MES Implementation in Thailand — Cost, Comparison and the ERP Boundary

“Head office rolled out ERP. We already have a production management system on site. And now the production engineering group in Japan is telling us to put in an MES.” When a Japanese-owned factory in Thailand asks us for help with an MES implementation, the starting point is almost always some version of that sentence. The problem is rarely a lack of understanding of feature lists. The problem is that systems are already running, and nobody has drawn the line that says which part of the business belongs to which layer. This article is not another general introduction to the manufacturing execution system category. It is about how to draw that line, and how the cost structure and the sequence of work fall out of it once the line exists.

The three layers you need to map before an MES implementation

Start by laying out what you already have, on ISA-95 levels

The single biggest reason MES discussions go in circles is that there is no common basis for comparison. You can put Vendor A’s proposal next to Vendor B’s proposal all day, but if nobody has defined where the gap in your own architecture actually is, all you can compare is how many features each one lists.

The first job is to put everything currently running onto one page, arranged according to ISA-95 — the international reference model for the system hierarchy in manufacturing. Siemens’ published explanation of the ISA-95 framework organizes the layers as follows.

MES sits at ISA-95 level 3 in that table. It takes the “what, when, how much” that ERP has decided and translates it into a granularity the shop floor can execute, then returns the actuals generated on the floor back to the management layer. A two-way pipe — that is the essential role of MES. It is not a one-way instruction system pushing orders downward, and it is not a one-way collection system pushing actuals upward. Implement it without understanding that bidirectionality and you end up with an expensive terminal for typing in daily reports.

The typical starting point at a Japanese-owned plant in Thailand

When we actually run this inventory at a Japanese-owned factory in Thailand, the result usually looks like this.

• Level 4 (ERP): Present. Rolled out under head office direction. But from the Thai site’s point of view it is a box for sending numbers back to Japan, and it is not used for decisions on the floor
• Level 3 (MES): This is the gap. Either nothing at all, or a scattered patchwork of Excel macros and small locally built tools
• Levels 1-2 (PLC/SCADA): Present at the machine level. But it differs by equipment maker, and there is no open path for getting data out. Where SCADA exists, it is closed off within a single line
• Level 0: Paper work instructions, paper daily reports, whiteboards, and hand transcription from whiteboard into a ledger

In other words, the gap is not only level 3. It is level 3 *and* the route that carries data up from levels 1-2 into level 3. Miss that and you commit the classic mistake: you take the MES license cost to your approval committee, and then a shop-floor data acquisition cost of comparable size lands on top of it later. We will get to cost in detail further down, but structurally it is very nearly decided at this stage.

What is MES — the scope a manufacturing execution system covers

The MESA-11 functions as a whole picture

The common language for describing the functional scope of MES is the set of eleven functions defined by MESA (Manufacturing Enterprise Solutions Association).

The important point is that you do not need to implement all eleven. In fact, the projects that try to implement all eleven are the ones that fail most reliably. The three functions where the benefit becomes visible first are data collection, process management, and product tracking. Those three alone establish the foundation for shop-floor visibility and traceability. Quality management and scheduling produce far more realistic requirements when they are layered on after those three are running stably and data has accumulated.

Why MES adoption in Japan has stayed limited

Commentary from Hitachi Solutions and ABeam Consulting notes that MES adoption in Japan is limited compared with European and American companies. The reasons given are the size of the implementation cost and the length of the implementation period, and adoption among small and mid-sized enterprises is almost negligible. Behind that lie a shortage of IT staff and the burden of implementation cost.

This does not mean Japanese manufacturers have no need for MES. If anything the opposite is true. In Japanese manufacturing, where shop-floor capability is strong, the individual mechanisms invented on the floor — Excel, homemade tools, paper routines — already work well enough that the motivation to replace them with an MES struggles to appear. Following ABeam’s point: whether to keep existing subsystems or replace them with MES is a judgment that has to be made carefully, weighing the load of existing operations and how accustomed the floor already is to them.

And that judgment is exactly what the next section — how to draw the line — is about.

MES vs ERP, and how both differ from a production management system

Separating the three layers’ roles in a table

Start with definitions.

That last row matters most in practice. Because MES reaches into the execution layer of the shop floor, its availability requirement is fundamentally different from ERP’s. That translates directly into infrastructure cost, as discussed below.

