Production Management System Cost 2026 — Where Quotes Really Split

“We asked three vendors for a quote, and the cheapest one and the most expensive one were several times apart.” When a Japanese-affiliated factory in Thailand comes to us about production management system cost, the conversation begins with that sentence remarkably often. The gap is not a gap in license prices. It is not a gap in the size or the reputation of the vendors either. What creates the gap is whether or not the quotation includes four line items that barely move no matter which delivery model you choose. This article first sets out the published price benchmarks, then breaks a five year total down across three delivery models for a factory in Thailand, and shows in numbers exactly where quotations diverge.

Production management system cost benchmarks — price ranges by delivery model

Let us start with the price benchmarks that are already in the public domain. The figures below are the ones commonly cited for the Japanese domestic market, and we keep them in Japanese yen as published. Procurement prices in Thailand sit on a different labor cost structure and a different exchange rate, so a separate baht-denominated estimate follows further down.

Production management system price varies widely by delivery model. The usual split is three ways: cloud, on-premise or packaged, and custom build.

Benchmarks for the Japanese market, based on the sources listed at the end. Amounts in JPY.

If your reaction to that table is that the ranges are so wide they are not really benchmarks at all, that reaction is correct. The on-premise range of JPY 1 million to over 10 million spans an order of magnitude between its lower and upper bound, and the custom build range of JPY 5 million to several hundred million spans two orders of magnitude. They are presented as benchmarks, but as raw material for building your own budget they are not usable as they stand.

Why do the ranges spread this far? The short answer is that these numbers describe the price tag on the software and the infrastructure and nothing else. Buy the same software and the cost of getting it into a state where your own operation can actually use it still varies enormously from factory to factory. Published production management software pricing answers the question of what it costs to buy. It does not answer the question of what it costs to get running.

Cloud pricing

Cloud is quoted at an initial cost of JPY 0 to 1 million and a monthly fee of JPY 30,000 to 150,000, with an implementation period of 1 to 3 months.

Initial cost can fall to zero because there is no server, operating system or database to procure. You load your own master data into an environment the vendor has already prepared and start using it. Monthly charges are predominantly per user, and the width of the range comes from the number of users and the range of functions in scope.

The point to watch on cost is that a low initial figure does not translate into a low total. The monthly fee keeps accumulating for as long as you keep using the system. A low initial cost is genuinely favorable from a cash flow point of view, but the total over five or seven years is a separate question. We take that question up again in the company estimate below, in concrete baht figures.

On the choice between cloud and on-premise itself, our comparison of cloud and on-premise deployment works through the non-cost angles as well, resilience against a network outage, where the data physically sits, and how much freedom you have to customize.

On-premise pricing

On-premise, or packaged, deployment is quoted at an initial cost of JPY 1 million to over 10 million for license and infrastructure combined, annual maintenance of around 5 to 15% of the implementation cost, and an implementation period of 3 to 6 months.

There is one point here that will catch your eye every time you read a quotation carefully. It is the maintenance percentage, and what that percentage is applied to.

Separately from the 5 to 15% of implementation cost cited above, other published material puts the software license itself at JPY 500,000 to 5 million and maintenance at 15 to 20% of the license fee. Compare the rates alone and 15 to 20% looks like the more expensive one, but the denominators are different. Is the 5 to 15% applied to the whole implementation cost, license plus infrastructure plus rollout support, or is the 15 to 20% applied to the license alone? For the same system, the annual maintenance amount changes depending on which reading applies.

When you compare competing quotations, our suggestion is to line up the annual maintenance amount in money before you line up the maintenance percentage. A percentage cannot be compared unless the denominator is visible. On how to take a maintenance line apart and read it, our breakdown of maintenance cost for factory business systems sets out how to move the discussion from percentages to amounts.

Custom build pricing

Custom development is quoted at JPY 5 million to several hundred million, with a development period of several months to over a year.

That range is less a benchmark than a spread of examples showing that projects of each size exist. The cost of a custom build is close to proportional to the volume of requirements. What sets the amount, therefore, is not the day rate of the development house but how much your own organization puts on the requirements list.

