Factory Wireless LAN and Industrial Network Design 2026: 5 Cost Layers

The handheld terminal drops out at the back of the warehouse. The AGV halts at the same corner on every run. The Wi-Fi that felt perfectly fine in the office building becomes untrustworthy the moment you step into the production hall. Factories that struggle with wireless LAN deployment tend to struggle in remarkably similar places, and the cause is almost never transmit power. It is that an office design philosophy was carried into a plant unchanged. This article walks through factory wireless LAN and industrial network design in the order you actually encounter the decisions: the cost broken into five layers with real THB ranges, the OT/IT boundary, AGV roaming requirements, the regulatory position in Thailand and Vietnam, and a 90-day roadmap you can start on Monday.

Why factory wireless LAN fails when you build it like an office network

“Add more Wi-Fi and it will connect” is broadly true in an office and broadly false in a factory. An office floor has plasterboard partitions, a ceiling height of around three metres, a layout that changes once a year at most, and devices that sit in a person’s hand — where a few seconds of interruption is absorbed by the user simply tapping again. In a factory, every one of those assumptions collapses.

It is not distance that weakens the signal — it is metal and liquid

What determines the RF environment in a production hall is not the distance from the access point. It is metal fixtures, steel racking, liquid tanks, and material handling equipment such as conveyors and overhead travelling cranes, which reflect the signal, block it, and move it around over time. Reflections off metal surfaces create multipath, producing a state where the received signal strength at a single fixed point fluctuates from moment to moment. Tanks filled with liquid absorb 2.4 GHz and 5 GHz energy without much resistance.

The awkward part is that none of these are fixed obstacles. A forklift passes and the shadow moves. Pallets stack up and a propagation path closes. The tank contents change with the season and so does the attenuation. This is why a design that consists of drawing overlapping AP coverage circles on a floor plan will always produce dead spots once the plant is live. Site surveys for factories and large retail sites have to account for interference from adjacent bays, the floors above and below, and neighbouring facilities — not just the room you are standing in.

Heat, dust and vibration — the equipment assumptions themselves are different

Standard office-grade APs and switches are designed on the assumption of an air-conditioned room with no particulate load. That assumption does not hold on a shop floor. You need equipment selected to survive high and low temperatures, dust, oil mist and vibration, and repurposing office APs is not a realistic option.

Because this difference shows up as a line item in the initial cost, someone in the approval meeting will always ask whether there is a cheaper AP available. There usually is. But the cost of rebuilding a network that became unstable after six months and lost the confidence of the production floor is far larger than the price difference on the hardware. The more damaging second-order effect is that once you have deployed cheap and lost credibility, the next request — “this time let us do it properly” — struggles to get funded at all.

“It connects” and “it does not drop” are entirely different requirements

What an office Wi-Fi network is asked to deliver is connectivity and speed. What a factory wireless network is asked to deliver is that it does not drop and that its latency is predictable. In design terms these are almost unrelated goals.

Throughput, for example, is essentially a non-issue for barcode scanning on a handheld terminal or for an IoT gateway pushing sensor values, because each device is moving a few kilobytes at a time. Handover time when a device moves between APs, on the other hand, is a make-or-break requirement for an AGV. Nobody notices a gap of several hundred milliseconds on a device held in a human hand; for a vehicle in motion it means a missing control instruction. If you pick equipment while measuring the wrong definition of “fast”, you end up with a network that looks impressive on the datasheet and is unusable on the floor.

Wi-Fi does have one clear advantage for factories that is worth stating plainly: you can add network coverage while leaving existing equipment exactly where it is. Within the scope that requires no cabling work, you can add a communication path without stopping the line. Wired networks cannot do that, and it is precisely why wireless gets chosen in plants — on the condition that the design is right.

The shape of industrial network design — wired, wireless and the OT/IT boundary

A request that starts as “we need a factory wireless LAN” almost always turns into an industrial network design project. It is very rare that the wireless portion can be carved out and ordered on its own. Holding three layers in your head makes it possible to read a quotation and understand what is and is not inside it.

Layer 1: the wired foundation — cabling and switches

Wireless quality can never exceed the quality of the wired network behind it. An AP is a radio, but it is also the last hop of a wired network. The fibre backbone between buildings, the Cat6A horizontal cabling to each AP, the industrial switches mounted inside field cabinets, the distribution switches that aggregate them, and the redundant core that is not allowed to go down. If this stays weak and you only refresh the APs, the symptoms do not disappear.

