Traceability System Build Cost: 4-Layer Model and 90-Day Plan

“Can You Tell Me Where This Lot Shipped Within Four Hours?”

The call comes from your customer’s quality department. “We found a defect in the parts you delivered last week. We need to know today where every product made from the same material lot went, and how many.” From the moment your floor starts moving to the moment an answer exists, how many hours does your plant need? A traceability system build is, at its core, the work of shortening that elapsed time from days to hours, and then to minutes. This article breaks down the cost structure, how to choose an identification method, and a 90-day sequence for getting a system running, all written for the realities of plants in Thailand and Vietnam.

What a Traceability System Build Actually Produces

When a plant asks us to “put traceability in,” the first thing we always align on is the definition: what has to be traceable, and how far. If you collect quotations while that is still vague, the numbers from different vendors will spread by a factor of three or more, and the comparison itself becomes meaningless.

Internal Traceability and Chain Traceability

Internal traceability means that, within your own four walls from receiving to shipping, you can follow which material lot passed through which machine, which operator, and which conditions, and which product lot it became. It is decided and funded entirely by the plant, so it is the main battlefield of any build project.

Chain traceability means lot information stays connected across the whole chain, from upstream suppliers to downstream customers and distribution. You cannot complete it alone: it requires agreeing label requirements with suppliers for incoming goods, and shipping label or electronic data formats with customers. In practice the only workable order is to solidify the internal picture first, then connect its entrance and exit to the chain. If you dress up the customer-required format while the internal chain is still broken, you end up in the most dangerous state of all: paperwork that looks correct with no underlying data.

Trace Back and Trace Forward

Trace back starts from a product or a defective part and works upstream to answer “what was this made from.” You use it to pin down the cause of a defect and to fix the suspect range of material lots. Trace forward starts from a material lot or an equipment abnormality and follows the flow downstream to answer “where did it go.” You use it to define the recall scope and decide which customers must be notified.

A common design bias on the floor is a strong back and a weak forward. There is a screen that shows the history of a single defective piece, but no single button that produces “the list of shipping destinations for every product that used this material lot.” Yet in a real recall, the second one is what you need in a hurry. When you write build requirements, state explicitly for both directions what the input is, what the output is, and within how many minutes it must appear.

Define “Traceable” With Numbers

To stop the discussion from staying abstract, there are three numbers we always ask the customer to fix early in design.

The “within 4 hours” figure is not set by any law. It is the level we use as a design target because it is what customer audits and initial-response situations realistically demand. Start by measuring where you are today. Pick one inquiry that actually happened in the past and simply ask the people involved how long the response took; in most plants the answer comes back as two to five working days. That number becomes the starting point of the project.

Why the Build Has Become Urgent in 2026

Traceability used to be a “nice to have.” Over the last two or three years it has been moving into the territory of “without it, business stops.” It is worth holding the external pressure in factual terms.

IATF 16949: Effectively Mandatory for Automotive Parts

Clause 8.5.2.1 of IATF 16949 requires traceability for automotive parts, and for safety-critical parts it requires tracking at serial level. The scope reaches beyond your own processes to parts originating from subcontractors and to quality evidence across the supply network. Record retention is normally aligned with “the service life of the vehicle plus regulatory requirements,” and 10 to 15 years is commonly cited. Customer-specific requirements (CSR) from an OEM may demand an even longer period.

What auditors check is equally concrete: a documented traceability map, an unbroken trace record for a sample part they select, the change management history of the system, and a demonstration that you can define a recall scope. In other words, they are not asking whether you bought a system; they are asking whether you can trace on the spot.

On the institutional side, IATF Rules 6th Edition took effect on 1 January 2025. A major revision of IATF 16949 itself is reported to be scheduled for late 2026 through early 2027. Since the content of that revision is not yet fixed, chasing individual clauses is less useful than getting into a state where you can produce a genealogy immediately for any sample the auditor points at. That state is an investment that survives any revision.

EU Battery Passport: The 18 February 2027 Deadline

Under Regulation (EU) 2023/1542, from 18 February 2027 a battery passport becomes mandatory for EV batteries, industrial batteries above 2 kWh, and LMT batteries placed on the EU market. The mandatory data set is reported as 111 items for LMT batteries and 98 items for EV traction batteries, made machine readable through a QR code and a URL conforming to GS1 Digital Link, and registered in the EU battery passport registry. Data is managed in a three-tier model: public, restricted, and authority-only.

