Barcode Management System 2026 — Master Data and Integration Design

When a plant starts evaluating a barcode management system, the first information that arrives is almost always about handheld scanners and label printers. Yet when a project goes live and people report that “it did not get much easier”, the cause is rarely the hardware. It is the design of the codes themselves, and the design of how scan data reaches the production management system. This article is written for production control and IT staff at Japanese-owned manufacturers operating in Thailand, and it works through two axes that most vendor material skips, namely numbering rules for work-in-process tags and master data design, and integration with the core system.

What a barcode management system actually is

The term covers very different scopes depending on who is speaking. Some people mean the handheld terminal. Others mean the label design software. Most stalled approval processes trace back to this mismatch rather than to any technical disagreement.

In practice it helps to treat the system as four parts.

• Code scheme — deciding what counts as one identifiable item and what character string represents it. Tag number structure, lot number derivation, and whether to follow GS1 standards all sit here.
• Issuance — assigning numbers under that scheme and printing them onto tags, labels or direct marks.
• Scanning — reading the code on the physical item and tying it to an operator action.
• Integration — converting each scan event into inventory and production transactions inside the production management system or WMS.

Most published buying guides concentrate on the third part, which is why they read as hardware spec sheets. But almost every deployment that underdelivers bought hardware while the code scheme and the integration were still vague. Scanning stabilises once operators get used to it. Rewriting a code scheme means physically pulling every tag off the floor, and rewriting the integration means modifying the core system. Both are expensive to reverse.

The issuance side, meaning printer integration and reprint control, is covered in Label Issuing Systems 2026 — where numbering rules split. Hardware selection, including the choice between a 1D laser and a 2D imager, is in Handheld Terminals 2026 — cost and selection. This article stays in the space between them.

Decide first what counts as one item — the four identification levels

Starting from a feature comparison matrix almost guarantees a drifting project. The first decision is which physical unit carries a number. On a manufacturing floor, four levels coexist.

These levels are not mutually exclusive. A common configuration puts item plus lot on the work-in-process tag, adds a pack number at shipping, and adds unit serials only for the models where a customer demands it.

The decision that matters is not how many levels you use. It is where the parent-child relationship between levels is stored. Being able to scan a carton number and learn the part number, lot and quantity inside does not work because that information is printed on the carton label. It works because, at the moment of issuance, a record was written stating that this carton number contains this part, this lot and this quantity. Skip that design step and the first time a carton is split or repacked, the ledger and the physical stock diverge, after which no scan in the plant can be trusted.

Once you know which levels to track, choosing the marking method for that granularity is a separate question. For that, see Traceability Barcode Selection 2026 — QR, direct marking and RFID. Here we deal with turning a chosen level into a character string and a master record.

Designing numbering rules for work-in-process tags — six decisions

This is where questions concentrate. Six points need an explicit answer.

1. Who assigns the number. Does the production management system assign it when the work order is released, does the label system assign it at print time, or does the terminal assign it at scan time? Pick exactly one owner, or the same number will eventually be printed twice. The safe default is that the system holding the inventory assigns the number.

2. Fixed length or variable length. Fixed length makes parsing trivial and lets you feed existing production screens without modification. Variable length is easier to extend later but requires an agreed separator convention. A practical split is fixed length for codes that stay inside the company and GS1 variable length for anything that leaves it.

3. Meaningful number or meaningless number. Covered in detail below.

4. Whether a reprint reuses the same number. If a tag becomes unreadable, do you reprint the identical number or issue a new one and void the old? Reusing the number is easier for operators but removes any defence against double-counting. Issuing a new number keeps the ledger clean but requires someone to own the voiding step.

5. What happens on split and merge. If thirty pieces are drawn from a box of one hundred, does the remaining seventy keep the same number or receive a new one? This is the most frequently undecided point, and it is the single largest source of stocktake variance.

6. Number reuse and retention period. When a sequential range runs out, do you reset at the fiscal year or add digits? If numbers are recycled on a shorter cycle than the period over which history is queried, past records become ambiguous.

Meaningful versus meaningless numbers

Should a tag number embed a plant code and line number, or be a plain sequence with attributes held in master data? This is not a matter of taste. Operating conditions decide it.

The realistic landing point is a hybrid. Making only the first two digits a site code and leaving the rest a meaningless sequence prevents cross-site collisions while surviving changes in line layout. What you must avoid is embedding attributes that can change, such as part numbers or customer codes. Every part revision then destabilises the numbering scheme.

