Unit-Level Traceability Management 2026 – Serial or Lot Control Is an Economic Decision

Unit-Level Traceability Implementation in 2026 – Drawing the Line With Lot Control on Economics

When a customer asks you to trace every piece, the first argument is financial, not technical. Unit-level traceability gives each product a serial number and links its manufacturing history to that item. Precision is highest, but marking, reading and records all grow heavier. This article draws the line between lot and unit-level control on three axes – unit price, required precision, part count.

Getting the Difference Between Lot Control and a Unit-Level Traceability System Straight

Before the discussion starts, both definitions need to be aligned. If a plant rushes into “let us strengthen traceability” while this point is still vague, the requirement specification held by the customer and the one held by the factory drift apart, and the investment proceeds on top of that gap.

Lot control treats a group of products manufactured under the same conditions as a single unit, assigns a lot number to that group, and manages it accordingly. Products made on the same day, on the same equipment, from the same material lot all carry the same lot number. Records are kept at the lot level.

Unit-level control assigns a unique serial number to each individual product and links the manufacturing history to that one piece. It is also called a serial number management system. Even for products made on the same day on the same equipment, every number is different. Records are kept at the individual product level.

Where this difference really bites in daily operations is the width of the net you have to cast when a defect is found. Under lot control, the unit for isolating suspect product is the lot. Suppose a product is built in lots of 5,000 pieces (actual lot sizes vary by process and product; this is only an assumed figure); if a problem is found, the scope is 5,000 pieces. Under unit-level control, you can name only the individual pieces that passed through the problem process. In practice, of course, the precision of that narrowing depends not only on whether individual product identification exists, but on how far the process records are actually tied to the individual item. If you assign numbers but keep recording at the process level, the width of the net is no different from lot control.

On the other hand, unit-level control carries a clear cost. Because each product has to be given a number and managed individually, the supplier side takes on more labour and more expense – a point that is almost universal across the literature in this field. In other words, the two methods are not related as “the good method” and “the inferior method”. They are related as two positions on a trade-off between precision and burden, and the question is where you strike the balance.

“Which One to Choose” Is Not a Matter of Taste – It Is an Economic Decision

This is the core argument of the article. The question of lot control versus unit-level control is often handled as a qualitative debate about “what is appropriate for our quality level”, but in reality the boundary can be drawn on three independent axes. If even one of the three tips toward unit-level control, there is a reason to consider it. If none of the three tips that way, you can say with confidence that lot control is sufficient.

Seen through these three axes, the industry-by-industry norms become explainable. The reason unit-level control is taken for granted in automotive parts, medical devices and semiconductors is not that those industries have a higher quality consciousness – it is that at least two of the three axes tip toward unit-level control. Conversely, the reason lot control continues to be used for resin pellets, screws and solder follows exactly the same logic.

The three boundary lines are examined in turn below.

Boundary Line 1 – The Marking Cost Break-Even Set by Product Unit Price

The first axis is the simplest. To make unit-level traceability work, you have to put a mark on the physical item that identifies it one piece at a time. That mark always carries a cost per unit.

Look at the numbers. Commonly cited comparison data puts the per-piece cost of the main media used for individual product identification in the following ranges. A standard thermal label runs USD 0.01 to 0.05, and a passive UHF-band RFID tag runs USD 0.05 to 0.35 when purchased in volume.

Set those figures against the product unit price and the boundary becomes visible. If you apply a label to a part with a unit price of USD 10, the lower bound of USD 0.01 is 0.1% of the unit price. Nobody objects. But apply the same label to a part with a unit price of USD 0.5 and USD 0.01 becomes 2% of the unit price. Use the RFID tag lower bound of USD 0.05 and it is 10% of the unit price. For a volume production part whose gross margin sits in the 10% range, spending 10% of the unit price purely on identification simply does not add up.

The table above is nothing more than the media cost divided by the product unit price. Two assumptions apply: the media cost is the lower bound of each range, and the figures exclude the labour of applying the media and of reading it. In practice that labour is added on top, so the actual burden ratio rises further. Even so, there are many situations where this simple division alone is enough to conclude that unit-level control is not viable for a given part. The general observation that assigning serial numbers to low-priced, high-volume parts is unrealistic because the management cost squeezes profit is simply this calculation expressed in words.