Where custom development comes onto the table for a Japanese-affiliated factory in Thailand, it is usually because the existing production model does not fit the standard functions of a package. An unusual order intake pattern, report formats mandated by the Japanese head office, tax and import and export documents specific to Thailand. When you try to absorb all of that, the judgment that a custom build is more straightforward than heavily customizing a package can genuinely hold up. What does need to be understood is that the moment you make that judgment, the weight of responsibility on the side that issues the requirements increases sharply.

Three hidden costs that never appear in the benchmark table

Alongside the model-by-model benchmarks, three costs are commonly flagged as easy to overlook.

• The double cost of the parallel run. For as long as the old and the new system run side by side, the same production results get entered twice. The overtime and the load on the shop floor during that period do not appear on the system quotation.
• Post-go-live customization. Once the system is live, requests appear without fail, along the lines of “this report is missing” or “we want one more field here”. Requirements definition before go-live cannot reduce that to zero.
• The cost of getting your data out at termination. This bites hardest on cloud. In what format, and at what price, can you receive your own data when the contract ends? Sometimes the contract does not say.

None of these three is a price for the system itself. They are the costs of starting to use it, continuing to use it, and stopping using it. That is precisely the territory this article deals with.

One clarification before we go further. This is not a product comparison article. For a comparison on the basis of which product suits your operation, see our 2026 comparison of production management systems. What this article deals with is narrower, the question that always arrives after you have chosen a product, or before you have, of how to read the line items in a quotation.

Why quotes split at the same scale — production management system cost has five layers

The same factory issues the same requirements, and the quotations that come back diverge widely. The quickest way to understand that phenomenon is to split production management system cost into five layers.

Layers 1 and 2 appear on every vendor’s quotation. They are the price tag of the product and the effort to configure it. For the vendor they are revenue, so there is no reason to leave them out.

What diverges is layers 3, 4 and 5. These three share a common property. The work extends onto the customer side as well.

Master data cleanup ends with your organization deciding what your own item master contains. On external interfaces, the only party that can supply the specification of the system at the other end is you. For acceptance testing, the people who write the test cases and decide pass or fail are your own business departments. Layers 3, 4 and 5 are territory where the boundary between the work the vendor does and the work you do is blurred, and that is exactly why they tend to drop out at the quotation stage.

Costs that drop out do not disappear. Once the project starts, they reappear somewhere without fail, as a change order or as your own overtime.

A “cheap quote” is not cheap, it is short

When an unusually cheap proposal turns up in a competitive bid, what is usually happening is not a discount. The scope of the quotation is short. Put a quotation that covers only layers 1 and 2 next to one that covers all five, and of course the first one looks cheaper.

That structure surfaces late in the project as a budget overrun. On failure statistics for ERP projects, secondary sources citing the 2026 ERP report from Panorama Consulting Group put the share of ERP projects in discrete manufacturing that fall short of their objectives at 73%, against an industry average of 68%. A production management system is either part of ERP or adjacent to it, so the tendency is a useful reference.

The same material also cites budget overrun figures such as 189% and 215%, but because the source is not clear about whether the figure is expressed as a share of budget or as the excess over budget, we do not treat those as multipliers in this article. We would suggest that readers do the same, and check the definition of the denominator whenever a number of this kind appears. Whether it means a percentage of the budget or a percentage by which the budget was exceeded changes the meaning considerably.

What sits behind an overrun cannot be explained by cost alone. On the organizational factors, our article on the five fault lines in failed production management system projects works through that side, and is worth reading alongside this one.

Breaking down the five year total across three models — our estimate

From here we present an estimate for a specific, concrete factory. Everything that follows is our own company estimate, and actual amounts move with specification, vendor and exchange rate. Please read it as a framework for thinking. All amounts are in THB.

Assumptions

• A Japanese-affiliated factory near Bangkok, Thailand, 120 employees
• 25 users of the production management system, 1 site
• 3 external interfaces
• 8,000 item master records, 600 customer and supplier master records, 3,500 BOM lines
• Evaluation period of 5 years

The three models

• Method A is cloud, a SaaS product used mainly with standard functionality.
• Method B is a package deployed on premise.
• Method C is a custom build by a local vendor.

Initial cost breakdown

Unit: THB. Company estimate.