GS1 standards or an in-house scheme

The test is simple. Does the code leave the company?

Shipping labels, codes subject to customer receiving inspection, and product codes that enter distribution are all easier to live with if they follow GS1. In GS1 encoding, each data element is prefixed by an Application Identifier, a short numeric prefix that tells the scanner which field follows. The ones encountered most often in manufacturing are these.

These formats follow the Application Identifier dataset GS1 publishes. The practically useful detail here is that AI 10 and AI 21 are variable length with a maximum of 20 characters. If there is any chance your internal tag numbers or lot numbers will later be carried in a GS1 code, keeping them within 20 characters means you will not have to rebuild the scheme at migration time.

Codes that stay internal, such as work-in-process tags and inter-process move tickets, are perfectly fine on an in-house scheme. Forcing GS1 onto them usually lengthens the code, which costs both print area and scan time.

One forward-looking caveat. GS1 has set an industry goal that, by the end of 2027, retail point-of-sale systems should be capable of reading and processing the GTIN from both existing linear barcodes and GS1 2D barcodes. This is what is meant by Sunrise 2027.

Two things about it are widely misread. The first is that it is an industry goal, not a regulation. The second is that linear barcodes are not being retired. GS1 describes the move to 2D as an expansion of capability rather than a replacement of existing barcode technologies, and states that linear symbols such as EAN and UPC will continue to coexist for the foreseeable future. The 1D tags on your shop floor today are not about to stop working.

Movement has nonetheless started in Thailand. GS1 Thailand is running a 2026 roadshow across seven provinces nationwide, Bangkok among them, and the transition from 1D to 2D barcodes is on its agenda. As a live implementation, 7-Eleven Thailand had by early 2023 applied GS1 DataMatrix to roughly one hundred ready-to-eat items produced by CPRAM, carrying the GTIN and the best-before date in a single code across more than 12,000 stores. The reason DataMatrix was chosen over QR was simply that it is smaller for the same payload.

So even if you run an in-house scheme today, reserve enough digits and enough print area to add a GTIN and a lot number later. If you are also reviewing paper-based shop floor tickets, Electronic Kanban 2026 — three boundaries to set before dropping paper covers the adjacent decision.

Master data design — three tables carry the load

A code scheme is inert without master data. In practice three tables do the work.

Item master — absorbing the “three part numbers” problem

The failure point in every item master is number multiplicity. An internal part number, a customer part number and a supplier part number all circulate at the same time, and which one appears on the physical label depends on who shipped it. Cartons arriving from suppliers carry the supplier’s number, while cartons going to customers are required to show the customer’s.

The answer is to fix the internal part number as the primary key and maintain one cross-reference table. That table needs four fields, namely the counterparty code, the counterparty’s part number, the internal part number and an effective date range. Drop the date range and you will have an incident the first time a part revision changeover date falls in the middle of a receiving batch.

The item master should carry at minimum:

• Internal part number and description
• Unit of measure, plus the display unit shown on the terminal
• Standard pack quantity, meaning how many pieces are in one carton
• Whether lot control applies, and where the lot number originates
• Whether unit-level serial control applies
• Cross-reference to counterparty part numbers

Standard pack quantity is routinely treated as optional and it should not be. It is the field that determines how much stock moves when a carton is scanned once. Leave it blank and carton-level scanning simply does not function.

Location master — settle the depth first

How many levels to use across warehouse, zone, rack and bin is the hardest thing to change afterwards, because adding a level means reprinting every shelf label and migrating existing stock records at the same time.

As a working guide, a few dozen storage points need two levels, a few hundred need three, and optimising picking travel justifies four. Alongside the physical hierarchy, define logical locations from the start, covering staging, awaiting inspection, quarantine for defects, and ready to ship. Locations without a physical shelf are exactly where stock stalls and variance accumulates.

Transaction type master — deciding what one scan means

Scanning the same code with the same terminal tells you nothing about whether the event is a receipt, an issue, a transfer or a stocktake. The transaction type the operator selected beforehand decides it.

This master needs the type name, the effect on inventory such as increase, decrease, transfer or record-only, any additional input required such as quantity, location or reason code, and whether approval is needed. The common mistake is creating too many types, after which operators pick the wrong one. Keep the types used daily to six to eight and isolate exceptions on a separate supervisor screen.