A significant share of complaints described as “the wireless is unstable” turn out to be wired-side problems: saturated switch ports, loops, insufficient PoE budget, ancient firmware. That is why the right first move — before any RF investigation — is an inventory of the wired side.

Layer 2: wireless access — survey, APs, controller

The wireless layer cannot be assembled out of hardware alone. A pre-deployment site survey (RF measurement and identification of interference sources), the APs themselves, the controller that manages them centrally, and a post-deployment survey after installation must be treated as one package. When the survey is stripped out as an “optional” item, the rework that follows costs several times what was saved.

Layer 3: the OT/IT boundary — segmentation and security

The third layer is the boundary between the production equipment side (OT) and the information systems side (IT). Connecting the network means putting control systems that were previously isolated by physics onto the same playing field as the corporate LAN and the cloud. Boundary firewalls, zone design and device authentication are not things you bolt on later; they belong in the first design. This layer sits directly alongside OT security design for factories, and the two cannot sensibly be planned apart.

These three layers map broadly onto the cost structure below (layer 1 splits into cabling L1 and wired hardware L2, and operation L5 is added on top). When you receive a wireless-only quotation and think it looks cheaper than expected, the other two layers are usually sitting outside the document.

Breaking factory wireless LAN cost into five layers

Now the numbers. What follows is TOMAS TECH’s own estimate. It is not based on public statistics or third-party research. It is a modelled figure for a model factory, assuming procurement inside Thailand in 2026. Real figures move with the number of buildings, ceiling heights, explosion-proof or dust-proof requirements, and the condition of the existing installation.

Model factory assumptions

The AP count is 32. That single number drives both the horizontal cabling runs in L1 and the AP procurement quantity in L3. When you read a quotation, check that the number of horizontal cable runs matches the number of APs. If they diverge, you can tell at a glance that the quotation was not derived from an actual design.

L1 to L4 — the initial cost breakdown

Initial cost splits into four layers. Adding L5 — operations, which recurs annually — gives you five layers to manage, which is the practical unit of control.

The initial total is approximately THB 2.6 million. Depending on scale, number of buildings and explosion-proof or dust-proof requirements, projects of this type most often land somewhere between THB 2.0 million and THB 3.3 million. Carrying that range in your head makes budget requests noticeably more accurate.

What deserves attention in this table is the ratio. The part that is genuinely “wireless” — L3 — is 994,000 THB, just under 40 percent of the initial cost. The remaining 60 percent and change is cabling, switching and boundary security. A quotation that answers “what does a factory wireless LAN cost?” with an AP unit price is a quotation that has not looked at that 60 percent.

The other line that tends to raise eyebrows is L4 at 700,000 THB. Seven hundred thousand baht for a firewall, zone design and authentication — roughly a quarter of the initial cost. The instinct to trim it is entirely human. But as explained further down, L4 is the one layer that is hardest to add after the fact.

L5 operations — the layer measured annually

Including an annual survey in operations is the point of this layer. Factory layouts change. Reconfigure a line and the RF environment changes with it, and the original design drifts a little further from reality each year. Without a routine of measuring and correcting once a year, the same conversation — “it has become unstable again” — comes back in three years’ time. Spare units follow the same logic: lead times for industrial hardware are hard to predict, so how many cold-standby units to hold is a decision that belongs in the design phase, not in the incident.

Comparison with Japanese figures — do not transplant them

For reference, here are typical network construction costs in Japan. These are Japanese domestic reference figures, not Thai procurement prices. Labour rates, import duties and the number of capable vendors all differ, so they cannot be used as a basis for a Thai budget.

Looking at those Japanese domestic reference figures, there is a data point of JPY 4,500,000 for 2,000 m². It is tempting to scale that by floor area up to a 10,000 m² model factory, but the proportion does not hold. Building structure, required availability and the presence or absence of OT/IT segmentation can move the unit cost by a multiple. Cross-border comparison is safe only as a sanity check on the order of magnitude, never as a pricing basis.