Interfaces are costed at 180,000 each for A, 200,000 for B and 160,000 for C, and three of them are included. The unit price differs by model because of the presence or absence of a standard connector and the difference in the development setup. The 160,000 for parallel running is carried at the same amount in all three models, as the overtime equivalent for the double entry period.

Recurring cost and later rework

Unit: THB. Company estimate.

Separately from the table, we carry rework over the five years at 300,000 for A, 500,000 for B and 900,000 for C. These are treated as one-off change costs and are not included in the recurring totals.

The license under Method B is assumed to be a site license that does not depend on the number of users. That assumption does the work later on, in the section on the crossover point by user count.

Five year total

Unit: THB. Company estimate.

Under these assumptions the five year total comes to 4,750,000 for Method A, 6,560,000 for Method B and 9,000,000 for Method C. The ranking is A < B < C on initial cost and unchanged on the five year total. At 25 users, cloud has the advantage on cost.

That conclusion, though, leans heavily on the assumptions. The next two sections take apart which assumptions are doing the work.

The THB 1,300,000 that does not move whichever model you choose

The core of this article starts here.

Look down the three breakdown tables above one more time. License, server, development cost. These can be zero under one model and THB 3,200,000 under another. It is natural that the debate about which model to choose concentrates on those rows.

There are, however, four rows that barely move whichever model you choose.

Unit: THB. Company estimate.

The four items add up to 1,300,000 for A, 1,400,000 for B and 1,320,000 for C. All three models land inside a band of THB 1,300,000 to 1,400,000.

That is the central proposition of this article. Quotations diverge according to whether or not this band has been written into them. Not according to license price, and not according to the size of the vendor.

Why they do not move

There is a clear reason in each case why the four items do not depend on the delivery model.

Master data cleanup and migration comes to 380,000 in all three models because the object of the work is your own data. 8,000 item master records, 600 customer and supplier records, 3,500 BOM lines. Those volumes do not change whether you pick cloud or a custom build. Consolidating duplicated item codes, dropping obsolete parts, standardizing units of measure, tidying up the BOM hierarchy. The volume of that work is the same regardless of what kind of system is waiting to receive it.

External interfaces move within a range of 160,000 to 200,000 per connection, or 480,000 to 600,000 for three, but the movement is small. What sets the effort on an interface is not the system on your side but the specification of the system at the other end. The journal format of the accounting system, the production result feed to the Japanese head office, inventory synchronization with the warehouse system. As long as the other party does not change, the effort to reconcile with it does not change much either.

Training and acceptance testing runs from 220,000 to 300,000. Writing test cases, getting the business departments to actually use the system, raising defects and fixing them, then judging pass or fail. The volume of that work is set by the number of user departments and the complexity of the operation, not by the implementation model.

Parallel running is 160,000 in all three models. It is the double entry that occurs while the old and the new system run at the same time, plus the overtime equivalent that goes with it. That too is unrelated to the implementation model.

What share of initial cost the four items take

Look at the share those same four items take of initial cost and the difference in how each model presents itself becomes obvious.

Unit: THB. Company estimate.

Under Method A, 74.3% of the initial cost is these four items. Cloud is described as having a low initial cost, and that low figure comes from not needing a license or infrastructure. It does not come from the four items disappearing. If anything, because everything else is cheaper, the four items loom relatively larger.

This is where the practical trap sits. When a cloud proposal states an initial cost of THB 450,000 and nothing more, that figure covers only the configuration and fit and gap row of the table. In reality THB 1,300,000 sits on top of it. The 450,000 presented is only about a quarter of the 1,750,000 initial total. A gap of roughly 3.9 times is created in exactly that way.

The license is not the main character in the total

Now the same point from another angle, the share that the license or subscription itself takes of the five year total.

Unit: THB. Company estimate.

Under Method B, the license is only 24.4% of the five year total. Even so, it is not unusual for the largest share of time in the vendor selection meeting to go into negotiating a discount on that 24.4%. The contract then proceeds with the remaining line items undiscussed, and shortly before go-live the parallel run overtime and the state of the master data turn into problems. In our experience that sequence repeats itself often.

Method A shows the highest ratio at 56.8%, but that is the result of accumulation over five years. The price itself, THB 1,800 per user per month and THB 45,000 a month for 25 users, looks small in an approval paper. What that becomes once it is multiplied out over the term needs to be confirmed before signature.