Draw the line by refresh frequency and point of origin, not by feature name

This is the core of the article.

Argue the MES vs ERP split in terms of feature names and the discussion never converges. The reason is that functions like “production dispatch”, “actuals collection”, and “inventory management” exist in the production module of ERP, in a production management system, and in MES. As commentary from Layers Consulting and Unifinity points out, ERP’s production module can handle part of production dispatch and actuals collection, so unless you make explicit how far ERP carries the load and where MES takes over, functional overlap and broken integration become likely.

So rather than feature names, we recommend drawing the line on the following two axes.

Axis 1: refresh frequency

A production management system updates at daily or weekly granularity; MES updates minute by minute and hour by hour. That gives you a decision rule.

Can this information do its job with one update per day? If yes, it belongs on the ERP / production management side. If no, it belongs on the MES side.

Some examples to check it against.

Axis 2: point of origin of the primary record

If the information is first created on the floor — equipment, operators, inspection — capture the primary record in MES. If it is created by human judgment or from a document, hold it in ERP / production management.

This axis earns its keep because it eliminates double entry. The classic failure is quadruple entry: the operator keys actuals into a handheld, the line leader transcribes them into Excel, production control enters them into the production management system, and at month-end finance enters aggregated values into ERP. Fix the point of origin in exactly one place and everything downstream becomes a matter of integration, not re-entry.

A practical procedure for drawing the line

1. Pick one target process: not the whole company — narrow it to one line or one group of operations
2. Collect every piece of paper and every Excel file used in that process: work instructions, daily reports, inspection records, defect tags, die inspection sheets. These become the primary source material for requirements
3. On each sheet, write down its refresh frequency, its point of origin, and who reads it
4. Nominate for MES only those with a refresh frequency finer than daily and a point of origin on the floor
5. For everything left over, check whether it already exists in ERP or the production management system: if it does, do not build it again
6. Among the MES candidates, separate what can be acquired automatically from levels 1-2 from what requires manual entry: this is what your shop-floor data acquisition estimate will be based on

Follow this procedure and your RFP to vendors changes from “please send us your MES feature list” into “of these 15 forms, we want to replace these 9 with MES; 4 of the 9 are acquired automatically from PLCs and 5 are tablet entry.” Both the accuracy of the quotations and the comparability between vendors are settled at that moment.

If you are still at the stage of selecting the production management system itself, it is safer to sort out the boundary first. We have organized the comparison criteria for production management systems on their own in our production management system comparison article, which is worth reading alongside this one.

Making the first three MES functions concrete

Of the MESA-11 functions, here is what the three you should tackle first actually involve in implementation terms.

Data collection — where you take it from, and how

“Data collection” tends to be discussed in the most abstract terms, but in practice each acquisition route is an entirely different piece of work.

At Japanese-owned plants in Thailand it is normal to find equipment vintages spanning more than twenty years mixed together. Within the same line you will find a machine that talks OPC UA standing next to a machine whose only external output is a lamp. Consequently, a data collection estimate is not final until you have built a machine-by-machine list. Any estimate that settles for machine count multiplied by a unit price will move later. It always does.

Process management — detecting delay and handing it to a human

The value of process management is not that progress becomes visible. It is that delay against plan is detected while there is still time to respond. The minimum implementation requires three things.

• Hold standard cycle time (or standard labor hours) per operation as master data
• Calculate the variance against measured actuals in real time
• When a threshold is crossed, surface it on the line leader’s tablet or on an andon

This is where multilingual operation enters the picture. If the threshold-exceeded message does not appear in Thai, the Thai line leader cannot react to it. Push localization into a “later” phase and process management will barely function at all.

Product tracking — decide the traceability granularity first

Product tracking has the widest requirement range of any of the functions. What you must decide is granularity.

Granularity is largely not yours to decide. It is determined by customer requirements and by regulation. For automotive parts, IATF 16949 requirements and the traceability clauses written into the customer’s quality assurance agreement are effectively your specification. Decide “lot level for now” without reading those documents and you will be rebuilding after the next customer audit. We cover the scope and cost logic of traceability in detail in our traceability system build article.