Five-year TCO and the reality that the network alone does not pay back

This is the part of the article that matters most. Most factory network refresh projects stall not for technical reasons but at the investment justification.

The five-year TCO is about THB 4.1 million

Take the initial cost of 2,604,000 THB and add five years of L5 operations at 300,000 THB per year.

Five-year TCO = 2,604,000 + 300,000 × 5 = 4,104,000 THB (approximately THB 4.1 million)

Over five years, the total is roughly 1.6 times the initial cost. If you get approval on the initial figure alone, you will fight for the maintenance budget every year from year two onwards. Presenting the TCO from the start is, in the end, the easier path.

The benefit estimate — THB 417,600 per year

Now the benefit side. Everything below rests on assumptions rather than measurements, and none of it is a guarantee that these benefits will materialise. The assumptions have deliberately been kept conservative.

Total annual benefit is 417,600 THB. From there, the simple payback period:

2,604,000 ÷ 417,600 ≈ approximately 6.2 years

What to do with a 6.2-year payback — evaluate it together with the projects above it

Approximately 6.2 years. Under most Japanese corporate investment criteria, that number does not get approved. The internal hurdle is typically three years, five at the outside. And note that 6.2 years is a simple calculation that has not deducted the L5 operating cost at all. Include operations and the payback period stretches further.

In other words, however carefully you build the model, the network on its own will not produce a number that passes an approval committee. That is not a flaw in the estimate. It is the nature of the asset.

A network does not create value by itself. IoT-based equipment monitoring, MES, AGV/AMR, electronic forms, traceability — the network is the precondition, the enabler, that lets those projects run. It is closer in character to electrical distribution or a compressed air line. Nobody calculates the ROI of running compressed air piping in isolation. You run the piping because there is equipment that consumes it; doing it the other way round inverts the logic.

So there is really only one way to construct the business case: evaluate the network together with the projects that sit on top of it.

• Put the AGV deployment, IoT equipment monitoring and electronic forms projects in the same fiscal year
• Treat the network as the shared foundation for all three and apportion the investment across them
• Calculate each project’s benefit within that project, and share only the foundation cost

Done this way, an investment that looked like a 6.2-year payback in isolation is evaluated inside a portfolio. The AGV project carries the material handling labour benefit; IoT equipment monitoring carries the benefit of visualising equipment downtime; electronic forms carry the quality cost benefit associated with building traceability. The foundation cost is common, so the more projects ride on it, the lighter each apportioned share becomes.

Conversely, the pattern most likely to fail is “there is no project above it, but let us refresh the Wi-Fi anyway.” Its effect cannot be measured, so it is the first item cut in the next budget round and is left half-finished. If you are starting to think about a network refresh, decide first what you intend to run on top of the foundation. If that is still blank, the right order is to sort that out before anyone discusses money.

Designing OT/IT convergence — the Purdue model and IEC 62443 zones and conduits

Of all the layers in factory network design, this is the one where reversing course is hardest. Taking it in order.

The Purdue model still works as a shared language

The Purdue model (Levels 0 to 5) remains valid as a shared language in 2026. Organising the plant from Level 0 sensors and actuators, through Level 1 PLCs, Level 2 supervisory control, Level 3 manufacturing operations, up to Levels 4 and 5 on the business side, gives OT and IT people a common drawing to argue over. Simply being able to say “this is a Level 2 discussion” in a design meeting changes the efficiency of the conversation completely.

But the Purdue model on its own cannot express the communication paths that exist today. That is the practical focus in 2026.

Zones are grouped by the severity of the consequence of compromise, not by hierarchy level

The mainstream way to implement segmentation now is IEC 62443-3-2 “zones and conduits”. The decisive point is how zones are grouped.

Zones are not grouped by Purdue level. They are grouped by the severity of the consequence if they are compromised. A safety instrumented system (SIS) and a general-purpose HMI may sit at the same Purdue level, but the severity of consequence if each is compromised is entirely different, so they can and often should end up in different zones. A design that has not made this conceptual shift produces the worst of both worlds: the hierarchy is separated but the risk is not.

The strength required of each zone is expressed as security levels SL1 to SL4, defined in IEC 62443-3-3. Not every zone needs to be SL4. Assigning levels in proportion to consequence is the only way to balance cost against safety.