The crossover at 42 users — choosing a production management system on cost

The estimate above stands on an assumption of 25 users. If the assumption changes the conclusion changes with it, so it is worth pinning down algebraically where the change happens.

The five year total for Method A rises in proportion to the number of users.

Five year total for Method A = 2,050,000 + 108,000 × number of users

The 2,050,000 is the initial cost of 1,750,000 plus 300,000 of rework over five years. The 108,000 is five years of subscription for one user, THB 1,800 × 60 months.

Note that initial cost is held constant here regardless of user count. In reality training and acceptance testing grow with headcount, so the crossover would come at a lower user count than the one derived here.

The license under Method B, on the other hand, is a site license that does not depend on user count, so the five year total stays flat at 6,560,000.

Here is where the two lines cross.

Unit: THB. Company estimate.

The crossover is at 42 users. Up to 41 users cloud is cheaper, and from 42 users the on-premise package is cheaper.

Two practical notes on how to use that number.

First, judge on the number of users five years from now, not the number today. You may go live with 25, but add production engineering and quality assurance the following year, then gain more people again when a second plant starts up, and 42 comes within realistic range. Conversely, if there is no clear plan to add users, the advantage of cloud continues.

Second, the number 42 moves when the assumptions move. If the unit price is THB 1,500 per user per month rather than THB 1,800, the crossover shifts to the right, and if the Method B license is charged per user, the comparison does not hold in this form at all. What matters is not the number 42 itself but the structure behind it, that cloud is a rising straight line and a site license is a horizontal one, and the two must cross somewhere.

When a quotation arrives, try drawing those two lines with your own conditions. If the intersection sits to the right of your expected user count, cloud has the cost advantage. If it sits to the left, the package does.

Four numbers to check on the quotation

Now to turn all of the above into what to check when a production management system quotation lands on your desk. There are four numbers. A quotation that does not state these four is, on the safe reading, a quotation in which the amount is not yet fixed.

1. The number of external interfaces

This is where the amount moves most. The estimate above assumes 3 interfaces, so here is what happens when that becomes 9, an increase of 6.

Unit: THB. Company estimate.

All three models increase by around THB 1,000,000. And connections get added in a requirements workshop through the lightest of remarks. “While we are at it, we would also like to pull production results from the PLC.” “Could we connect the handheld inventory scanners too?” Three or four remarks of that kind stack up and the five year total moves by THB 1,000,000. That is the sense of scale we would suggest bringing into the requirements workshop.

Rather than a single line reading external interfaces as a lump sum, we recommend asking for the name of each connection and the total count to be stated. The name of the system at the other end, the direction, outbound, inbound or bidirectional, and the frequency, real time or daily. With those three recorded, the basis for the increment is clear if the count grows later.

Connection to a warehouse management system, or WMS, is one of the most frequently added interfaces on production management projects. The same cost structure arises on the WMS side too, so if you are evaluating both at the same time, checking our breakdown of WMS cost alongside this article helps avoid both double counting and gaps on the interface portion.

2. The record count and the quality of the master data to be migrated

This estimate assumes 8,000 item master records, 600 customer and supplier master records and 3,500 BOM lines, and puts master data cleanup and migration at 380,000.

What to check in the quotation is the assumption about quality, not only the record count. “Migrate 8,000 records” is one statement. “Migrate 8,000 records, resolve duplicates and obsolete parts, and standardize units of measure” is a completely different volume of work. The first is transcription. The second is a stocktake of the operation.

In real factories it is not unusual for an item master said to hold 8,000 records to have only a fraction actually used in the past year, with the rest made up of one-off items from the past and duplicate registrations. Migrate that without tidying it and search in the new system becomes unusable, and the work gets redone after go-live anyway. How many items are actually alive is a number you should count in house before you request a migration quote.

Some quotations state as an assumption that the customer will supply cleansed master data. Where that is the case, it means work equivalent to the 380,000 remains on your side. It has left the vendor’s quotation, but the cost has not become zero.

3. The number of acceptance test cases

Training and acceptance testing is carried at 220,000 to 300,000 depending on the model. Two things are worth checking here, who writes the test cases and how many cases are assumed.

