Electronics Traceability 2026 | Four Units to Trace, Five Cost Layers

When a customer sends one board back to you, the first question is almost always the same. Which reel did the parts on this board come from? The part number is in your ledger. So is the lot number. And still, most factories cannot answer. The reason is simple — reels get spliced mid-production. Electronics traceability built on the same lot-tracking design used for automotive parts stops dead at that one question, because the units you actually have to trace are not one thing but four.

Why lot-based tracking is not enough for electronics traceability

Most factories starting a traceability project take the automotive template as their reference. Record the incoming lot, carry the lot forward through each process, and connect the lineage all the way to the shipping lot. It is a proven design, aligned with what IATF 16949 asks for. We covered how that burden of proof works in our article on automotive parts traceability.

Bring that same template onto an electronics assembly line, though, and it will break. It breaks in four specific places, and every one of them has the same root cause — the lot number is being used as the primary key.

First, the same part number and the same lot still will not tell you which board it went onto. When a tape reel runs low, it is spliced onto the next reel. Across that splice, the manufacturing lot changes and so does the date code. Which means the answer to “which lot went onto which board” is decided only by comparing the time the reel was changed against the time the board was placed. You can record lot numbers meticulously and still have nothing to link them to, if you have no timestamps.

Second, the floor life of an MSD (Moisture Sensitive Device, a surface-mount part sensitive to humidity) is neither an inventory figure nor a quality record. Floor life is cumulative exposure time from the moment the bag is opened. Putting the reel back on the shelf does not stop the clock. There is no home for that column in an inventory system or in a quality record, so it needs a ledger of its own.

Third, a panelized board branches one-to-many. The moment one panel becomes five individual boards, the parent-child relationship is severed. A factory with no unit-level ID can only contain a defect at panel granularity. If the containment scope differs by a factor of ten, the sorting cost differs by a factor of ten as well.

Fourth, lot numbers do not stop counterfeit parts. Lot numbers and date codes are printed marks, and printed marks can be faked. What you need to capture at the receiving gate is not the number on the part but the route it came through — who you bought it from.

What these four have in common is that none of them connect through “lot.” They connect through time or through route. So the thing an electronics factory needs to design is not a lineage of lots. It is four ledgers keyed on time.

There are four units to trace — reels, MSD, unit-level boards, purchase route

Let us put the whole picture down first. These four differ in their primary key, in when the record is written, and in how long it has to be kept. Try to force them into a single table and one of them always gets sacrificed.

Building the four ledgers in isolation is pointless. The key that joins them is time. The set time in the reel ledger against the placement time in the mounter log. The opening time in the MSD ledger against the set time in the reel ledger. The depaneling time in the unit ledger against the placement time of the panel. If those three joints hold on time, you can start from one shipped board and pull the reel, the lot, and the purchase route of every major part on it in a single continuous query.

Put the other way around, the precision of your timestamps is the precision of your traceability. In a factory that records reel changes only to the day — “9 August” — every board that ran that day falls inside the suspect scope. In a factory recording to the minute, only the boards within thirty minutes either side of the splice do. Same four ledgers, different time granularity, and the containment scope changes by a factor of tens.

From here we take the four units one at a time, down to the level of which columns you actually record.

Unit 1 — reel balance and splicing link only through time

Splicing is the act of feeding in new parts without stopping the mounter. The leading end of the next reel’s carrier tape is joined to the trailing end of the nearly empty one with a metal clip or splice tape. On a single line it happens dozens of times a day.

The problem is that this joint is the boundary line of the parts’ identity. Before and after the splice, the manufacturing lot changes. The date code changes. In some cases it is even the boundary between what you bought from a franchised distributor and what you bought from a broker. And the mounter does not recognize that boundary at all, because from the machine’s point of view the only fact is that parts came out of the feeder.

So the link from board to part lot is decided only by matching the reel change or splice time against the board placement time. Factories running this on paper forms will almost certainly have gaps. Keeping a 100% fill-in rate on a task that happens dozens of times a day is not realistic.

The columns a reel ledger needs

At minimum, a reel ledger needs the columns below. Fix the columns first, then decide the input method. Do it in the reverse order and you will find columns missing later.

Once the columns are settled, the next job is to make the operation something the floor can actually sustain. This is where many factories fail. Add manual entry fields and they will be skipped during the busy hours, every time. Make the splice record nothing more than scanning two barcodes back to back. The ID of the reel running low, then the ID of the reel being spliced in. The terminal stamps the time automatically. The number of fields an operator has to type is zero. Whether you can achieve that design decides whether the reel ledger survives.

What you use for identification depends on environment and read distance. Does it go through the reflow oven? Is there a cleaning process? Is it read with gloves on? We have collected what you need to make that call in our article on choosing barcodes for traceability.