The heavy part in practice is that a carbon footprint must be calculated and declared per battery model and per manufacturing plant. That demands linking lot-level production history to the plant’s energy consumption data. Batteries manufactured before February 2027 are also reported to fall in scope if they are imported into the EU on or after that date. For non-compliant products, refusal of customs clearance at the EU border, withdrawal from the market, and fines have all been flagged as possibilities.

Even more important is the positioning. The battery passport is treated as the leading case for the wider Digital Product Passport (DPP) under ESPR, and as the precedent for other categories such as textiles, electronics, and furniture. Even if your products have nothing to do with batteries, it is worth designing on the assumption that this exact requirement pattern, disclosing data at product level, will reach your category within a few years.

GS1 Sunrise 2027: The Move to 2D Codes

Sunrise 2027, the industry initiative led by GS1, aims for retail point of sale to be able to read both one-dimensional barcodes and GS1-compliant 2D codes (GS1 QR and GS1 DataMatrix) by the end of 2027. A 2D code can hold, in addition to the GTIN, the best-before date, batch number, serial number, and weight in a single code. Because the lot number sits inside the code, a recall can be narrowed to individual items, and that is the biggest significance when you look at it from the traceability side.

During the transition, dual labeling with both an EAN and a 2D code is recommended. There is no legal penalty, but it is pointed out that delaying the change increases pressure from retail partners and customers. If you supply consumer goods or food products, the rational move is to add the 2D code at the next moment you revise your label design. The specific design issues in a food plant are covered separately in our article on building traceability in food factories.

Vietnam’s New Decree: Decree 37/2026/ND-CP

For manufacturers with a site in Vietnam, the most concrete near-term deadline is already moving. Decree 37/2026/ND-CP, dated 23 January 2026, sets out rules for the management of product identification codes and traceability codes, and encourages the use of identification technologies such as QR codes, DataMatrix, RFID, and NFC.

According to reports, traceability became mandatory from 1 January 2026 for “high-risk goods” such as chemicals, industrial explosive precursors, and tobacco, while other categories are treated as recommended. Businesses are to register accounts, obtain identification codes, and verify product information from 1 July 2026, and 1 January 2027 is reported as the deadline for full compliance for goods entering distribution. The national portal is verigoods.vn, and reporting indicates that more than one million product codes had been certified as of the end of May 2026. Consumers can check, free of charge, the product name, images, the manufacturer or distributor, address, brand, manufacturing lot number, and serial number, plus the best-before date where applicable; for imported goods, importer information and authorized dealer information are also required.

Note that the details of dates and scope vary between sources. Whether your own products fall in scope must be confirmed with the competent authority or with local legal and regulatory specialists. What matters here is less the scope determination than a different question: if you do fall in scope, do you already hold, inside your own systems, a data structure that can publish a manufacturing lot number and a serial number at product level to an external party?

The Whole Picture: Breaking the Build Into Four Layers

The main reason quotations cannot be compared is that traceability is treated as one lump. We always break it into four layers. Each layer has different decisions to make, and the nature of its cost is different too.

Layer 1, Identify: What Gets an ID, and at What Granularity

The first layer is giving an ID to the thing you want to track. There are four decisions. First, the object: do you put an ID on materials, work in process, finished products, returnable boxes, jigs, or dies. Second, the granularity: lot level or serial level, or something in between such as a pseudo lot of “one shift times one machine.” Third, the code scheme: the digit structure, whether the code carries meaning or not, and whether you follow the GS1 standard or your own scheme. Fourth, the physical marking method: label, printing, stamping, or RFID tag.

The most expensive decision to reverse here is the code scheme. A part number stuffed with meaning is easy for humans to read, but the scheme collapses every time the product specification changes. As a rule we recommend a design where the ID is a meaningless sequential number plus a check digit, and all meaning lives in the database. The only exception is information the floor genuinely needs to read by eye, such as machine number and date, which should be printed in human-readable form alongside the code.

Layer 2, Capture: How the Data Gets Off the Floor

An ID that nobody reads and records leaves nothing behind. Capture methods fall into three broad groups: people read them (handheld terminals, tablets, fixed scanners), equipment supplies them automatically (PLCs, sequencers, measuring instruments, vision inspection machines), and records are matched afterwards (inspection reports, material mill sheets, delivery data from subcontractors).

The key to good design is keeping human input to a single action that follows the natural flow of work. Any design that asks an operator for more than 30 extra seconds will break down in a busy month. The benchmark we use when designing on the floor is to keep the added input to within 30 seconds per process step, and within 10 seconds where possible. The ideal is to strip it down to scanning two codes at the start of the step: the code on the workpiece and the code on the work instruction.