Acceptance testing is not a step that confirms whether the system behaves as specified. It is a step that confirms whether your operation runs. The test cases therefore need to be written by your own people, who know the operation. Test cases written by the vendor can only confirm what is written in the specification.

Where the quotation says acceptance test support as a lump sum and nothing else, ask what the support consists of. Does it stop at providing a test case template, does it include attendance during test execution, does it include the effort to fix defects? Proceed with that unclear and change orders appear during the test period.

4. The duration of the parallel run

This is the item carried at 160,000 in all three models. The amount looks small, but it grows in proportion when the period is extended.

A parallel run is a period in which the same production results are entered twice, into the old system and the new one. For the operators on the floor the work simply doubles, so the longer it drags on, the more input quality drops, which then creates a new problem, that the data in the new system does not agree.

Check how many months of parallel running the quotation assumes. One month or three months. And check whose budget carries the overtime for that period. Is it in the system budget or in the manufacturing department budget? Leave that undecided and an interdepartmental negotiation lands on you just before go-live.

Extra costs that come with a factory in Thailand — localization, local support, time difference

The estimate above already assumes a Japanese-affiliated factory in Thailand, and factory system implementation cost here carries items that are structurally added compared with an implementation in Japan. Here are three.

Multilingual operation

The screens of a production management system may need to be in Japanese for Japanese managers, in Thai for Thai staff, and in English for head office reporting. Packaged products with standard Thai UI support are limited, and even where support exists, field labels and master data display names have to be registered by you.

What bites in practice is not the UI itself but multilingual master data. Register item names in Japanese and the Thai operators on the floor cannot read them. Register them in Thai only and translation is required when reporting to the Japanese head office. Carrying Japanese, Thai and English across an item master of 8,000 records increases the volume of master data work accordingly.

Which languages you hold, and how far, is something to settle at the requirements definition stage, because changing it after go-live means revisiting every record.

Local support

When the system stops, who responds, and within how many hours? The options here are narrower than they are in Japan.

Where a package from a Japanese vendor is used in Thailand, the usual arrangement is first line support from a local partner with second line support in Japan. There is nothing wrong with that arrangement in itself, but it is worth checking whether the local partner’s cost is inside the scope of the maintenance contract. In some cases a separate support contract with the local partner is required on top of the maintenance contract with the Japanese vendor.

The level of support cost is influenced by the labor cost structure on the Thai side. For reference, the minimum wage in Thailand as of 2026 is THB 337 to 400 per day, with a national average of about THB 374. It is set province by province as a rule, and the upper band covers Bangkok, Phuket, Chonburi, Rayong and Chachoengsao provinces together with Koh Samui district in Surat Thani. In addition, the employer social security contribution rose from THB 750 to THB 875 per month in January 2026. That is an increase of THB 125 per person per month, or THB 1,500 per year.

None of that is directly a system cost, but these are the numbers sitting behind the parallel run overtime carried in the estimate above and behind the labor cost of local support staff. As long as labor costs are trending upward, an operation that fills gaps with manual effort gets less favorable as time passes. That is a premise worth holding in mind when the decision is to keep initial cost down and cover the difference in operations.

Time difference and support hours

The time difference between Japan and Thailand is 2 hours. It sounds small, and it tells on support desk hours. When a factory in Thailand starts at 08:00, the desk in Japan reads 10:00. At the end of the day, 17:00 in Thailand is 19:00 in Japan, and the Japanese desk is closing.

Trouble with a production management system clusters immediately after the start of the shift, when the previous day’s results have not come through, and just before the end of it, when the day’s results cannot be closed. In other words, incidents cluster in exactly the hours where the time difference tells most.

Whether you have a support desk that can respond in local time feeds directly into the maintenance amount. When you compare quotations, we recommend asking for support hours to be written in Thai time. A quotation that says weekdays 09:00 to 18:00 and nothing more describes a materially different service level depending on whether that is Japan time or Thailand time.

Investment climate and the supply of IT talent

Since this has been a long run of cost figures, here is one number from the wider environment. Investment applications to the Thailand Board of Investment, or BOI, in the first half of 2026 reached THB 1.473 trillion, up 37% year on year. Of that, the digital industry accounted for THB 1.115 trillion.