Moving scan data into the production management system — four integration patterns

This is the second axis of the article. A successful scan is worthless if the data does not land correctly in the core system. There are four broad patterns.

In Thai plants the realistic answer is usually a combination of the middleware pattern and the batch pattern. Work that must be stopped in the moment, such as shipping verification, runs real-time through the relay server. Work that is confirmed in bulk, such as the monthly stocktake, runs on file transfer. Trying to make everything real-time rarely pays for itself.

Keyboard wedge is the cheapest to start and looks attractive because existing screens stay untouched. Its weakness is fundamental, though. A successful read is not the same as a correct entry. Scan a part number that does not exist, or a lot that has expired, and the screen accepts it. Choose this pattern for a process whose whole point is to stop errors, and the visual double-check survives anyway, which means the labour never actually goes away.

Five things the integration design must state explicitly

Having chosen a pattern, five points belong in the specification. Leave them vague and they will surface at month-end close.

1. When the transaction posts. Does stock move the instant the code is read, or only after a confirmation action at the end of the task? Posting on read is faster for operators but requires a reversal path when work is interrupted. Posting on confirmation is accurate but leaves unconfirmed work in limbo. Whichever you choose, make the “scanned but not confirmed” state visible on screen.

2. Transaction granularity. Does each scan become one line in the core system, or does one task become one document? Fine-grained lines improve traceability at the cost of data volume and processing time. For work such as shipping, where a single order can involve dozens of scans, one document per task is usually what the core system can sustain.

3. Duplicate scan handling. If the same tag is read twice, is that an error or an addition? The correct answer inverts depending on the process. In receiving inspection the second read should be rejected. In a counting task it should add. The standard design is to attach a unique key to every scan, built from terminal ID, timestamp and a sequence, and to have the core system refuse a key it has already accepted. This is also what protects you when a network retry delivers the same payload twice.

4. Offline behaviour and replay order. Wireless coverage in a plant always has blind spots caused by metal racking and operating equipment. If the terminal buffers scans for later transmission, the transmission order must be guaranteed. An issue arriving before its matching receipt drives stock negative and gets rejected. Write into the specification that buffered scans are always sent in ascending order of scan time.

5. Consistency with the accounting close. What happens when scan data for a closed period arrives late? Is it automatically reassigned to the following month, or held as an error for a person to judge? This requires agreement between finance and production control, and IT cannot settle it alone.

If you are also weighing whether to stand up a separate WMS, draw the boundary between what production management holds and what the WMS holds first. The variables behind that judgement are set out in WMS Comparison 2026 — six variables that decide the level you need.

What changes when you extend to QR code inventory management

Search for QR code inventory management and you will find long lists of ways 2D beats linear. In practice the difference reduces to three things.

• Data capacity. A linear code tops out around 20 to 30 characters in practical use, which a part number alone consumes. A 2D code can hold part number, lot, quantity and date together.
• Print area. For the same payload a 2D code is physically smaller. On small components and electronic component reels this is decisive.
• Orientation and damage tolerance. A 2D code reads from any angle and its error correction recovers from partial damage. In processes with cutting fluid or airborne dust this matters.

For a sense of scale, a GS1 US study of 400 US-based organisations published in May 2026 reports that organisations relying primarily on 1D barcodes see a 7 percent first-pass scan failure rate and a 1.5 percent relabeling rate. The cost-reduction figures in the same release are explicitly described as modeled projections, so they are not numbers to transplant into your own business case. The failure and relabeling rates are still worth designing around, because rescanning and relabeling cost more in practice than the price gap between scanner models.

The trade-off is that a 2D code requires a 2D imager. Legacy 1D laser scanners cannot read it, so if you intend to reuse existing terminals, that constraint sets the pace of the migration.

There is also a consequence of the larger payload that is easy to miss. Writing lot and quantity directly into the code means that information cannot be corrected afterwards. Change a quantity and the tag must be reprinted. The principle holds for 2D exactly as it does for 1D, namely that the code should carry only stable identifiers, while attributes that can change belong in master data.

Considerations specific to a Thai site

One number is worth having in mind first. In the JETRO survey of Japanese-affiliated companies operating overseas, published in November 2025, 46.1 percent of respondents in Thailand said they had introduced, were using, or were investing in digital technology, against 583 valid responses. That sits below the ASEAN average of 52.1 percent. The obstacles most often cited in the same survey were a shortage of digital talent at 60.7 percent and high implementation and running costs at 56.0 percent. Read the other way, manual-first operation is still the majority position at Thai sites, which means a plant that moves now separates itself from its peers relatively cheaply.