Layer 3, Link: The Lot Genealogy Database

This is the heart of the build. The captured events are connected into a genealogy. Technically, you build a structure that holds the many-to-many relationship between material lots and product lots through process events.

The easiest thing to get wrong is splitting and merging. One material lot divides into several product lots (split), several material lots mix into one product (merge), and the remainder of the previous lot carries into the head of the next one (carry-over). In processes handling liquids, powders, resins, or plating baths, there is no way around these three. Unless you accept, as a data structure and up front, the reality that the material in a tank is a mixture of the previous lot and the current one, you end up at the same conclusion later: the suspect range grows so wide that the system is useless.

At the same time, master data has to be put in order: the item master, the process master, the equipment master, the operator master, and the relationship between materials and items (BOM). Where these already exist in a production management system or ERP, the design should reference them rather than duplicate them. We cover the division of roles with an existing production management system or MES in our article on production management systems and MES for plants in Thailand, so if you already run such a system, read that one as well.

Layer 4, Retrieve: Trace Inquiry and Evidence Output

The final layer turns “traceable” into “usable.” Three outputs are needed. First, a trace inquiry screen with a tree view that expands both upstream and downstream, starting from a product ID, a lot number, or a shipping document number. Second, evidence output in PDF or Excel form that holds up in an audit or a customer submission, including timestamps, operator, equipment, inspection values, and change history. Third, scope extraction: a function that takes a material lot or a piece of equipment and exports a CSV list of the affected products and shipping destinations.

This layer is the one most often cut from a quotation, and cutting it removes the value. We have seen plant after plant where the data does accumulate in the database, but no inquiry can be answered without asking the IT department to write SQL. That does not shorten the turnaround at all. Put it in the requirements: a quality assurance staff member must be able to operate it personally and produce an answer within four hours.

Cost Breakdown: Estimate by Four Layers Times Initial and Running

Now to the cost, which is the point of this article. The ranges below are offered as something you can use for comparison, but every range carries assumptions. Change the assumptions and the numbers move.

Assumptions Behind the Estimate

The figures below are the indicative ranges we use when quoting this type of project for Japanese-affiliated manufacturing sites in Thailand and Vietnam. They are not the actual results of any specific customer. The assumptions are as follows.

• Plant size: 100 to 600 employees, 1 to 4 production lines
• Scope: internal traceability from receiving to shipping, retrofitted onto existing lines
• Identification: mainly thermal transfer labels carrying a 2D code
• Platform: cloud or in-plant server, with 1 to 2 interfaces to an existing production management system
• Exchange rate assumption: 1 USD is approximately 33 THB, and 1 THB is approximately 4.4 JPY (indicative, July 2026). Every figure in this article uses that conversion consistently

Cost Ranges by Layer, Initial and Annual

The total is the sum of the layer figures: 4,100 + 8,300 + 10,300 + 6,200 + 5,500 = 34,400 at the low end, and 17,200 + 41,300 + 48,200 + 27,500 + 24,100 = 158,300 at the high end. The annual running cost adds up the same way: 1,000 + 1,700 + 2,800 + 1,400 + 700 = 7,600, and 5,500 + 8,300 + 11,000 + 6,200 + 3,400 = 34,400. In baht that is roughly THB 1.14 million to THB 5.22 million for the initial build, and roughly THB 251,000 to THB 1.14 million per year to run.

Looked at as a five-year total cost of ownership, the low end is 34,400 + 7,600 times 5 = USD 72,400, and the high end is 158,300 + 34,400 times 5 = USD 330,300. Comparing initial cost alone can invert the ranking once running costs are included, so always line vendors up on a five-year TCO basis.

One note on Layer 1: the initial cost assumes you are adding a label printing and application step to an existing line. Equipment that is built into the line itself, such as a laser marker or an inkjet printer, is better treated as separate capital expenditure, and the range is wide depending on the method: roughly USD 5,500 to 55,100 per unit (inkjet USD 5,500 to 17,200, laser USD 17,200 to 55,100; see the comparison table of identification methods for details). It is not included in the table above.

What a Small Start Contains: The USD 34,400 Example

Here is what the low end of the range actually buys. It is a launch limited to one line and one product family, validation included.