Breaking MES cost into five layers

First, understand that license cost is only part of the total

MES cost discussions fail to connect because the scope of the figure each vendor presents differs by vendor. A software vendor quotes licenses and basic configuration. A systems integrator quotes a figure that includes the physical work on the floor. Two documents both labeled “MES quotation” contain different things, so a comparison that lines up only the totals will lead you to the wrong decision.

We therefore recommend splitting cost into the following five layers and obtaining an estimate for each layer separately.

Layer 1: a sense of MES license and subscription pricing

For reference, here are the ranges circulating as Japanese domestic market pricing (source: AI Souken, a Japanese industry research site. These are typical Japanese domestic price ranges; for deployment at a Thai site they will move with exchange rates, local labor rates, and differences in local support structure. USD figures below are approximate conversions at roughly 1 USD = 155 JPY and are for reference only.)

That range corresponds to layer 1 only. If the figure on your approval document is only this, it will diverge from the real total.

How layers 2 to 5 move the total

In most cases the total cost of an MES implementation is decided not by layer 1 but by layers 2 through 5. Layer by layer, here is why.

Layer 2, shop-floor data acquisition, depends completely on equipment vintage. A machine with an open communication port is a configuration task. A machine whose only output is a lamp requires wiring work. A machine that emits no signal at all requires added sensors, plus study of where to mount them and fabrication of the mounting hardware. If a line has 20 machines and 12 of them are retrofit candidates, it is not at all unusual for this layer alone to exceed the license cost.

Layer 3, infrastructure, spikes on the availability requirement. As noted above, when MES stops the floor stops, so you cannot build it on the same architecture as ERP. You need redundancy in the industrial network, local continued operation at the edge (a design in which the floor keeps working when the upstream link drops), and UPS provisioning. The in-plant wireless environment deserves particular care: in an environment with metal structures and noise from running equipment, reusing office-grade access points as-is will not give you stable communication. We explain the design philosophy for this area in our factory wireless LAN and industrial network article.

Layer 4, upstream integration, is determined by whether the existing ERP can be modified and by the state of the master data. ERP rolled out under head office direction often cannot be modified at the request of a site, in which case the design falls back on intermediate tables or file-based integration. And where master data is dirty — the same part number registered under multiple codes, BOM revisions that do not match the floor, routing masters that differ from reality — cleaning and reconciling it becomes the single largest effort in practice. A vendor cannot do this work on your behalf, so it has to appear in the estimate as internal effort.

Layer 5, operational adoption, is reliably heavier at a Thai site than in Japan. The next section deals with it in detail.

For a cost breakdown limited to the process management scope, we have organized finer ranges in our process management system cost article.

MES comparison: three product types and vendor direction in 2026

Cloud SaaS, Japanese packaged, and high-end suites

Here are the items you should always verify when comparing.

• Support presence in Thailand and languages covered (Japanese, English, Thai — how far does it go, and are support hours in ICT or Japan time?)
• Offline behavior (on a cloud product, when the link drops, can shop-floor terminals keep accepting entry, and do they synchronize on recovery?)
• Proven PLC protocols (is there a track record connecting to your own equipment makers?)
• Multilingual UI (can master data names, printed forms, and error messages be localized, not just screen labels?)
• How ERP integration is implemented (standard interface, or bespoke development?)
• Upgrade path once customized (will modifications survive updates without breaking?)

Vendor direction in 2026 — redefinition around AI copilots

The moves by major vendors as of 2026 share a common direction (source: AI Souken).

The broad trend is that MES is being redefined from a shop-floor efficiency tool into a data foundation connecting management and the floor, with natural-language search and AI copilot decision support getting under way in earnest.

There is a practical caveat, though. An AI copilot only functions usefully when clean data has accumulated underneath it. If equipment data is not being captured, if lot linkage is broken, if master data differs from reality, then evaluating AI features in that state is meaningless. The realistic approach is to treat 2026 vendor direction as headroom for future AI use, and to keep the six items in the previous subsection as your primary selection criteria.

On market size, figures vary by research firm, but forecasts put the market at roughly USD 30–40 billion around 2032, with a CAGR of about 10% (10.04%) (source: market research report via GII; note that estimates from different research firms vary in range). The analysis attributes the growth to quality management and regulatory compliance requirements in heavily regulated sectors — pharmaceuticals, food and beverage, aerospace. That is the same dynamic as the situation of automotive parts suppliers in Thailand, where customer audit requirements are what pull the trigger on traceability investment.