What gets overlooked here is that the cost structure differs fundamentally depending on the marking method. Labels and tags are consumables, so the cost keeps rising in proportion to production volume. Direct laser marking, by contrast, requires an up-front investment in equipment, but because it consumes no consumables the marginal cost per piece is close to zero.

The break-even therefore has to be considered in terms of cumulative production quantity. The break-even point is the up-front investment in the marking equipment divided by the cost of one label. If a label costs USD 0.02 and a complete marking system requires an up-front investment of USD 20,000, the break-even lands at a cumulative 1,000,000 pieces. This investment figure is an assumed value used to show the shape of the calculation – replace it with a quotation for the actual machine when you run the numbers yourself. What matters is the structure: for a part produced at tens of thousands of pieces per month and run for several years, there is a real region in which direct marking is cheaper than the consumable approach. Before concluding that unit-level control is impossible because the unit price is low, it is worth confirming that changing the method moves the break-even point.

Boundary Line 2 – Required Tracking Precision Is Set by Customers and Regulators

The second axis is the one your own company cannot decide. If a customer requirement specification, or the regulation of the country you export to, calls for individual product identification, you have to make unit-level traceability work regardless of the unit price. In that case the discussion shifts from “whether to do it” to “how to minimise the cost of doing it”.

Automotive Parts – The Granularity of Identification Required by IATF 16949

For automotive parts, clause 8.5.2.1 of IATF 16949 covers identification and traceability. What the clause calls for is a traceability plan defined on the basis of the risk and severity level of a failure to employees, customers and consumers, together with defined traceability systems, processes and methods appropriate for each product, process and manufacturing location.

What deserves attention is that the standard does not require blanket serialisation of every part. The same clause calls for serialised identification of individual products to be provided where required by the customer or by regulatory standards. In other words, serialisation is triggered by customer requirements and regulations, not by the standard on its own.

This has an important practical consequence – it means you are free to apply a different granularity to safety-related parts than to the rest. The clause also requires that, for products in the field or in the customer’s hands that may contain quality or safety defects, the start point and end point can be clearly identified. You only need to choose the smallest granularity that satisfies this “identify the start point and end point” requirement – there is no reason to go finer than necessary. Items where a failure translates directly into a safety issue, such as brake parts and steering parts, are handled at the individual product level, and the rest at lot or sub-lot level. That is the realistic design. If you want to organise your IATF 16949 response systematically, Traceability for Automotive Parts and IATF 16949 Compliance covers it from the angle of designing your burden of proof.

Note also that the same clause requires the identification and traceability requirements to extend to externally provided products with safety or regulatory characteristics. Applying unit-level traceability to your own finished goods alone will not make the plan work if records cannot be obtained on the purchased part side.

Medical Devices – Individual Identification and Direct Marking Required by UDI

For medical devices, the UDI regulation in the United States is more explicit. A UDI is composed of a device identifier (DI) and a production identifier (PI), and the PI covers items such as the lot or batch in which the device was manufactured, the serial number of a specific device, the expiration date and the date of manufacture.

What matters is that the components of the PI are conditional. If a serial number appears on the label, then the PI portion of the UDI must include that serial number. If a lot number and an expiration date appear on the label, then the UDI must include PI segments for both of them. The structure is that every production identifier appearing on the label has to be embedded in the UDI.

On top of that, 21 CFR 801.45 makes direct marking mandatory. Paragraph (a) of that section provides that, among devices that must bear a UDI on their label, a device intended to be used more than once and intended to be reprocessed before each use must bear a permanent marking on the device itself that provides the UDI. Under paragraph (c), this direct marking may be provided as easily readable plain text, or through automatic identification and data capture (AIDC) technology or any alternative technology that provides the UDI on demand, or as both.

Paragraph (d) of the same section also sets out exceptions: where any direct marking would interfere with the safety or effectiveness of the device, where it is not technologically feasible, where a single-use device is subject to additional processing and manufacturing for additional single use, and where the device has already been marked under paragraph (a). The very existence of these exceptions is the flip side of the fact that direct marking is required as a matter of principle.