In this configuration you read the code at four points on the target line: receiving, issue to the line, process completion, and shipping, and build the genealogy from those events. It is not the whole plant, so you cannot claim that everything is traceable, but it stands up as a model case to show in a customer audit, and it fixes the unit cost for rolling out to the remaining lines.

Running Costs That Get Overlooked

It is very common to focus the comparison on initial cost and sign a contract without ever pinning down the real running numbers. Three items in particular deserve hard figures up front.

The first is label and tag consumables. Even at THB 0.18 to 0.68 per label (roughly USD 0.006 to 0.02), labeling individual pieces at 250,000 units a month means THB 45,000 to 170,000 per month, or THB 540,000 to 2,040,000 per year, which is about USD 16,400 to 61,800. At the top end that single line item is roughly 1.8 times the upper limit of the annual running cost in the table above, so the choice between labeling every piece and stopping at box or tray level changes the cost structure itself.

The second is data retention. In an IATF 16949 context, 10 to 15 years of retention is commonly cited. Estimated against metered cloud pricing, that is not a trivial amount. Design the split between hot data (the most recent one to two years) and archive from the very beginning.

The third is retraining as people turn over. Operator turnover cannot be avoided in Thailand or Vietnam. Build SOP updates and new-hire training into the running cost as an annual workload.

An ROI Example With Every Input Disclosed

The following follows the format we use for an early-stage estimate, calculated on a fictional model plant. It is not the actual result of any real company.

Model plant assumptions

• Industry: Tier 2 automotive parts, 500 employees, 2 production lines (2 shifts each), 250 operating days a year
• Volume: 250,000 units a month, 3 million units shipped a year
• Major customers: 8 companies
• Labor cost: quality assurance and process engineering staff at THB 45,000 per month, multiplied by 1.3 for statutory benefits and similar, giving THB 58,500 per month. At 160 working hours a month, the hourly cost is THB 58,500 divided by 160 = THB 365.6, which we round to THB 366 per hour, or about USD 11.1 per hour. That rate is used throughout what follows

Investment assumptions

Slightly below the midpoint of the ranges above: THB 2,700,000 (about USD 81,800) initial and THB 450,000 (about USD 13,640) per year to run. The initial breakdown is Layer 1 300,000 + Layer 2 740,000 + Layer 3 880,000 + Layer 4 450,000 + common 330,000 = THB 2,700,000. The running breakdown is Layer 1 90,000 + Layer 2 113,000 + Layer 3 145,000 + Layer 4 68,000 + common 34,000 = THB 450,000 per year.

Estimated benefits

The arithmetic is as follows. Inquiries: 1,392 person-hours times THB 366 = THB 509,472. Audits: 280 person-hours times THB 366 = THB 102,480. Sorting: THB 366 per hour divided by 240 pieces per hour (at 15 seconds per piece) = THB 1.525, rounded up to THB 1.55 per piece, so 104,000 pieces times THB 1.55 = THB 161,200. Transcription: a difference of 24 minutes times 2 lines times 2 shifts times 250 days = 24,000 minutes = 400 person-hours, and 400 times THB 366 = THB 146,400. The total is THB 919,552, or about USD 27,870 at 33 THB to the dollar. USD figures in the table are rounded to the nearest 10.

Risk avoidance as an expected value

Labor savings alone will not make an investment of this size pay back quickly. In practice we put a value on being able to narrow the recall scope, expressed as an expected value. Assume the cost of one recall or special sorting event at a customer site is THB 6,800,000 (about USD 206,000), covering retrieval, sorting, transport, special shipments, and customer handling; assume traceability lets you narrow the scope to one fifth; and assume a 10 percent probability per year. The expected annual reduction is then THB 6,800,000 times (1 minus 1/5) times 0.10 = THB 544,000 per year, or about USD 16,480. All three of those numbers are assumptions. Replace them with figures from your own past events.

Payback period

On a three-year cumulative view, the cost is 2,700,000 + 450,000 times 3 = THB 4,050,000 (about USD 122,700). Labor savings alone give 919,552 times 3 = THB 2,758,656, leaving THB 1,291,344 unrecovered. Including risk avoidance gives 1,463,552 times 3 = THB 4,390,656, a surplus of THB 340,656. Over five years, against a cost of THB 4,950,000 (exactly USD 150,000 at this rate), labor savings alone reach THB 4,597,760 and the version including risk avoidance reaches THB 7,317,760.

The conclusion this table points to is clear. If you propose a traceability system build as a labor-saving investment, the payback period in most plants will exceed five years. The accurate way to frame the decision is to put conformity to customer requirements and limiting the damage in a crisis at the center, and explain labor savings as a by-product. That framing is also the one that passes internal approval more easily.