Five reasons MES implementations stumble at factories in Thailand

What follows is not a theoretical risk list. Every item bears directly on cost and on uptime.

1. Multilingual UI — running three languages is design, not translation

At Japanese-owned plants in Thailand the normal structure is three-tiered: operators work in Thai, local management in Thai plus English, and Japanese expatriate staff in Japanese. Build the MES screens in one language and some tier will always be unable to use it.

Three points bear on cost. First, localization is needed not only for screen labels but for master data names — item names, operation names, defect names — and that depends on how the system is designed. Even products that advertise multilingual support sometimes hold master data in only one language. Second, translation quality. Literal translations of process terminology and defect names do not mean anything to operators, so you need effort budgeted for building a glossary with local line leaders involved. Third, printed forms and error messages. Leave those in Japanese and local staff cannot respond when something goes wrong, and the operation reverts to paper.

2. Retention and turnover of local staff

MES operation tends to depend on specific individuals. If there is exactly one person who can touch data collection settings, one who maintains master data, and one who handles first response to abnormalities, then a resignation stops maintenance even though the system keeps running.

The countermeasure is removing that individual dependence, and “write a manual” is not enough. In practice three things work: (1) record operating procedures as video in Thai, (2) design so that master data changes are completed through the UI, requiring no SQL or scripting, and (3) secure an alternative route for first response through a maintenance contract with a local systems integrator. All three cost money, so they should be loaded into layer 5, operational adoption, from the start.

3. Requirements are not handed over when expatriate staff rotate

This is the most reproducible failure mode at a Thai site. The expatriate who drove the implementation returns to Japan after three to five years, and the successor inherits the operation without knowing why the system was built the way it was. The result: unused functions stay in place, needed enhancements stop, and the next expatriate starts saying they want to rebuild it.

The countermeasure is leaving the rationale for requirements in a document. Not merely a specification, but a record of the reasoning behind decisions: why this operation was chosen for MES capture, why this form was kept, which customer requirement each element answers. The practical procedure for drawing the line described earlier — writing refresh frequency and point of origin onto each form — doubles as exactly that handover document.

4. Stability of power and network links

It varies with plant location, but you should design on the assumption that voltage dips, outages, and link failures occur more frequently than in Japan. Because the floor stops when MES stops, make the following three points explicit requirements.

• UPS provisioning: servers and edge devices, network equipment, charging for shop-floor terminals
• Continued offline operation: can shop-floor terminals keep accepting entry when the upstream link drops, and synchronize after recovery?
• Recovery of data consistency: is there a mechanism so that actuals recorded during an outage are neither duplicated nor lost?

If you are considering a cloud SaaS product, the second and third of those can become the deciding factors in selection. They cannot be verified in a demo environment, so make sure your PoC evaluation criteria include deliberately cutting the link and observing what happens.

5. Traceability requirements demanded in customer audits

For automotive parts it is IATF 16949; depending on the industry it may be another standard or a customer-specific quality assurance agreement. Either way, that document is effectively the MES specification. A common pattern: requirements are frozen based only on internal wishes, then a post-implementation customer audit produces findings — “this parameter also has to be recorded”, “the retention period is insufficient” — and the system has to be rebuilt.

Avoiding it is simple. Bring the quality assurance department in at the very start of requirements definition, and include the customer’s requirement documents and past audit findings in the primary source material for requirements. This is not additional cost; it is a matter of sequence, to prevent rework.

A 90-day roadmap for starting an MES implementation

The general implementation process is usually organized into four steps: (1) current-state analysis and requirements definition (several months), (2) PoC (several months), (3) horizontal rollout (six months to eighteen months), and (4) company-wide operation and AI integration (ongoing) (source: AI Souken). Overall that is a project on the scale of one to two and a half years.

But starting from that whole picture is what freezes decision-making. The practical move is to set a boundary: produce the material for a decision in the first 90 days.