Semiconductors and Electronic Components – Direct Marking as a Counterfeit Countermeasure

In semiconductors and electronic components, the motivation for individual product identification is somewhat different. Here the driver is less the tracking of quality defects and more the prevention of counterfeit and non-conforming parts entering the supply chain.

According to the ERAI annual report for 2025, the number of suspect parts reported that year fell by 29.1% compared with the previous year. That decline does not mean the problem has gone away. In the same year, global semiconductor sales rose by 25.57%, so the number of reports fell while transaction volume was growing. Among the reasons cited is that roughly 35% of the industry growth in 2025 was driven by demand for high-performance AI-related components, which are difficult to counterfeit because of their technical complexity and export controls. The same report indicates that just over 38% of reports came from organisations based in the United States, which implies that a substantial share of reporting originates outside the United States.

The takeaway is that even though counterfeit reports are down across the market as a whole, that is a consequence of growth in high-value parts that are hard to counterfeit – it does not mean that conventional general-purpose ICs have become safe. Programmable logic, analogue ICs, microprocessors and memory continue to be cited as targets for counterfeiting.

It is in this context that direct marking on the component itself carries weight. Laser marking does not disappear through abrasion, scratching or corrosion, it does not bleed like ink, and it does not peel off like a label. Because it uses no consumables its environmental load is also low, and it can be applied to metal, ceramic, resin, wood and rubber. A mark can be applied in seconds, with no pre-treatment or finishing process required. If a label peels off, the individual item becomes unidentifiable, whereas a direct mark stays with the part for its whole life. That permanence eliminates the possibility of the mark being swapped somewhere along the supply chain. Implementation topics on the electronic component side are covered in Building an Electronic Component Traceability System.

Boundary Line 3 – Part Count and Depth of Linkage Determine the Hybrid Design

The third axis concerns what happens when you look beyond the finished product to its components. A common misconception in practice is to treat unit-level control and lot control as mutually exclusive options. In reality, many factories run both at the same time.

The thinking goes like this. The finished product is given an individual product number, and the lot number of each component used is linked to that individual product. The pattern of needing both serial tracking of the assembled product and batch tracking of the input materials is common to automotive parts and medical devices alike. Individual identification on the finished product side, lot identification on the component side, and a table that links the two – that is the standard design.

With this design, narrowing down becomes a two-step operation. If a problem is found in a given material lot, you can pull the list of individual product numbers for finished goods that used that lot. Conversely, if a problem is found in a given finished product, you can pull the material lots that piece consumed. Being able to follow the trail in both directions comes from the linkage table holding both the individual product number and the material lot – not from assigning individual numbers to every component.

The more parts a product contains, the more pronounced the economics of this design become. Suppose one finished product uses 80 components. If you apply unit-level control to every component, you need 81 records for a single finished product. If you apply unit-level control only to the finished product and handle the components by lot, you get one finished-product record plus 80 linkage rows, and the marking cost for the components themselves drops to zero. The amount of traceable information barely changes, while the marking labour on the physical items and the media cost disappear. This contrast between 81 records and 1 record is simply a side-by-side comparison of record structures – it is not the actual measured figure for any particular product.

So when should components be brought under unit-level control? The deciding factor is whether the component can be removed from the finished product and replaced. A replaced component changes its correspondence to the finished product after replacement. With a lot number alone, you can no longer tell which lot the currently installed component came from. Replaceable assemblies such as circuit boards, batteries and motors are worth giving individual product numbers separate from the finished product. Conversely, components such as solder, adhesive and paint, which are consumed during the process and cannot be extracted again, are inherently outside the scope of unit-level control.

Designing Serial Number Assignment Rules – The Foundation of Manufacturing History Management

Once the method is decided, the next step is designing the number itself. Treat this lightly and there is no way back later, because the number is marked on the physical item and shipped, and it cannot be changed after the fact.