Choosing an Identification Method: Label, Print, Stamp, 2D Code, RFID

The identification method is an expensive decision to reverse. The ranges below are the benchmarks we use on projects; they move with the model and specification you choose.

There are three axes for the decision.

First, does the marking survive the process conditions? A label will not survive on a workpiece that passes through heat treatment, washing, painting, or cutting fluid. In that territory you have no choice but laser or stamping even at a higher equipment cost, and the freedom of choice is low.

Second, do you need bulk reading? The essential value of RFID is reading many items at once without contact. If all you want is to identify individual pieces, a 2D code wins overwhelmingly on unit cost. Conversely, if you want to unattend the movement of returnable boxes in and out of stores, or the stocktaking of work in process, a tag at THB 2.7 to 13.6 still pays for itself. We break down the cost structure and the conditions for payback in our article on the cost of RFID in factories; if RFID is under consideration, reading that one first will speed up the decision.

Third, conformity to customer requirements. If GS1 compliance is demanded, the code scheme is no longer yours to decide. GS1 DataMatrix and GS1 QR hold more information in the same area and align with the Sunrise 2027 direction. If you already plan to revise your labels, taking that opportunity to move toward GS1 compliance is the rational call.

As a practical recommendation, most plants get the best cost efficiency from a hybrid: thermal transfer labels with a 2D code at box or tray level, laser or stamping only on the safety-critical parts that require individual identification, and RFID on returnable boxes and jigs. Trying to cover everything with a single method always leads to over-investment somewhere or a functional gap somewhere else.

Designing So Shop-Floor Input Does Not Break Down

Builds fail for operational reasons, not technical ones. You can deliver a system that works, but if the floor does not enter data, the data has holes, and you cannot trace with holes in it.

Put Read Points at the Seams Between Steps

The principle for minimizing added input is to make the points where a record is needed coincide with the moments the operator already stops their hands. The start and end of a process step, a material change, an equipment changeover, a pass or fail judgment at inspection. Those are moments when the operator’s hands are already still, and inserting a scan there adds almost no perceived load. Insert an input in the middle of machining, on the other hand, and it will always be postponed, which leads to batches of entries made later, meaning the timestamps can no longer be trusted.

Increase the Share That People Do Not Enter

The ideal is to narrow what a person enters down to scanning an ID and making a pass or fail judgment, and to acquire everything else automatically. Equipment run and stop status, cycle time, processing conditions, and values such as temperature, pressure, and torque can be taken directly from a PLC or a measuring instrument. Making an operator copy numbers by hand costs the time to copy them and creates transcription errors.

Even picking up the signal from a stack light is enough to record equipment status and time reliably. You do not have to make the whole line an IoT installation: starting with just two signals, the stack light and the production counter, is highly cost-effective and lets you phase the investment in the capture layer. We deal with that step-by-step approach concretely in our article on IoT implementation in factories.

Retrofitting Existing Equipment

The assumption that “the machine is old, so we cannot get data” is usually wrong. Even equipment with no communication interface can give up its status through the following means.

One important caution: wiring into existing equipment or tapping signals from it can affect the machine maker’s warranty conditions. Start with non-contact methods (optical sensor, current sensor, camera), and where a voltage tap or intervention in the PLC is unavoidable, confirm with the equipment maker in advance.

Language and Making It Stick

On Thai and Vietnamese shop floors, the language design of screens and SOPs decides whether the system sticks. The practical minimum is that screens operators touch are in the local language, analysis screens for managers are in English or Japanese, and error messages are in the local language and say what to do next. A message that only reads “a system error has occurred” creates an operating pattern where the line stops and the supervisor is called.

Linking the 4M: Man, Machine, Material, Method

If traceability ends at “the lot numbers are connected,” it cannot be used to find the cause of a defect. To reach a cause, you have to link the 4M to each process event.

When this 4M linkage works, the quality of analysis changes. If in-process defects rise, for example, you can produce a three-factor intersection within minutes: the defect rate is high on machine B, using die C, and with material lots from supplier X. That is work which used to take days of matching a veteran’s memory against paper logbooks.

Version control of Method is especially effective. If you record the date and time of every change to a work standard or recipe, you can narrow the candidate causes simply by overlaying the timing of the defects on the change history. Without that record, you fall into the hardest state to resolve: defects have increased, but nobody knows since when. If you are building this as a 4M management system, put version control in the initial scope without exception. Adding it later is effectively impossible, because the earlier revisions no longer exist.