Treat the Level 3.5 DMZ as critical infrastructure

All communication between the IT side and the OT side is brokered through the Level 3.5 DMZ. Leave even one path that reaches from IT into OT directly, and that path becomes the single way through. Equally, if the DMZ is treated casually as “the place where relay servers live”, it becomes the weakest point in the architecture. The DMZ should be treated as critical infrastructure and included in monitoring and redundancy.

The practical procedure is three steps.

1. List every conduit on the drawing — cloud connections, vendor remote maintenance, edge brokers, sensor paths. Write out every path that carries traffic, without exception.
2. Derive firewall rules from the listed conduits — do not build the ruleset by accumulating exceptions. The moment that inverts, you are a few years away from a several-hundred-line ruleset that nobody dares delete from.
3. Review through change management every time a new connection is added — the zone and conduit drawing is not a deliverable you produce once. It is maintained as a living document.

Paths the Purdue model cannot express

Here are five paths that do not sit neatly in the Purdue hierarchy and must be listed explicitly on the drawing.

None of these five can be discovered by physically tracing cables. Even if the network diagram is current, if these are missing the zone design does not hold together. This is exactly where factory OT security design and network design stop being separable disciplines.

Wireless design for AGVs and AMRs — roaming handover requirements

Why the AGV stops at the same corner

AGVs and AMRs communicate while moving. As the vehicle moves away from the AP it is currently associated with, it must switch to an AP with a stronger signal, and that triggers a rescan. The time consumed by that rescan and reauthentication is, from the AGV’s point of view, an outage.

The symptom of stopping at the same spot means that location coincides with a cell boundary. On a handheld carried by a person, a few hundred milliseconds there goes unnoticed. On a moving AGV, a gap in control instructions or navigation data means the vehicle either fails safe and stops or loses its route.

Where 50–100 ms is actually guaranteed

In industrial wireless design, handover time is treated as an explicit numerical requirement. There are products that achieve handover times of under 50 ms through intelligent roaming control, and industrial WLAN implementations that state cell-to-cell handover under 100 ms.

What matters is that this number is not determined by the AP datasheet alone. It emerges from three things meshing together: the AP-side capability, the roaming control on the controller, and the implementation of the wireless client on board the AGV. So what belongs in a procurement specification is not “Wi-Fi 6 capable AP” but “AP-to-AP handover time of under 100 ms along the AGV travel route, verified by measurement during the post-deployment survey.”

And that can only be validated by measurement after the APs are installed. Run the AGV for real during the post-deployment survey and measure handover time at every point along the route. Projects that skip this step reliably surface the problem later in the form of “the AGV keeps stopping.” For the AGV/AMR deployment decision itself, travel environment and fleet sizing are covered in our article on AGV and AMR deployment.

Should you wait for Wi-Fi 7 MLO?

Wi-Fi 7 MLO (Multi-Link Operation) targets deterministic low latency by connecting across multiple bands simultaneously, and its fit with industrial use cases is widely discussed. However, volume production of enterprise-class silicon is put at 2026, which makes “wait until MLO is widespread” an unrealistic position for a factory planning a deployment now.

The sensible approach is to lock down what can be decided today — AP placement, roaming control, wired backbone capacity — while leaving headroom in cabling and PoE budget so that MLO can be adopted at the next AP refresh. Physical cabling is a ten-year asset, so the decision to build slack into it now has an outsized effect on refresh costs five years out.

Regulations to check before you choose equipment in Thailand and Vietnam

Check the regulations before selecting equipment. Reverse that order and you risk the worst outcome: hardware you have already procured that you are not permitted to operate.

Thailand’s 6 GHz band — 250 mW indoor, 25 mW indoor and outdoor

In Thailand, the NBTC has opened 5,925–6,425 MHz (lower 6 GHz, 500 MHz of bandwidth) for wireless LAN use, with the technical standard NBTC TS 1039-2566. The EIRP ceiling splits into two categories.

In addition, equipment exceeding 25 mW EIRP may not be battery powered.

The gap between those two categories drives the design directly. Inside a production building you have 250 mW available, so a design that exploits the 6 GHz band is viable. Blanketing a large outdoor yard entirely in 6 GHz, on the other hand, is not realistic under the 25 mW constraint. Plan on covering outdoor yards and inter-building transit routes with 5 GHz, wired links or other means. The assumption that “it is Wi-Fi 6E capable, so it works everywhere” does not survive contact with Thai regulation.