Days 1-30: current-state analysis narrowed to a single process

The boundary definition table and the equipment data acquisition feasibility table produced in these four weeks are the foundation for every estimate and RFP that follows. Put the other way around: collecting quotations from several vendors before those two documents exist does not give you a comparison.

Days 31-60: run a PoC limited to three functions

Limit PoC scope to data collection, process management, and product tracking. Do not include quality management, scheduling, or maintenance management. Three reasons.

First, whether the floor works with those three functions is the largest uncertainty; verify that and you can make the remaining judgments. Second, adding functions extends the PoC period, and once the period extends the floor loses interest. Third, if it fails you can no longer identify what caused the failure.

PoC evaluation criteria should include more than whether functions work.

• Time required for shop-floor entry: measure how many seconds one entry takes and confirm it is at a level the floor can accept relative to the current paper routine
• Whether the Thai-language UI is workable: can Thai operators run it on their own, with no Japanese speaker explaining?
• Behavior when the link drops: cut it deliberately and confirm continued entry and post-recovery synchronization
• Data trustworthiness: do the system’s numbers match measured and visually counted values on the floor?
• Time from abnormality detection to response: how many minutes from detection until a person moves?

Days 61-90: evaluate and decide on rollout

Based on the PoC results, choose among three options.

1. Roll out as is: the three functions have taken hold and the data is trustworthy. Move to the next line
2. Add functions on the same line before rolling out: the three functions work, but without quality management included the floor ends up running two parallel systems
3. Revisit the approach: a structural problem has surfaced — entry burden too high, too many machines from which data cannot be acquired

“Do not implement all functions at once” is not a compromise for the sake of cost reduction. It is a design decision to raise the probability of success. If you move to a company-wide rollout before the three functions have taken hold on a single line and a single process, you will be unable to isolate problems when they appear, and the scale of the backtracking will be much larger.

As a precursor to shop-floor data acquisition, an approach that starts from visualizing equipment operation is also effective. On phasing the work, our factory IoT implementation guide is also worth consulting.

BOI incentives and their relationship to an Industry 4.0 investment plan

Whenever an MES implementation is under consideration in Thailand, the question of whether BOI (Thailand Board of Investment) incentives can be used always comes up.

Under BOI’s criteria for digitalization and smart operations, the applicant is required to propose an Industry 4.0 transformation investment plan certified by NSTDA (National Science and Technology Development Agency) and to implement that plan in full. The elements called for include the use of automation and network technology, data analytics and smart operations, and the introduction of digital technology into production processes.

An MES implementation maps directly onto those three elements, so its institutional fit is high. But two things should be made explicit.

First, the specific incentive rates and exemption periods have to be confirmed against the latest BOI announcement. Within what we could verify at the time of writing, we cannot state them definitively, so we do not present numbers. If you are structuring an investment plan on the assumption of a BOI application, always work from the announcement in force at the time of application and confirm with BOI or a local specialist.

Second, an NSTDA-certified investment plan is a precondition. In other words, it is not “we implemented MES, therefore incentives apply.” The sequence is “we implement MES as part of a certified Industry 4.0 investment plan, and we complete that plan.” Because preparing the plan and obtaining certification takes time, it has to run in parallel from an early stage, on a separate track from the MES implementation schedule.

In practical terms, if a BOI application is in view, it is efficient to begin organizing the material as an investment plan in parallel during the days 1-30 stage of the 90-day roadmap above. The boundary definition table and the equipment data acquisition feasibility table can be used directly in the plan’s narrative.

How to read the macro environment in Thai manufacturing

To think about the timing of an investment decision, here is the recent state of Thai manufacturing in numbers.

Production is expanding, but the momentum swings month to month

The S&P Global Thailand Manufacturing PMI came in at 53.5 in February 2026, 54.1 in March (the fastest improvement since December), and 52.7 in April (a three-month low). As of April that is twelve consecutive months of expansion (source: S&P Global).

There are two ways to read it. In that it has stayed above 50, the expansion phase is continuing, and the environment for investment in production equipment and systems is not bad. On the other hand the month-to-month swing is large, and no single month’s figure should drive a judgment.

In industrial production, January 2026 was +1.46% year on year. Within that, automotive production was reported at +6.3% (led by domestic demand for HEV and BEV) and electronic components including PCBs and ICs at +18.2% (global demand). Factory activity for full-year 2026 is expected to expand +1.5% to +2.5% (source: official statistics).