Five Patterns Where the Build Stumbles

From experience across many shop floors, the shapes of failure converge on about five.

1. Paper logbooks live on forever alongside the system

If you decide during the transition to “keep the paper just in case,” it usually continues for years. Double entry doubles the load on the floor, and eventually the accuracy of one side degrades. It is always the system side that degrades, because the supervisor checks the paper while nobody looks at the system. The countermeasures are to set a cut-over date and abolish the paper, and for paper you cannot abolish, such as customer-specified forms, to change it so the system prints it.

2. Lot granularity set too fine, or too coarse

One case: “if we are doing this at all, let us do it at serial level,” which triples the number of scans on the floor and leaves the operation a formality within six months. The opposite case: “one lot per day,” which makes the suspect range in an incident the entire day’s output, so the recall scope cannot be narrowed. Both happen often. The decision criterion is the size of the suspect range you can tolerate in a crisis. If your business cannot survive recalling a full day’s shipments, the granularity has to be finer than a day.

3. Missing design for splitting and merging of work in process

As noted above, splitting, merging, and carry-over are not problems only for liquids and powders. Batch furnaces processing several lots at once, shared plating baths, leftover material from several lots mixed and sent to the next step. If you push these outside the design as “exception handling,” exceptions turn out to be 30 percent of real operation, and that 30 percent becomes the hole you cannot trace. During requirements definition, walk every process step and list every place where a split or a merge occurs.

4. Nobody starts on master data

The item master does not match reality, discontinued items are still in it, the process master does not match the actual layout. Put a system on top of that and the data accumulates but the aggregates do not add up. Master data cleanup is unglamorous and hard to assign an owner to, but in our experience it accounts for 20 to 30 percent of the build effort. Explicitly reserve the period and the people for it in the project plan.

5. Audit requirements bolted on afterwards

After the system is built, you discover that the form the customer audit needs cannot be produced, or that no change management history was kept. This is the most painful rework of all. In an IATF 16949 context, what gets checked is the traceability map, an unbroken trace record, the change management history of the system, and a demonstration of defining a recall scope. During requirements definition, write out on paper a sample of every form you will present at the next audit, and work backwards to confirm that every field on it is actually captured as data.

Practical Issues in Thailand and Vietnam

A build in Japan and a build at an ASEAN site require attention to different things.

Language and Training

Multilingual screens, labels, and SOPs are mandatory. Beyond that, aim for a state where local staff alone can run the operation. A system that cannot answer an inquiry unless a Japanese expatriate is present gets reset every time the expatriate rotates out. Treat training cost not as a one-off at the start but as an annual running cost.

Staff Turnover

If your design links operator IDs into the system, you need an operating flow for updating the master data as people join and leave. Deleting a leaver’s ID breaks the historical records, so design it as a logical delete with an inactive flag. It sounds obvious, but we have seen a case built with a physical delete where, at audit time, the operator for a past lot could no longer be identified.

Thailand: IATF 16949 and TISI

In the Thai automotive parts industry, IATF 16949 certification is effectively mandatory if you are targeting OEMs or Tier 1 customers. Certain automotive parts also carry mandatory certification from TISI, the Thai Industrial Standards Institute. On the market environment, Thai automotive parts production is forecast to grow 1.5 to 2.5 percent a year from 2026 to 2028, OEM parts account for 80 to 85 percent of export value, and export value is projected to grow 1.0 to 2.0 percent a year on average. A structure that is export-led with a high OEM share is also, by definition, a structure in which customer requirements are the trigger for the build.

BOI investment incentives are worth examining too. Under the BOI’s “Smart and Sustainable Industry” measures, applications aimed at machinery upgrades, digital technology adoption, and automation or robot integration were reported at 132 cases worth about USD 507.6 million in the first half of 2026. Across all BOI approvals in the first half of 2026, more than 82,000 jobs are expected to be created. Eligibility and conditions differ case by case, however, so always confirm in advance whether a traceability-related investment qualifies.

Vietnam: Responding to the New Decree

For Decree 37/2026/ND-CP described earlier, check three things on the internal system side. First, that the manufacturing lot number and serial number are fixed at product level. Second, that data can be exported in a form submittable to the external portal, whether CSV or API. Third, if you handle imported goods, that importer information and authorized dealer information can be held. The national portal is reported as verigoods.vn, and reporting indicates more than one million product codes were certified as of the end of May 2026. Because dates and scope vary between sources, treat confirmation with the competent authority or a specialist as a precondition for deciding whether you are in scope.