Thailand’s local 5G — the 4.8 GHz framework is still being finalised

The NBTC has indicated an intention to make 100 MHz in the 4.8 GHz band available for private 5G. That corresponds to 3GPP Band n79 (4.4–5.0 GHz). The mechanism is a PNO (Private Network Operator) licence framework, under which factory and industrial estate operators would be able to receive an assignment free of charge on an application basis. There are conditions: use is limited to internal, non-commercial use by the operator, and diversion to public mobile services or integration and roaming with existing 2.6 GHz networks is not permitted.

Most importantly, as at the publication date of this article, no official commencement date and no application deadline have been announced. Local 5G in Thailand is therefore not yet available — the framework is still being finalised — and it is not at a stage where it can be built into a 2026 capital plan as an assumption. Deferring a wireless refresh while waiting for local 5G is not a decision we would recommend.

Vietnam — Circular 01/2025 and QCVN 136:2025

In Vietnam, the Ministry of Information and Communications (now the Ministry of Science and Technology) issued Circular 01/2025/TT-BKHCN on 31 March 2025, opening 5,925–6,425 MHz for licence-exempt WLAN use. It took effect on 15 May 2025.

Further, the technical standard QCVN 136:2025/BKHCN was issued on 30 November 2025 and takes effect on 1 January 2027, replacing the existing QCVN 47:2015/BTTTT.

For companies with Vietnamese sites, the practical implication is clear. Because the applicable conformity standard changes in January 2027, ask your vendor whether equipment procured during 2026 will require recertification under QCVN 136. Compliance at the time of purchase does not guarantee that a future expansion can be added using the same part number. We recommend getting the vendor’s answer in writing rather than verbally.

Type approval is a precondition for procurement

Equipment used in Thailand requires NBTC type approval. In Vietnam, conformity with the QCVN standards above is required. Equipment that head office has standardised on in Japan cannot be assumed to be usable as-is in Thailand or Vietnam.

Ironically, the companies most likely to miss this check are the ones running a disciplined global standardisation programme from head office. At the point of requesting quotations, put two questions to the vendor in writing: has this part number obtained NBTC type approval in Thailand, and is there a plan for QCVN 136 conformity in Vietnam. This is not a design conversation. It is a customs clearance and operations conversation.

Can the 200% tax deduction be applied to network investment?

Thailand offers a 200% corporate income tax deduction for SME investment in digital products and services. It took effect on 7 February 2026, applies retroactively to expenses incurred on or after 24 June 2025, and runs until 31 December 2027.

What qualifies are products and services from vendors registered in depa’s “Thailand Digital Catalog”. As at early 2026, more than 400 entries were registered.

Here is where care is needed. Cabling work and general-purpose hardware do not automatically qualify. Whether an item qualifies is determined by whether it appears in the depa catalog. The practical sequence is therefore:

1. Before requesting quotations, check whether the candidate vendors are registered in depa’s Thailand Digital Catalog
2. If they are, ask the vendor to identify explicitly which line items in the quotation correspond to catalog-registered products and services
3. Confirm the final eligibility determination with your own accounting and tax advisors

The order matters. Realising after contract signature that the investment would have qualified does not let you apply the deduction retroactively if the conditions were not met. Conversely, if you have several vendor candidates offering equivalent functionality, catalog registration becomes one legitimate selection criterion. With the window closing on 31 December 2027, it is well worth examining within a 2026 investment plan.

Six common failure patterns in factory network design

Six failures we see repeatedly on site.

1. Repurposing office-grade APs

They do not withstand dust, temperature and metal reflection, and become unstable within six months. The amount saved on initial cost is small relative to what is lost.

2. Skipping the site survey

The order is placed based on AP positions drawn on a floor plan, and coverage holes appear once the plant is live. Two surveys are required: pre-deployment and post-deployment. The first is for design; the second is for verification and correction by measurement after the real hardware is installed. The post-deployment survey is the one most often cut on the grounds that “the installation is finished”, but including the AGV roaming validation, it is the final quality gate.