The first fork is whether the number should carry meaning. Significant numbering embeds information such as a plant code, a line code, a production year and month and a sequence number inside the number itself, with the advantage that you can read the item’s origin straight off the number. Non-significant numbering uses a simple unique sequence or random value, with all the information held in the system-side ledger.

In practice, significant numbering frequently causes problems years later. Every time something changes – a new plant is added, lines are reorganised, the digit count runs short – the embedded meaning stops matching reality. Numbers on products already shipped cannot be changed, so number systems with different rules end up coexisting. The recommendation is to make non-significant numbering the default and look up the item’s origin on the system side. If meaning absolutely has to be embedded, limit it to attributes that do not change, such as the year of manufacture.

Set the digit count with margin against the cumulative production quantity you anticipate. If the numbers run out during the production period, you are forced into the worst possible choice – reusing numbers. Using the same number twice under unit-level traceability destroys the very premise of tracking. Numbers belonging to scrapped items must not be reused either.

On code systems, riding on an existing international standard is the safe approach. In the GS1 system, combining application identifier (01), which represents the trade item code GTIN, with (21), which represents the serial number, forms a serialised GTIN (SGTIN) that uniquely identifies each individual item. Lot numbers are represented by (10). The information carried by these identifiers is then encoded into GS1 DataMatrix or GS1-128. In pharmaceuticals, the GTIN is used in combination with the lot number and expiration date, and for prescription drugs a serial number is mandatory.

As a more general framework, ISO/IEC 15459 defines the standard for unique identification. Under that standard, each individual transport unit must be uniquely identified by an identifier and a string defined by an Issuing Agency. That string is intended to be represented on a barcode label or other AIDC media. The syntax for high-capacity AIDC media is defined by ISO/IEC 15434, so when large volumes of data are involved you select that syntax and then choose a suitable medium such as Code 128 or Data Matrix. Before you invent your own numbering rules, check which system your customer is already using.

Choosing a Marking Method – Where 2D Codes, RFID and Direct Marking Each Fit

Once the numbering rules are set, you decide how to put that number onto the physical item. There are four main options.

The equipment cost on the reading side cannot be ignored either. An RFID handheld reader runs USD 1,200 to 3,000, whereas a barcode handheld scanner runs USD 50 to 200. Looking at a complete system, the ranges cited are USD 25,000 to 150,000 for RFID and USD 2,000 to 15,000 for barcode. A difference of one order of magnitude always makes itself felt in the implementation decision.

RFID still gets chosen, and the reason is the labour it saves at the point of reading. RFID can read multiple tags at once, while barcode reads one item at a time. This difference is expected to cut labour cost by 60 to 80%, and sites processing more than 10,000 items per day tend to see a return on the investment, with a payback period cited as 12 to 24 months. A 5-year TCO estimate for one distribution centre gives USD 4,545,000 for RFID against USD 5,480,000 for barcode, a difference of USD 935,000 and a break-even at 14 months. Note, however, that this estimate covers a distribution centre, and the conditions differ from a production line where the number of read operations scales with the number of processes. Substitute your own process count and read points before evaluating it.

In the automotive industry there is an industry guideline for direct 2D code marking on parts. AIAG B-17, the “2D Direct Parts Marking Guideline”, covers applying Data Matrix or QR codes directly to parts using laser, dot peen and inkjet technologies and reading them back, and it is positioned as a complement to B-4, the application standard for part identification and tracking. If your customer is an automotive OEM or a Tier 1, check the applicable revision of this guideline first.

It is also worth noting that combining multiple media in a single configuration is becoming common in practice – 2D codes for identifying parts one at a time, RFID for managing containers or sequence units. The detailed decision criteria for media selection are set out in Selecting Barcode, QR and RFID for Traceability.

Linkage Design That Makes Manufacturing History Tracking Work – Material Lot Linkage and MES Integration

Assigning a number and making it readable does not by itself make a unit-level traceability system function. Manufacturing history tracking is only established once the number you read is linked to the records of the processes that piece went through.

There are broadly four kinds of information to link. First, input components – the lot number of each component used in that individual item. Second, process conditions – the equipment number, setpoints, measured values and work timestamps at the time of processing. Third, inspection results – the measured values and judgements at each inspection process. Fourth, shipping information – when, to which destination, and which individual items were shipped. The finished form of unit-level traceability is a state where all four can be pulled together on a single key, the individual product number.