With that as background, several points make a Thai deployment differ from the same design in Japan.

Supplier tags are not uniform. Japanese-owned suppliers, Thai suppliers and imports from China all arrive with different label layouts and different code schemes. Demanding that every supplier conform is not realistic, so the choice is between relabelling to your own tag at receipt, or maintaining a cross-reference for the major suppliers only. Working down from the largest suppliers by spend, and relabelling the rest, is the practical staged approach.

Screen and label languages. Terminal screens are in Thai, while management reports are in Japanese and English, which is a three-way structure. If you intend to hold Thai product descriptions, build the multilingual fields into the item master from the beginning. Adding them later means data entry across every part number.

Wireless environment. A warehouse dense with metal racking and a production area with large moving equipment have very different radio conditions. Access point design has to follow an on-site survey, and importing an access point layout from a Japanese plant will leave blind spots. The offline design described above is not an option in Thailand, it is a premise.

Workforce turnover. Where staff retention cannot be assumed, a design that relies on training to prevent errors will not hold. The terminal has to narrow the available choices so that invalid combinations cannot be selected at all. The earlier guidance to keep transaction types to six to eight comes partly from this.

BOI incentives. BOI operates a Smart and Sustainable Industry measure aimed at upgrading existing operations, and its digital sub-measure grants a three-year corporate income tax exemption capped at 50 percent of the investment. The minimum investment is 1 million THB, and the condition is not a single tool but the linking of at least three business functions. A barcode management system is by nature an investment that connects scanning, inventory and production reporting, so it sits comfortably against how that requirement is framed. Eligibility still turns on the conditions of the specific project, and applying bundled with equipment differs procedurally from applying separately. Check the current application guide and align with whoever handles applications internally at an early stage.

Cost model

The figures below are a model estimate based on our own assumptions. A real quotation moves substantially with the processes in scope and the state of the existing systems. The premise is 2,000 part numbers, 1,500 work-in-process tags printed per day, 10 scanning terminals, three processes in scope covering receiving, inter-process movement and shipping verification, and an interface added to an existing production management system.

The line worth attention is the second one. If preparing the item master takes an average of eight minutes per part number, including physically confirming the standard pack quantity and deciding whether lot control applies, then 2,000 part numbers come to roughly 267 hours. That is internal labour, so it never appears on a vendor quotation. In reality it is the most common reason a go-live date slips.

The work also spans production control, purchasing and the warehouse, so it will not progress without one person assigned to it. Put master data preparation in the project plan as a task in its own right, separate from system construction.

A 90-day approach

Plans that switch every process at once almost always overrun. Taking one process end to end, closing the design gaps it exposes, and only then expanding, lands sooner.

Days 1 to 30. Choose one process. Shipping verification or receiving inspection work well because the benefit is visible and the scope is bounded. In parallel, decide which identification levels apply and document all six numbering rule decisions. During this window, extract the list of part numbers whose standard pack quantity is missing from the existing item master.

Days 31 to 60. Choose the integration pattern and freeze the interface specification. Write the five points above into the design document, especially posting timing and duplicate scan handling. Start master data preparation at the same time, working first through the part numbers used by the chosen process.

Days 61 to 90. Run the chosen process in parallel. Do not stop the existing paper process. Record both and reconcile the differences. The target is not that the difference reaches zero, but that every difference has an explanation. Stopping the paper process while unexplained differences remain removes your ability to trace where things went wrong afterwards.

Five common failure patterns

• Starting from hardware selection. Fixing the terminal specification first drags the code scheme toward whatever that terminal reads comfortably. The order should be the reverse.
• Proceeding with standard pack quantity blank. Nothing determines how much stock a carton scan moves, so manual entry survives.
• Creating too many transaction types. Past roughly twenty, operators choose wrongly and time goes into correcting mis-posted data.
• Stopping paper too early. Without a parallel period there is no mechanism to detect early design errors.
• Leaving master data preparation unowned. Outside the vendor scope, deprioritised by the floor, and months pass.

Frequently asked questions

What is a barcode management system?