It would be premature, though, to read that figure straight as a tailwind. Most of the digital industry application value is investment in data centers and cloud, and it does not directly increase the pool of people who implement production management systems. If anything, supply and demand for IT talent is tightening, which tells on the man-month rates in a quotation. When you consider implementation timing and securing a partner, it may be worth keeping that supply and demand picture in view.

The implementation flow and how long it takes

The phase structure of a production management system implementation is broadly common across delivery models.

1. Current state review and requirements definition

Write out the current flow from order intake to shipment process by process, and decide what the new system will handle. What to settle at this stage is threefold: scope, meaning which processes; target, meaning which product lines come first; and what is explicitly out of scope this time. The third one matters most.

2. Product selection and choice of delivery model

Decide whether cloud, package or custom build answers the requirements. The band of four items above, and the crossover point by user count, are the material for that production management system selection.

3. Fit and gap and configuration

Identify the differences between the standard functions of the package and your operation, and decide one by one whether to close each gap by configuration, by changing the way you work, or by customization. Add customization here and both maintenance cost and later rework grow.

4. Master data cleanup and migration

As described above, this is the area where the volume of work does not change with the delivery model. In factories with high record counts, this phase sets the schedule for the whole project. Starting it in parallel with requirements definition is the standard approach.

5. Building and testing the external interfaces

Coordination with the owners of the systems at the other end is involved, so progress is not yours to control alone. The more connections, the higher the schedule risk.

6. Training and acceptance testing

The business departments write the test cases, use the system for real, and judge pass or fail. Unless time to fix the defects raised here is built into the schedule, the go-live date moves.

7. Parallel running and go-live

Run the old and the new system together, confirm that the numbers agree, then stop the old one.

Typical implementation duration

The implementation durations assumed in our company estimate are as follows.

Company estimate.

The Japanese market benchmarks quoted at the top of this article give 1 to 3 months for cloud, 3 to 6 months for on-premise and several months to over a year for a custom build. The durations in our estimate run somewhat longer because they include 3 external interfaces and the cleanup of 8,000 item master records and 3,500 BOM lines.

Here too, what stretches the schedule is not the product but the four items. Master data cleanup, external interfaces, acceptance testing, parallel running. None of the four can be shortened by changing the delivery model. If you want to compress the implementation period, bringing the start date forward is the only lever. Master data cleanup in particular is work that can begin before the product has been chosen.

Can you use a subsidy — the scope of the Digitalization and AI Adoption Subsidy 2026

Many people considering production management system cost also look at subsidies. Here is a summary of the Japanese scheme.

For fiscal 2026, the former IT Introduction Subsidy is running as the Digitalization and AI Adoption Subsidy 2026. The standard track is as follows.

It does not apply to a Thai local entity

This is the most important point in this section.

The subsidy is available to small and medium sized enterprises and small businesses in Japan. Costs incurred by a Thai local entity implementing locally are outside its scope.

In practice this is a point that needs confirming repeatedly. On a project where the Japanese head office leads a renewal of the production management system across the whole group, it is entirely normal to end up with the Japanese plants inside the subsidy scope and the Thai plant outside it. If that premise is left out when the overall project budget is built, the procurement plan slips later by several million yen.

We are also often asked what happens if the Japanese head office bears the cost of an implementation at the Thai plant. The purpose of the scheme is to support the digitalization of small and medium sized enterprises inside Japan, so where the implementation site is outside Japan, the reasonable reading is that it would not be accepted as an eligible cost. Individual cases require checking the published application guidelines and confirming with the secretariat, but planning on the premise that the Thai side of the budget is built without a subsidy causes fewer problems later.

What is available on the Thai side

Does that mean there is nothing at all on the Thai side? Not quite. Board of Investment incentives include support measures relating to the adoption of digital technology. The conditions for applying them, however, are divided finely by industry and by the nature of the investment, so whether a production management system implementation on its own qualifies for an incentive needs checking case by case.

As noted above, investment applications to the Thailand Board of Investment in the first half of 2026 reached THB 1.473 trillion, of which the digital industry accounted for THB 1.115 trillion, so the policy push behind the digital field is strong. If your plant already holds a BOI promotion certificate, it is worth checking whether any part of the work can be handled within your existing incentives.