3. Deploying AGVs with no roaming design

This causes stoppages and route deviation. The 50–100 ms handover requirement has to be fixed up front as a specification on the AP and controller side. Starting the wireless conversation after the AGVs have been ordered is too late.

4. Letting the IT department decide alone

The requirements of the control network (Levels 0 to 2) do not get reflected, a separate network is later added on the OT side, and you have paid twice. Put manufacturing engineering and maintenance staff in the design meetings. The OT knowledge that “this line absolutely cannot be stopped” does not appear anywhere on the IT department’s drawings.

5. Deferring segmentation

Attempting to redo IEC 62443 zone design after the plant is live runs into the fact that too many lines cannot be stopped, which makes it effectively impossible. That is why we recommend not deferring L4 even though it accounts for about a quarter of the initial cost. Segmentation is not a later change request; it is the original design.

6. Treating 6 GHz and local 5G as “the future solution” and postponing

Given the regulatory and equipment reality — Thailand’s 25 mW outdoor constraint, the unannounced commencement date for n79, enterprise Wi-Fi 7 silicon in volume production put at 2026 — decide what can be decided now. Deferring a decision on the grounds of future technology means the stoppage cost keeps accruing every month in the meantime.

The 90-day roadmap

Finally, the execution plan. Five phases across 90 days.

The failure log in Day 1–15 is the highest-value activity

The highest return in this roadmap comes from the first 15 days. A record of when, where and how many devices failed to communicate is simultaneously an input to the design and the evidential basis for the benefit estimate in the approval paper. The “6 hours per month of line stoppage” figure used above remains an assumption without that record. With measured data, you can rebuild the estimate using numbers specific to your plant.

Confirming in the same phase whether upstream projects exist connects directly to the bundling argument made earlier. Move into design with that field blank and the five-year TCO ends up walking around on its own with nothing to justify it.

Why zone definition runs in parallel during Day 16–35

Placing the site survey and IEC 62443 zone definition in the same phase is deliberate. If zone definition is pushed later, AP placement and VLAN design get frozen first, and physical placement then obstructs any attempt to draw zone boundaries afterwards. If the zones are settled first, which AP belongs to which zone falls out of the design naturally.

What to protect during installation in Day 56–80

Work requiring line stoppage goes into planned shutdown days. That sounds obvious, but network work is routinely squeezed into a weekday afternoon on the basis that “it only takes a moment”, and unforeseen trouble then affects production. Also, the L4 boundary firewall and authentication must be verified in a test environment before production cutover. This is the layer where discovering a problem only at cutover means a slow and painful rollback.

Frequently asked questions

How much does a factory wireless LAN cost?

For a model factory of 10,000 m² with 32 APs, TOMAS TECH’s own estimate puts the initial cost at approximately THB 2.6 million (2,604,000 THB), with a range of THB 2.0 million to THB 3.3 million depending on scale and requirements. Operations add THB 250,000–400,000 per year, and using THB 300,000 per year gives a five-year TCO of 4,104,000 THB. Wireless equipment (L3) accounts for just under 40 percent of the initial cost; the remainder is cabling, switching and boundary security.

What is the difference between an industrial network and an office LAN?

Three things. First, environmental conditions: equipment must be selected to withstand heat, cold and dust, and office APs cannot be repurposed. Second, the RF environment: reflection and shadowing from metal fixtures, racking, liquid tanks and material handling equipment dominate, so distance-based design does not work. Third, the target metric: an office wants “connects and is fast”, while a factory wants “does not drop and has predictable latency”. When AGVs are involved in particular, AP-to-AP handover time becomes an explicit numerical requirement.

Can Wi-Fi 6E be used in Thailand?

Yes. The NBTC has opened 5,925–6,425 MHz (500 MHz of bandwidth) for wireless LAN use, with the technical standard NBTC TS 1039-2566. However, there are two EIRP categories: a maximum of 250 mW for indoor-only use and a maximum of 25 mW for indoor and outdoor use. Equipment exceeding 25 mW may not be battery powered. Inside a production building you can use 250 mW, but covering an entire outdoor site with 6 GHz is not realistic under the 25 mW constraint. Equipment used in Thailand also requires NBTC type approval.

Should we choose local 5G or Wi-Fi?