Automotive and electronic components — the two main sectors for Japanese-owned manufacturing in Thailand — are both positive, with electronic components in particular growing at double digits. When volume rises, paper-and-Excel routines tend to break. That is a practically strong reason to implement MES.

The fact that this is a rising-cost phase

There is a figure to be careful about at the same time. Input costs recorded their fastest rise since September 2022, attributed to increases in crude oil, fuel, transport, and raw material prices in connection with the situation in the Middle East (source: S&P Global).

In other words the current phase is “production is growing, but costs are rising.” In that situation, the role of an MES implementation is not only handling increased volume. If cost increases cannot be passed through in price, the remaining means of protecting profit narrow to internal yield improvement and higher equipment utilization. And to do either, the current state first has to be visible in numbers. That is precisely where the data collection and performance analysis capabilities of MES apply directly.

Investment environment and policy direction

FDI (foreign direct investment) application counts roughly doubled in January to September 2025, reported to be centered on digital infrastructure, batteries, electronics, and EV-related projects. The theme of the World Bank’s Thailand Economic Monitor for February 2026 is “Advanced Green Manufacturing for Growth,” indicating that upgrading manufacturing is a policy focus.

None of this leads directly to the conclusion that you should implement MES immediately. But a situation in which three conditions overlap — (1) production is in an expansion phase, (2) rising costs create pressure for internal improvement, and (3) policy and investment are pointed at manufacturing sophistication — can fairly be assessed as good material for an investment decision.

Frequently asked questions

What is MES?

MES (manufacturing execution system) is a system for managing and executing work on the manufacturing floor. In ISA-95, the reference model for the system hierarchy in manufacturing, it sits at level 3, between ERP (level 4) and control systems such as PLC and SCADA (levels 1-2). Its proper role is a two-way pipe: it expands the production plan ERP has decided into a granularity the floor can execute, and it returns the actuals generated on the floor to the layer above.

What is the difference between MES and ERP?

The time horizon and the origin of the data are different. ERP works at monthly, weekly, and daily granularity, handling information generated from documents — sales orders, purchasing, accounting. MES works at minute and hour granularity, handling information generated on the floor by equipment, operators, and inspection. Function names such as production dispatch and actuals collection exist in both, so trying to separate them by function name produces overlap and broken integration. The practical approach is to draw the line by refresh frequency and by where the primary record originates.

If we already have a production management system, do we still need MES?

You cannot say it is unnecessary, but the scope you need becomes narrower. Because a production management system handles plans and actuals at daily and weekly level, “how much are we planning to build today” and “how many did we make yesterday” should already be covered. MES becomes necessary for the parts where minute-level and hour-level awareness is operationally required — detecting equipment stoppages, changing today’s release sequence, recording lots passing operations, immediate quarantine when a defect appears. If that part is being run on paper and Excel and is not working, it is MES territory.

How much does an MES implementation cost?

Looking only at software license and subscription, the ranges published as typical Japanese domestic pricing are: cloud SaaS at a few million to JPY 20 million (approx. USD 19,000–129,000) initial and JPY 1–5 million (approx. USD 6,500–32,000) per year; Japanese packaged products at JPY 10–30 million (approx. USD 64,500–193,500) initial and JPY 2–5 million (approx. USD 13,000–32,000) per year; high-end suites at JPY 30 million to several hundred million (approx. USD 193,500 to several million) initial and JPY several million to several tens of millions (approx. USD 19,000 to several hundred thousand) per year (source: industry research article, typical Japanese domestic price range; USD figures are the same approximate conversions noted earlier). What actually moves the total, however, is four other layers: shop-floor data acquisition (PLC connection, retrofit sensors, terminals), infrastructure (industrial network, servers and edge devices, UPS), upstream integration (ERP interface development, master data cleanup), and operational adoption (localization, training, maintenance). Shop-floor data acquisition in particular depends on equipment vintage, so the figure is not final until you have built a machine-by-machine list.

Can a small or mid-sized factory implement MES?