Either way, we would suggest not placing subsidies or incentives at the center of the cost plan. Whether an application is accepted is not settled in advance, and although eligible costs cover not only software purchase and cloud subscription but also implementation-related items such as configuration and training, costs that a Thai local entity spends in Thailand are outside the scope in the first place. The band of THB 1,300,000 to 1,400,000 arises whether or not a subsidy exists.

Frequently asked questions

How much does a production management system cost?

It differs greatly by delivery model. The benchmarks cited for the Japanese market are an initial cost of JPY 0 to 1 million and JPY 30,000 to 150,000 per month for cloud, an initial cost of JPY 1 million to over 10 million with annual maintenance around 5 to 15% of the implementation cost for on-premise or packaged deployment, and a development cost of JPY 5 million to several hundred million for a custom build.

In our company estimate for a Japanese-affiliated factory in Thailand, with 120 employees, 25 users, 3 external interfaces and an item master of 8,000 records, the five year total came to THB 4,750,000 for cloud, THB 6,560,000 for a package and THB 9,000,000 for a custom build.

What appears in a benchmark table, though, is the price of the software, and the actual total has the cost of the implementation work on top of it. In our estimate, master data cleanup and migration, external interfaces, training and acceptance testing, and parallel running come to THB 1,300,000 to 1,400,000 on their own in all three models.

What determines the price of a production management system?

Not the license price. It is closer to reality to treat the price as being set by four things: the number of external systems to be connected; the record count and the quality of the master data to be migrated; the number of acceptance test cases; and the duration of the parallel run.

In our company estimate, the license or subscription share of the five year total came to 56.8% for cloud and 24.4% for a package, while for a custom build the development cost accounted for 35.6%. For a package, the license, the item that carries the price tag, accounts for only 24.4% of the total.

The amount that moves most easily is the number of external interfaces. In our estimate, increasing them from 3 to 9 takes the five year total to THB 5,830,000 for cloud, THB 7,760,000 for a package and THB 9,960,000 for a custom build, an increase of around THB 1,000,000 in all three models.

How long does a production management system implementation take?

The benchmarks for the Japanese market give 1 to 3 months for cloud, 3 to 6 months for on-premise or packaged deployment, and several months to over a year for a custom build.

Under the assumptions of our company estimate, which include 3 external interfaces and the cleanup of 8,000 item master records and 3,500 BOM lines, we put cloud at 3 to 4 months, a package at 6 to 9 months and a custom build at 9 to 14 months.

What drives the duration is not the product but the same four items, master data cleanup, external interfaces, acceptance testing and parallel running. Because none of the four can be shortened by changing the delivery model, where you want to compress the schedule, the realistic approach is to start master data cleanup without waiting for product selection.

What are the steps in a production management system implementation?

It divides broadly into seven phases: current state review and requirements definition; product selection and choice of delivery model; fit and gap and configuration; master data cleanup and migration; building and testing the external interfaces; training and acceptance testing; and parallel running and go-live.

Of these, the most important thing at the requirements definition stage is deciding what is out of scope this time. Collect requirements without settling the scope and the gaps grow at fit and gap, customization increases, and maintenance cost and later rework stay with you for years.

Master data cleanup is also the one phase that can start before the product has been chosen. Resolving duplicate items and taking stock of obsolete parts is work that is needed whichever system you select, so running it in parallel during the evaluation period creates room in the schedule.

Can Japanese subsidies be used for a production management system at a Thai factory?

No. The Digitalization and AI Adoption Subsidy 2026, formerly the IT Introduction Subsidy, is available to small and medium sized enterprises and small businesses in Japan, and costs incurred by a Thai local entity implementing locally are outside its scope.

Even on a group-wide project led by the Japanese head office, the safe way to build the budget is to treat the Thai implementation cost as ineligible. Plan on the basis that only the Japanese plant portion falls inside the scope.

On the Thai side, Board of Investment incentives include support measures relating to the adoption of digital technology, but the conditions for applying them are divided finely by industry and by the nature of the investment. If your plant already holds a BOI promotion certificate, it is worth checking whether any part of the work can be handled within your existing incentives.

Summary

Here are the key points of this article on production management system cost.