It is the collective term for the mechanism that reads an identification code attached to physical stock and converts each scan event into inventory and production data inside a core system. It consists of a code scheme, issuance, scanning and integration. It is not a word for the hardware alone.

How much does it cost to implement?

The model in this article puts three processes, ten terminals and integration with an existing production management system at 1,500,000 THB initially and 225,000 THB annually. Narrowing scope to shipping verification alone starts smaller. Conversely, if the core system has no existing interface and one must be built, the integration line grows.

Who should decide the numbering rules for work-in-process tags?

The owner of the system that holds inventory, which in practice means the head of production control. Deciding it inside IT alone tends to skip points that reach into floor operations, such as how split and merge are handled.

Should we choose QR codes or linear barcodes?

If the code carries only a part number, linear is sufficient. If you want lot, quantity and date in a single code, or if the available print area is small, 2D has the advantage. Confirm first whether existing terminals can be reused, because 2D requires an imager.

Can barcode integration work with our existing production management package?

Usually yes, and the condition is whether the package exposes an interface. If a standard import file format or an API exists, a relay server can bridge to it. If not, a modification quotation is required, and that line tends to dominate the total.

We already have large stocktake variances. Will barcodes fix that?

Scanning reduces keying errors, but if the variance originates in how numbers are handled on split, or in undefined locations, adding codes will not remove it. Classify the causes of your variance before implementing and confirm which of them scanning actually addresses. On the inventory level question itself, see Optimal Inventory Levels 2026 — five layers to settle before the formula.

Summary

Whether a barcode management system succeeds is decided by the parts you have to think through, namely the code scheme and the integration design, not by scanner performance. The order that works is to decide which of the four identification levels apply, document the six numbering rule decisions for work-in-process tags, build out the item, location and transaction type masters, and only then choose an integration pattern and write down posting timing and duplicate scan handling. Hardware selection can safely come last.

The largest hidden cost is the internal effort to prepare master data. Whether that work, which never appears on a vendor quotation, is planned as a task with an owner is what separates a 90-day go-live from a six-month one.

Talk to us

How to draw numbering rules for work-in-process tags, and how far to connect into an existing production management system, depend on conditions that differ from plant to plant in Thailand. TOMAS TECH supports Japanese-owned manufacturers with Thai operations across the whole range, from code scheme design through to core system integration. It is perfectly fine to be at an early stage, still working out which identification levels your operation actually needs. Tell us how things run today and we will start by separating out the decisions. You are welcome to get in touch through our contact page.

References

• GS1 – GS1 Application Identifiers, the source for the AI table above | https://ref.gs1.org/ai/
• GS1 – 2D Barcodes at Retail Point-of-Sale Implementation Guideline, ratified December 2025 | https://ref.gs1.org/guidelines/2d-in-retail/
• GS1 – Solution provider 2D readiness, the end-of-2027 industry goal | https://www.gs1.org/industries/retail/solution-provider-2d-readiness
• GS1 US – Sunrise 2027 | https://www.gs1us.org/industries-and-insights/by-topic/sunrise-2027
• GS1 US – 2D Barcodes Can Cut Warehouse Costs by Over 60%, May 2026 | https://www.gs1us.org/industries-and-insights/media-center/press-releases/2D-Barcodes-Can-Cut-Warehouse-Costs-by-Over-60-
• GS1 – 7-Eleven Thailand case study | https://www.gs1.org/insights-events/case-studies/7-eleven-thailand-boosts-consumer-safety-and-satisfaction-next-generation
• GS1 – GS1 General Specifications | https://www.gs1.org/standards/gs1-general-specifications
• GS1 – GS1 Digital Link Standard | https://www.gs1.org/standards/gs1-digital-link
• GS1 – Serial Shipping Container Code (SSCC) | https://www.gs1.org/standards/id-keys/sscc
• GS1 – Global Trade Item Number (GTIN) | https://www.gs1.org/standards/id-keys/gtin
• JETRO – Survey on Business Conditions of Japanese-Affiliated Companies Overseas, FY2025 | https://www.jetro.go.jp/world/reports/2025/01/231fa237934b5b0c.html
• Thailand Board of Investment – A Guide to Investment Promotion, April 2026 | https://www.boi.go.th/upload/content/BOI_A_Guide_Web_Th.pdf
• GS1 Thailand – Next Gen Barcodes Roadshow 2026 | https://gs1th.org/next-gen-barcodes-2026-01/