Yes, provided the scope is narrowed. Even in Japan, MES adoption among small and mid-sized enterprises is limited, with implementation cost and a shortage of IT staff cited as the reasons. So rather than implementing all functions, the realistic form is to limit yourself to data collection, process management, and product tracking, and to start with a single target process. A cloud SaaS product can hold the initial investment down, but always confirm in advance how it behaves when the link drops and whether it can accommodate custom shop-floor requirements.

How long does an MES implementation take?

The general framing is several months for current-state analysis and requirements definition, several months for PoC, six months to eighteen months for horizontal rollout, and continuous operation after that — one to two and a half years overall. That said, if you get through “30 days of current-state analysis on one process, 30 days of PoC limited to three functions, 30 days of evaluation and the rollout decision” in the first 90 days, you will have the material you need for the investment decision. Cutting those 90 days out first moves decision-making along better than working backwards from the full schedule.

What is the difference between SCADA and MES?

SCADA sits at ISA-95 levels 1-2 and handles equipment monitoring and control. What it deals with is equipment signals, alarms, and running state, and its subject is the equipment. MES is at level 3; what it deals with is manufacturing orders, lots, operators, and actuals, and its subject is manufacturing as a business process. SCADA tells you whether equipment is running normally, but it does not trace which lot of material went into a given product, or when and by whom it was made. The two do not compete: the relationship is that MES ties the data SCADA acquires into the context of the business process. That linkage between the control layer and the business layer is what the term OT IT convergence describes.

Summary

The success or failure of an MES implementation is very nearly decided before product selection begins. The key points.

• Inventory the layers first: put what you have today onto one page across ISA-95 levels 0 to 4. The typical picture at a Japanese-owned plant in Thailand is “ERP exists / a production management system exists / level 3 and the route that lifts data from levels 1-2 are both blank”
• Do not draw the line by function name: draw it on two axes — refresh frequency (does daily suffice, or is minute-level needed?) and where the primary record originates (a document, or the floor?). Arguing by function name invites overlap and broken integration
• Implement three functions first: data collection, process management, product tracking. Quality management and scheduling come after data has accumulated
• Take estimates in five layers: license, shop-floor data acquisition, infrastructure, upstream integration, operational adoption. What moves the total is layer 2 onward, and shop-floor data acquisition in particular depends on equipment vintage
• Build the five Thailand-specific factors into the cost: three-language operation, retention of local staff, requirements handover when expatriates rotate, stability of power and network links, traceability requirements demanded in customer audits
• Produce decision material in 90 days: 30 days of current-state analysis on one process, 30 days of PoC limited to three functions, 30 days of evaluation and the rollout decision. Do not implement everything at once
• Work the BOI track in parallel: an NSTDA-certified Industry 4.0 investment plan is the precondition. Confirm specific incentive figures against the latest BOI announcement

Thai manufacturing has now posted twelve consecutive months of PMI expansion, while input costs are in their fastest rise since September 2022. When production grows and costs rise at the same time, paper-and-Excel routines hit their limit first. MES is not an investment to postpone that limit; it is an investment to know, in numbers, where the limit actually is.

TOMAS TECH is a systems integrator based in Bangkok, Thailand, working in factory IT and OT for Japanese-owned manufacturers. Before you get to product selection or quotations, you are welcome to talk to us about the earlier question — given that ERP and a production management system already exist, which parts of the business should MES carry. We can start simply by looking at your current system architecture and the forms in use on your floor, and inventorying together where the gaps and the overlaps are. So even if you are still at the stage of assembling material for a decision, feel free to get in touch via our contact page.

References

• What is a Manufacturing Execution System (MES)? — AI Souken (Japanese-language industry research site)
• Layers of the ISA-95 framework — Siemens
• The difference between MES and production management systems — Unifinity (Japanese-language article)
• Division of roles between ERP and MES — Layers Consulting (Japanese-language article)
• Smart manufacturing and MES — Hitachi Solutions (Japanese-language article)
• Insights on MES use in manufacturing — ABeam Consulting (Japanese-language article)
• Manufacturing Execution Systems market report — GII (Global Information, Inc.)
• Thailand Manufacturing PMI — Trading Economics (S&P Global)
• Thailand Economic Monitor February 2026: Advanced Green Manufacturing for Growth — World Bank
• Investment Promotion Policy for Automation and Robotics Industries — BOI Thailand (PDF, Japanese-language edition)