Rehearse the Audit for Real

What a customer audit tests is not the existence of a system but a live demonstration. The auditor designates one sample and has you trace from it to the material lot, the operator, the equipment, the inspection values, and the shipping destination. Rehearse that sequence internally before the audit. The reason we always put an audit demonstration rehearsal in the final phase of a build project is that this is usually where a missing output field is discovered for the first time.

A 90-Day Roadmap

This is the execution plan that keeps a build from stalling. Each phase has deliverables and a decision gate, and the rule is that you do not proceed until the gate is passed.

Three points to add.

First, do not skip Phase 0. A project that never measured the current turnaround cannot claim any benefit once it is finished. Whether you can say “what used to take two to five working days now takes four hours” depends entirely on those 15 days in Phase 0.

Second, do not get greedy with the scope of Phase 2. Take one line and one product family and drive vertically from receiving to shipping. Spreading horizontally across a few steps on all lines does not complete a genealogy, so the value cannot be confirmed. Driving one vertical path through is the rule.

Third, what finishes in 90 days is the launch, not a company-wide rollout. Depending on the number of lines and product families, a realistic view is a further 6 to 18 months to complete the rollout. Because Phase 3 fixes the unit cost of expansion, however, the budget planning from that point onward becomes far more accurate.

Frequently Asked Questions

How much does a traceability system build cost?

Under the assumptions used in this article (100 to 600 employees, 1 to 4 lines, mainly labels with a 2D code, internal traceability), the indicative ranges are USD 34,400 to 158,300 initial and USD 7,600 to 34,400 a year to run. A small start limited to one line can be launched from about USD 34,400. On a five-year TCO basis that is USD 72,400 to 330,300. If you need line-integrated equipment such as a laser marker, add USD 17,200 to 55,100 per unit as separate capital expenditure.

What is the difference between a lot management system and an MES?

A lot management system focuses on retaining history at lot level so that things can be traced. An MES, or manufacturing execution system, is a broader framework covering the release of production instructions, progress management, actual data collection, quality management, and equipment management, with traceability included as one of its functions. In practice, building first with lot tracing as the only objective and expanding into other MES functions later gives you better control over both the investment and the schedule. In plants that already run an MES or a production management system, the shortest path is to build the genealogy database so that it references that existing data.

Which is better, RFID or a QR code?

They serve different purposes, so it is not a question of better or worse. If you only need to put an ID on individual pieces, a 2D code wins overwhelmingly on unit cost (THB 0.18 to 0.68 per label against THB 2.7 to 13.6 per RFID tag). Where you need to read many items at once without contact, however, such as returnable boxes moving in and out of stores, locating jigs and dies, or bulk stocktaking of work in process, only RFID can do it. For most plants the realistic answer is a hybrid: 2D codes on products and boxes, RFID on returnable containers and jigs.

What do the IATF 16949 traceability requirements actually demand?

Clause 8.5.2.1 requires traceability for automotive parts and requires serial-level tracking for safety-critical parts. Parts originating from subcontractors and quality evidence across the supply network are included in scope. At audit, what gets checked is reported to be a documented traceability map, an unbroken trace record for a sample part, the change management history of the system, and a demonstration of defining a recall scope. Record retention follows the principle of aligning with the service life of the vehicle plus regulatory requirements, and 10 to 15 years is common. Customer-specific requirements may demand a longer period, so check the CSR of your major customers first.

Can we get data from old existing equipment?

In most cases, yes. If there is a stack light, an optical sensor or a voltage tap gives you the run status; if there is a completion signal, a contact input gives you the production count. Where a PLC exists but is not connected to the network, adding a communication unit lets you read from it. Even with no signal at all, you can fill the gap with a camera and OCR, an estimate of running status from a current sensor, or manual entry on a tablet. The indicative added cost is USD 200 to 3,400 per machine. Because wiring or tapping signals can affect the equipment warranty, start with non-contact methods and confirm with the machine maker in advance where necessary.

How long does the build take?

For a launch on one line and one product family that runs through to the inquiry screen, 90 days is a realistic target (Phase 0 assessment 15 days, Phase 1 design 25 days, Phase 2 pilot 30 days, Phase 3 rollout preparation 20 days). Allow a further 6 to 18 months for a company-wide rollout across all lines. The single biggest factor affecting the schedule is the state of your master data; where the item, process, and equipment masters have drifted from reality, that cleanup alone can add one to two months.

