the top 3 of 12 destinations account for 80% of the volume, automating only those 3 gives better investment efficiency. Keep the option of not automating everything on the table from the start.

Summary

What stops the flow in conveyor design is not speed but the design of merging, diverting and accumulating. It is rare for a factory to stop because transfer capacity was insufficient. What stops it is the merge point where two flows overlap, the divert point with no escape route when a destination backs up, and the accumulation zone that cannot absorb the takt difference between upstream and downstream. These three cannot be fixed by adjustment after installation, because they are locked into the layout of the support structure.

Then there is the money. Justify a conveyor in Thailand on labour cost reduction alone and the payback is 16.1 years. Saving 1,380 hours a year, roughly 0.66 of a person, gives 111,780 baht. Subtract 46,620 baht of running cost and the net is 65,160 baht. Divide the initial investment of 1,050,000 baht by that and you get 16.1 years. That is a conclusion of “do not proceed.”

What creates the payback is not the transfer itself but the line time lost waiting for supply. Stoppage of 18 minutes a day across all causes is 90 hours a year, and at a contribution margin of 2,800 baht per hour that is 252,000 baht. The 60% share attributable to transfer, meaning the part the conveyor removes, is 151,200 baht. Total benefit is 216,360 baht and the payback becomes 4.9 years. Same equipment, same investment, and the only difference is what was counted.

Two reservations go with that, however. The first is that the 111,780 baht of labour cost is realised in cash only as headcount avoidance. Unless you actually remove a person, the labour cost in the books does not fall, so the accurate wording for an approval request is “approximately 0.66 of a person’s additional headcount will not be needed when production increases.” Do not book it as a cost reduction. The second is that valuing the stoppage time at contribution margin applies only where the lost operating time cannot be recovered. In a plant that recovers it with overtime, replacing the figure with the actual cost of that recovery lowers the valuation. Present 4.9 years without writing both of these down and the finance department will send it back.

So what you should do before placing an order is not to collect three quotations. It is to measure, for just one month, how many minutes your line is stopped for transfer-related reasons. Without that number, no choice of conveyor gives you a basis for an investment decision. With that number, you only have to put your own values into the formulas in this article to produce a payback figure you can put into an approval request.

Transfer design cannot be judged without seeing the actual product and the actual layout. Even at the same 25 m, the required specification and the cost change depending on whether it crosses an aisle, whether there is a level change, and what pattern the upstream process discharges work in. TOMAS TECH is based in Bangkok and has worked on the design of material transfer and production management in the factory IT and FA field for Japanese-owned manufacturers. Still at the concept stage with no drawings, still undecided between a conveyor and an AGV. Conversations at that stage are in fact the ones that remove the most rework later. If you have your current transfer frequency and a layout drawing to hand, we are happy to start by working through the decision axes with you. Get in touch here.