The key design point is that the individual product number must already be fixed at the moment each record is created. If you design the number to be assigned at the final process, the records from every process before it will not be linked to the individual item. If you want to keep the upstream records too, you have to fix the number at an early stage of machining and have that number read at every subsequent process. Conversely, if linking only the final inspection result to the individual item is enough, assigning it at the final process is fine. How far back you want the linkage to reach is what determines where the number is applied.

The second key point is how records are entered. Because unit-level control generates as many records as the production quantity, manual entry is guaranteed to break down. Commentary on manufacturing DX in Thailand cites initiatives such as automatically collecting data from inspection equipment on a work-instruction basis, issuing inspection certificates and recording physical item tags together with serial numbers, and making daily work reports paperless with handheld terminals so that aggregation becomes easy. Records under unit-level traceability should be captured either automatically from equipment or simply by scanning on the shop floor. A design that relies on people transcribing data will become a hollow formality the moment volume production starts.

For the record repository, an MES or a quality management system is the usual choice. What needs examining here is how the record count grows. What was one record per lot under lot control becomes one record per piece under unit-level control, and process records multiply further by the number of processes. If you set a retention period of 10 years, that full 10 years of records has to remain in a searchable state. Confirm the database design and the storage and retrieval performance before you commit. How to hold quality data is organised in Designing a Quality Data Management System.

Handling of non-conforming product belongs in the design as well. The commentary cited above describes practices such as recording non-conformity on the physical item tag to prevent it flowing downstream while tracking the corrective action status, and automatically notifying the relevant departments of a non-conformity occurrence by email. Because unit-level control records non-conformity at the individual item level, you can later check by quantity whether the same defect is concentrated on a particular machine or a particular time of day. This is a secondary effect of unit-level traceability, but it is the one the shop floor welcomes most.

The Cost Structure of a Unit-Level Traceability System – Estimate Where the Weight Falls Before You Start

The cost of unit-level traceability is weighted toward operation rather than up-front investment. The items most often overlooked at the decision stage are organised below by where they occur.

Of all these, the one most often underestimated is the cycle time of the reading process. Even a scan of 2 seconds per piece becomes 10 seconds if it is read at 5 processes. At a daily output of 2,000 pieces that is 20,000 seconds per day, roughly 5.6 hours. This figure of 5.6 hours is simply the arithmetic of 2 seconds times 5 processes times 2,000 pieces = 20,000 seconds converted into hours – it is not the measured result of any particular factory. Change the assumptions and the result changes too. Even so, this simple multiplication is enough to show that designing the read operation so that a person has to stop and perform it generates labour on a scale that matters. That is precisely why fixed readers that read as the item passes by, or placements that read during machining, become important.

One more point: the retention period setting feeds straight into cost. Automotive parts can require long-term retention that takes the service life of the vehicle into account, and as the retention period lengthens, the cost of keeping that many years of records in a searchable state accumulates. Choose a system without confirming the customer’s required retention period and you will hit a wall on both capacity and search performance a few years later.

How to Proceed When Unit-Level Traceability Is Demanded at a Thai or ASEAN Plant

For Japanese-owned companies with manufacturing sites in Thailand, the demand for unit-level traceability usually comes from the parent company in Japan or from the customer in the export destination. Cases where the site itself feels the need and starts on its own are rarer than cases where the requirement comes down as a specification.

The first thing to confirm in that situation is the granularity of the requirement. It is not unusual for a requirement specification to say nothing more than “traceability shall be ensured”, without stating whether that means individual product level or lot level. As noted above, even under IATF 16949, serialised identification is required where the customer or regulatory standards require it – the standard does not impose unit-level control across the board. Ask the customer to confirm in writing which granularity is required, for which items and for which characteristics. This confirmation alone can change the investment amount by an order of magnitude.