In Thailand as at 2026, designing on the assumption of Wi-Fi is the realistic choice. The NBTC has signalled an intention to make 100 MHz in the 4.8 GHz band (3GPP Band n79) available for private 5G, with factory and industrial estate operators able to receive a free assignment on an application basis under a PNO licence framework — but no official commencement date or application deadline has been announced as at the publication date of this article. Use is also limited to internal, non-commercial use, with no diversion to public mobile networks and no integration or roaming with existing 2.6 GHz networks. Treat it as an option whose framework is still being finalised: leave room for future expansion, and decide on Wi-Fi now.

Do we really need two site surveys?

Yes. The pre-deployment survey serves the design (RF measurement, identification of interference sources, draft AP placement); the post-deployment survey serves verification and correction by measurement after the real hardware is installed. The purposes are different. When AGVs are deployed in particular, roaming handover time along the travel route can only be confirmed by live running trials during the post-deployment survey. The estimate allocates 80,000 THB to the pre-survey and 60,000 THB to the post-survey.

Can the 200% tax deduction be applied to network investment?

It depends. Thailand’s 200% deduction for SMEs (effective 7 February 2026, retroactive to expenses on or after 24 June 2025, running until 31 December 2027) applies to products and services from vendors registered in depa’s “Thailand Digital Catalog”. Cabling work and general-purpose hardware do not automatically qualify. Before requesting quotations, check whether the candidate vendors are registered and ask them to identify which line items correspond, then confirm final eligibility with your own accounting and tax advisors.

Can we refresh the network without stopping production?

Partly. Wi-Fi has the advantage that network coverage can be added while existing equipment stays exactly where it is, so within the scope requiring no cabling work you can proceed without stopping the line. Laying fibre backbone, swapping switches inside field cabinets and cutting over the boundary firewall, on the other hand, may require stoppages. That is exactly why the 90-day roadmap places work requiring stoppage into planned shutdown days during Day 56–80.

Summary

Six points for making a factory wireless LAN deployment succeed.

1. Do not import the office design philosophy. Metal reflection and shadowing, temperature and dust, and a target metric of “does not drop” make the factory fundamentally different from the office.
2. Manage cost in five layers. L1 physical cabling, L2 wired hardware, L3 wireless and L4 segmentation and security up front, plus L5 operations annually. Wireless equipment is just under 40 percent of the initial cost.
3. Do not calculate ROI in isolation. The model estimate gives a simple payback of approximately 6.2 years, which exceeds most internal criteria. The network is an enabler for IoT, MES, AGV and traceability, and must be evaluated together with those projects.
4. Do not defer L4. Group IEC 62443 zones by severity of consequence, list the conduits, and derive firewall rules from that list. Redoing it after the plant is live is effectively impossible.
5. Check the regulations before selecting equipment. Thailand’s two 6 GHz EIRP categories and type approval, Vietnam’s switch to QCVN 136 in January 2027, and the fact that local 5G in Thailand is still being finalised.
6. Run it in 90 days. The failure log from the first 15 days becomes both a design input and the basis of the business case.

Most requests that arrive as “the Wi-Fi is unstable, we want more APs” are in reality conversations about cabling, switching, boundary design and the projects above. Follow the order and these five layers assemble without drama.

Talk to us

If you are considering a factory wireless LAN or industrial network refresh, get in touch through the TOMAS TECH contact form. There is no need to hand over the whole scope at once. “We would like a site survey only” or “we just want to understand how to write the bundled business case” are perfectly good places to start. Working from Bangkok with Japanese-affiliated factories across Thailand, we can help you organise the situation around what is actually on your shop floor.

References

• Thailand NBTC Notification: Wi-Fi 6E and Wi-Fi 7
• Thailand NBTC frees 4.8 GHz for private 5G in factories
• QCVN 136:2025/BKHCN — RF for 6 GHz Wi-Fi
• Purdue Model 2026: OT Network Architecture
• OT Network Segmentation: When Purdue Falls Short and IEC 62443 Takes Over
• Cisco Industrial Automation Wireless Design Guide
• Private 5G vs Wi-Fi 6 in Smart Factories
• Allied Telesis — Wireless LAN for Manufacturing (Japanese)
• Panasonic — Network Column (Japanese)
• Network Construction Cost Benchmarks (Japanese)
• Thailand introduces 200% tax deduction for SME digital spending