Can we not just start with Excel?

Using Excel for the initial assessment and for thinking through granularity is genuinely useful. Continuing to run production operations on Excel is not recommended, for three reasons. First, holding a many-to-many genealogy (splits and merges) in tabular form makes the trace query a manual task, so the turnaround does not shorten. Second, no change history is kept, so you cannot show the change management evidence an audit requires. Third, concurrent editing and access control break down, and you lose the ability to assert that the data is valid. Use Excel as a design tool and put the operation on a database.

Summary

Break a traceability system build into four layers and both the estimate and the sequence become clear. Layer 1 Identify decides the IDs and the granularity; Layer 2 Capture collects data while minimizing the load on the floor; Layer 3 Link holds a genealogy that includes splits and merges; Layer 4 Retrieve creates a state where a quality assurance staff member can produce an answer alone within four hours. Under the assumptions in this article, the cost is USD 34,400 to 158,300 initial, USD 7,600 to 34,400 a year to run, and USD 72,400 to 330,300 over a five-year TCO.

Three things not to get wrong in the investment decision. First, proposing it on labor savings alone usually pushes the payback beyond five years; the real value lies in conformity to customer requirements and in limiting the damage when something goes wrong. Second, lot granularity should be derived backwards from the suspect range you can tolerate in a crisis, because operations break down whether it is too fine or too coarse. Third, audit requirements must be worked backwards from sample forms during requirements definition, and never bolted on later. As for sequence: measure the current turnaround in Phase 0, confirm the value with a pilot that drives vertically through one line, and only then expand horizontally. In our view that order carries the lowest risk of failure.

What you need at the entrance to a build is not a system selection but a decision about what will be traced and how far. If you are weighing where to set your granularity, how much data you can realistically get from existing equipment, or whether your current scope is enough for what your customers demand, we are happy to talk at the concept stage. TOMAS TECH is a Bangkok-based system integrator providing production management, IoT, and automation to manufacturers in Thailand and ASEAN, and we support the whole path from current-state assessment through design to launch on the floor. If you bring your process flow and the documents describing your customer requirements, we can complete the identification of untraceable points and present an indicative cost range within the first meeting. Feel free to reach us through this contact form.

References

• IATF 16949 traceability requirements, record retention, and audit checkpoints: https://tracio.com/compliance/iatf-16949
• Enhancing traceability under IATF 16949: https://www.qmii.com/iatf-16949-requirements-enhancing-traceability-in-automotive-manufacturing/
• Entry into force of IATF Rules 6th Edition and the outlook for the standard revision: https://www.smithers.com/resources/2026/april/iatf-16949-news-2025-rules-and-2026-changes
• EU battery passport requirements and data items: https://www.circularise.com/blogs/eu-battery-passport-regulation-requirements
• Timeline of EU Digital Product Passport deadlines: https://traceable.digital/resources/blog/eu-digital-product-passport-deadlines-the-complete-2025-2030-timeline/
• GS1 2D barcode standards and Sunrise 2027: https://thetraceabilityhub.com/gs1-2d-barcodes-standards-gs1-datamatrix-qr-code-sunrise-2027/
• Practical guide to GS1 Sunrise 2027: https://www.macsa.com/gs1-sunrise-2027-2d-barcodes-guide/
• Vietnam mandating traceability for high-risk goods: https://en.vneconomy.vn/vietnam-mandates-product-traceability-for-high-risk-goods.htm
• Commentary on Vietnam’s new regulations on goods traceability: https://www.productcomplianceinstitute.com/2026/02/12/vietnam-new-regulations-on-traceability-of-goods-and-products/
• Latest rules on managing identification codes and traceability codes: https://www.vietnam.vn/en/quy-dinh-moi-nhat-ve-quan-ly-va-su-dung-ma-dinh-danh-ma-truy-xuat-nguon-goc-hang-hoa
• Outlook for the Thai auto parts industry, IATF 16949 and TISI: https://www.krungsri.com/en/research/industry/industry-outlook/hi-tech-industries/auto-parts/io/auto-parts-2026
• Thailand investment trends in the first half of 2026 and BOI measures: https://panafricanvisions.com/2026/07/thailand-secures-43-6bn-1h-2026-investment-surge-as-big-tech-accelerates-southeast-asia-ai-infrastructure-push/
• BOI investment incentives for 2026 and 2027: https://emerhub.com/thailand/thailands-renewed-boi-incentives-for-2026-2027/