Next, factor in the local constraints. It has been pointed out that at local subsidiaries in Thailand the scope of responsibility carried by IT staff is broad, and that most companies struggle with the manpower side of their DX implementation structure. A unit-level traceability system is not finished when it goes live – detection of missed reads, master data maintenance and responses to changing customer requirements all continue indefinitely. Put in the mechanism without anybody locally to run it and the operation will have stopped six months later. Decide who owns day-to-day operation at the same time as you evaluate the implementation.

Staff turnover has to be factored in as well. Unit-level control only works when every operator on the floor follows the procedure of “have it read, then pass it on”. Procedures documented in Thai, and a mechanism that prevents the item from advancing to the next process if a read is skipped – without these two, holes appear in the records the moment the trained people leave. A design that does not depend on people, meaning an interlock that physically blocks progress unless the item is read, works out cheaper in the end.

Check the documentation requirements for export too. Whether individual product numbers need to appear on shipping documents or inspection certificates changes how the shipping process must be built. If the operating pattern involves the customer coming back after shipment with a query by individual product number, you need an internal screen that pulls the history from that number.

A Staged Approach to Implementation

Rolling out unit-level control across the whole company at once is not advisable, either on cost or on shop floor load. Starting with a narrow scope and expanding it as requirements dictate is the realistic path.

The first stage is narrowing the scope. Limit it to the items for which a customer requirement or a regulation explicitly calls for individual product identification. Not expanding this into “if we are doing it anyway, let us do everything” is the single biggest lever for holding cost down. Apply the three boundary lines above item by item, and leave the items that do not tip under lot control.

The second stage is designing the number and deciding where it is applied. Fix the numbering rules, the digit count, the code system, and the process at which the number becomes final. Confirming which system the customer uses at this stage prevents rework later.

The third stage is physical validation of the marking method. Even when the catalogue says there is no problem, the actual material, the actual surface condition and the actual soiling pattern can make the mark unreadable. Direct marking in particular sees its read rate change with the interaction of material and contrast. Always insert a sample validation under volume production conditions.

The fourth stage is designing the read points. Decide at which processes the item is read, whether readers are fixed or handheld, and how missed reads are detected. Build a design that depends on human attentiveness here and holes will appear later without fail.

The fifth stage is the record repository and system integration. Fix the record count, retention period and search requirements, and design the connection to the existing production management system and quality management system.

The sixth stage is building a simulated enquiry into operations. Pick an arbitrary individual product number and periodically confirm that the manufacturing history for that item can be produced within the required response time. Whether the mechanism runs and whether it can produce an answer when you need one are two different things.

Frequently Asked Questions

Should we choose unit-level control or lot control?

Judge it on the three boundary lines. Is the marking cost negligible against the product unit price? Does a customer requirement or a regulation demand individual product identification? Is history required per component? If even one applies, there is a reason to consider unit-level control. If none of the three apply, lot control is sufficient. This is not a question to be settled by how high your quality consciousness is.

How much will management cost increase if we introduce a unit-level traceability system?

The places where it increases are predictable. Marking on the physical item occurs in proportion to production quantity, the reading process loads onto cycle time, and the record count scales as production quantity times number of processes. The monetary amount cannot be generalised because it differs by item, but running these three through your own volumes will give you a rough estimate. On marking, note the difference in structure – consumable-based methods such as labels scale with quantity, while direct laser marking is dominated by up-front investment and carries a marginal cost close to zero.

How should we set the numbering rules for a serial number management system?

Use non-significant numbering as the default. Embed plant codes or line codes and the number stops matching reality every time sites or lines are reorganised, leaving number systems with different rules coexisting. Take ample margin on the digit count against the anticipated cumulative production quantity, and never reuse numbers, including those of scrapped items. Riding on an existing standard such as the GS1 SGTIN or ISO/IEC 15459 makes exchanges with the customer far easier.

Is material lot linkage impossible unless every component gets an individual number?

No, it is not necessary. Give the finished product an individual product number and hold a table that links the lot number of each component to that individual item, and you can narrow down in both directions. Components worth giving individual numbers are those that may be removed from the finished product and replaced – assemblies such as circuit boards, batteries and motors. Components consumed during the process that cannot be extracted again are inherently outside the scope of unit-level control.