What appears in a benchmark table is the price tag on the software and the infrastructure. The Japanese market benchmarks are an initial cost of JPY 0 to 1 million and JPY 30,000 to 150,000 per month for cloud, an initial cost of JPY 1 million to over 10 million with annual maintenance around 5 to 15% of the implementation cost for on-premise, and JPY 5 million to several hundred million for a custom build. The ranges open by one to two orders of magnitude because these are price tags and do not include the cost of getting to a usable state.

Cost divides into five layers, and the lower three are where quotations diverge. Layer 1, license or subscription, and layer 2, rollout support and fit and gap, appear on every quotation. What diverges is layer 3, master data cleanup and migration, layer 4, external interfaces, and layer 5, training, acceptance testing and parallel running. Because those three mix with work on the customer side, they drop out of quotations easily, and dropping out does not make them disappear.

The four items do not move whichever delivery model you choose. In our company estimate, master data cleanup and migration, external interfaces, training and acceptance testing, and parallel running total THB 1,300,000 for cloud, THB 1,400,000 for a package and THB 1,320,000 for a custom build. All three land inside a band of THB 1,300,000 to 1,400,000. Their share of initial cost is 74.3% for cloud, 32.4% for a package and 25.3% for a custom build. Choosing cloud does not make the band go away.

The license is not the main character in the total. As a share of the five year total, the cloud subscription is 56.8%, the package license is 24.4%, and for a custom build the development cost is 35.6%. The five year totals were THB 4,750,000 for cloud, THB 6,560,000 for a package and THB 9,000,000 for a custom build.

The crossover is at 42 users. The five year total for Method A is expressed as 2,050,000 + 108,000 × number of users, while Method B on a site license stays flat at 6,560,000. At 41 users A comes to 6,478,000 and stays below B, and at 42 users it reaches 6,586,000 and passes above B. What you should judge on is not the user count today but the user count five years out.

Check four numbers on the quotation. The number of external interfaces, the record count and the quality of the master data to be migrated, the number of acceptance test cases, and the duration of the parallel run. Increasing the interfaces from 3 to 9 alone raises the five year total by around THB 1,000,000 in all three models.

Japanese subsidies do not reach the Thai side of the budget. The Digitalization and AI Adoption Subsidy 2026 is available to small and medium sized enterprises and small businesses in Japan, and costs incurred by a Thai local entity implementing locally are outside its scope. The band of THB 1,300,000 to 1,400,000 arises whether or not a subsidy exists.

TOMAS TECH supports the factory IT and FA domain for Japanese-affiliated manufacturers from our base in Bangkok. On production management systems as well, you are welcome to come to us at the stage before a product has been chosen, with the questions that have not yet turned into a requirements document. How many of your item master records are actually alive, how many systems really need to be interfaced, and roughly where the user count is likely to land five years from now. If you have a list of your existing systems and something that shows the current flow from order intake to shipment, we can go as far as working out with you how to read the quotations you receive, so even at an early stage of evaluation, feel free to get in touch through our contact page.

Sources

• OSK, A detailed guide to production management system cost benchmarks, in Japanese — https://www.kk-osk.co.jp/promotion/it_solutions/guide022_seisan.html
• BOXIL, Price benchmarks for production management systems, in Japanese — https://boxil.jp/mag/a9525/
• Small and Medium Enterprise Agency of Japan, Application guidelines for the Digitalization and AI Adoption Subsidy 2026, in Japanese — https://www.chusho.meti.go.jp/koukai/hojyokin/kobo/2026/260310001.html
• Hojokin Portal, What the Digitalization and AI Adoption Subsidy is, fiscal 2026, in Japanese — https://hojyokin-portal.jp/columns/degital_ai_summary
• Godlan, ERP Implementation Failure Statistics 2026 Research — https://godlan.com/erp-implementation-failure-statistics/
• ERP Research, ERP Implementation Cost Breakdown 2026 — https://www.erpresearch.com/en-us/erp-implementation-cost-breakdown
• Thai Law Online, Thailand Minimum Wage 2026 337 to 400 Baht by Province — https://www.thailawonline.com/minimum-wage-in-thailand/
• Employer Records, Minimum Wage In Thailand 2026 Provincial Rates Employer Costs — https://employerrecords.com/eor-location/thailand/minimum-wage/
• The Nation Thailand, BOI says first-half investment tops B1.47tn — https://www.nationthailand.com/business/economy/40068948