Should we choose RFID or 2D codes?

It comes down to unit price and reading conditions. A 2D code label runs USD 0.01 to 0.05 against USD 0.05 to 0.35 for a passive RFID tag, and the reading equipment differs by an order of magnitude too. RFID still gets chosen because it can read where line of sight is unavailable and because it can read many items at once. Labour cost reductions of 60 to 80% are expected, and sites handling more than 10,000 items per day are cited as seeing payback in 12 to 24 months. In practice, combining the two is increasingly common – 2D codes for identifying parts one at a time, RFID for managing containers or sequence units.

Which process should manufacturing history management start from?

Decide first which process fixes the individual product number. Assign the number at the final process and none of the earlier records will be linked to the individual item. If you want to keep the upstream records too, design it so that the number is fixed at an early stage of machining and read at every subsequent process. How far back you want the history to reach determines where the number is applied.

When starting unit-level control at a Thai plant, what should we confirm first?

The granularity of the customer requirement. If the specification says only “traceability shall be ensured”, confirm in writing whether that means individual product level or lot level. Even under IATF 16949, serialised identification is required where the customer or regulatory standards require it – the standard does not impose it across the board. Next, decide who will own day-to-day operation.

Summary

Whether to introduce a unit-level traceability system is an economic decision, not a question of which method is superior. There are three inputs to that decision.

The first is product unit price. Against media costs of USD 0.01 to 0.05 for a label and USD 0.05 to 0.35 for an RFID tag, where your product unit price sits determines whether the numbers work. That said, choosing an up-front investment method with a marginal cost close to zero, such as direct laser marking, moves the break-even point.

The second is required tracking precision. Under IATF 16949, serialised identification is something required where the customer or regulatory standards require it – the standard does not impose it across the board. For medical device UDI, if a serial number appears on the label it has to be included in the PI portion of the UDI, and devices used more than once and reprocessed before each use are subject to mandatory direct marking under 21 CFR 801.45. Identifying where the requirement originates is the first task.

The third is part count and depth of linkage. A hybrid design – individual identification for the finished product, lot identification for components, and a table linking the two – is the optimum for most products. Assigning individual product numbers to every component should be considered only for assemblies that may be replaced.

And assigning numbers alone does not complete unit-level traceability. Creating a state where input components, process conditions, inspection results and shipping information can all be pulled together on a single key, the individual product number, and not relying on people transcribing records, is what determines whether the operation survives.

The hardest part of implementing a unit-level traceability system – and the part with the biggest effect on cost – is deciding which granularity applies to which items. Interpreting customer requirements, and judging how much record data can already be captured from your existing equipment, are areas where organising the thinking at the evaluation stage alone can be of use. We are glad to talk even at the concept stage, before requirements are settled.

Contact TOMAS TECH here

References

• Traceability Management Methods – Difference between lot control and unit-level control
• What Lot Control Is – Comparison of labour and cost against serial control
• IATF 16949:2016 Clause 8.5.2.1 Identification and traceability – Traceability plan and the requirement for serialised identification
• 21 CFR 801.45 Devices that must be directly marked with a unique device identifier – Direct marking obligation for medical devices and its exceptions
• 21 CFR Part 801 Subpart B Labeling Requirements for Unique Device Identification – Composition of the DI and PI of a UDI
• RFID Tags vs Barcodes Cost – Media unit costs, reading equipment costs, TCO estimate
• ERAI Report on counterfeit electronics – Trend in counterfeit part reports for 2025
• 7 Ways Laser Marking Improves Part Traceability – Permanence of direct laser marking and compatible materials
• AIAG B-17 2D Direct Parts Marking Guideline – Guideline for direct 2D code marking on automotive parts
• GS1 Application Identifiers – Composition of an SGTIN from a GTIN and a serial number
• ISO/IEC 15459-1:2014 Unique identification – Framework for unique identification and issuing agencies
• ISO/IEC 15434:2019 Syntax for high-capacity ADC media – Syntax for high-capacity AIDC media
• Manufacturing DX at Local Subsidiaries in Thailand – Recording physical item tags and serial numbers, automatic collection of inspection data