Unit 2 — MSD floor life is a separate ledger of cumulative hours

An MSD starts absorbing ambient moisture the instant the moisture barrier bag is opened. Send it into the reflow oven still holding that moisture and the water vaporizes violently, delaminating the package internally — the popcorn effect. To prevent it, IPC/JEDEC J-STD-020 classifies each part into an MSL (Moisture Sensitivity Level), and IPC/JEDEC J-STD-033 then specifies, for each MSL, how the part must be handled and how much cumulative time it may sit in an ambient environment after opening.

The critical point here is that floor life is cumulative time, not continuous time. As long as the part sits in an environment at or below 30 degrees Celsius and 60% relative humidity, the clock keeps running. Putting it back on the shelf does not reset it. Use it for two hours in the morning, return it to the shelf, use it again for three hours the next day, and you have consumed five hours.

To stop the clock you put the reel into a dry cabinet below 10% relative humidity, or into a nitrogen-purged environment below 5%. If the limit has already been exceeded, you either bake the moisture out — for example 192 hours at 40 degrees Celsius still in tape and reel, with conditions set by the part’s heat tolerance — or scrap the parts.

The primary key of this ledger is the opening time, not the lot number. That is the decisive point. You can have three reels of the same part number from the same lot, and if all three were opened at different times, they have three different remaining floor lives. A table keyed on lot number has nowhere to write those three values. Which is exactly why adding a column to an existing inventory system does not work, and a separate table is required.

The minimum you have to do operationally comes down to three things.

• At opening, scan the reel ID and the bag label so the opening time is stamped automatically
• Scan again when the reel goes back into dry storage, so the time the clock stopped is recorded
• Show a list of reels whose remaining floor life has dropped below the threshold on the shop-floor terminal

Recording the storage environment itself looks like a different topic, but it feeds the same ledger. If there were periods when the “30 degrees Celsius and 60% RH or below” assumption was not holding, floor life during those hours was being consumed faster than the ledger assumes. How to capture temperature and humidity in dry cabinets and material warehouses is covered in our article on temperature and humidity monitoring systems.

Unit 3 — panels branch one-to-many into unit-level boards

With a five-up array, one panel yields five individual boards. At the moment of depaneling — router cutting or V-cut breaking — the panel as a unit ceases to exist. If you have not preserved the parent-child link there, then when one board comes back from the field there is no road back to its assembly history.

There are three ways to give an individual board an ID. They differ in cost and in how well the ID survives downstream processes.

The third option, the logical ID, looks attractive. It costs nothing extra and you can start today as long as you manage panel IDs. It has one decisive weakness, though. After depaneling, the position number can no longer be read off the physical board, so it is useless for traceback after shipment. When a customer returns a board, you have no means of saying which position of which panel it was. Choose it with clear eyes — usable for internal defect analysis, not usable for answering customer requirements.

There is a second trap that factories fall into right after adding unit IDs. They issue the ID but never link it to the AOI and ICT judgments. Having a unit ID means nothing if there is no record of what AOI judged that board to be and where it was pulled from the line. Only when you also retain “the board judged defective was pulled at this time, by this person, into this rack” can you squeeze the suspect scope down to a few dozen boards. Issuing unit IDs and linking process judgments look like two separate projects but are in fact one design. Implement ID issuance first and defer the judgment linkage, and you will find the judgment data has no column to hold the unit ID — which means rebuilding it.

Unit 4 — without a purchase route you cannot stop counterfeit parts

Lot numbers and date codes are marks printed on the package surface. Printed marks can be reproduced. Grinding the surface and re-marking it — remarking — is an old technique, and parts good enough to defeat visual inspection are in circulation. Record numbers as carefully as you like and the counterfeit entry point stays open.

In its 2024 annual report, ERAI recorded 1,055 suspect parts reported during the year, up 25% year on year and the highest count since 2015. Note that the 25% includes a single batch of 248 fan assemblies reported by the U.S. Government in May; strip that batch out and the underlying increase is around 3%. The direction is still upward. On top of that, 29.40% of them carried a brand that had never been reported before. Narrowing your watch list by brand concentration is getting less effective every year.

There is a second fact here that runs against practical intuition. It is not only hard-to-source obsolete parts that come back counterfeit. ERAI’s 2024 report includes plenty of current, freely available parts bought through normal channels. Setting your alert level purely by sourcing difficulty is not enough on its own.

So what you capture at the gate is the route, not the number. Your receiving ledger must carry a column with these four categories.

The realistic answer is not to raise the inspection level across the board. That cost is unsustainable. It is to raise the level only on broker purchases. For that to work, you must be able to retrieve later which reels came through a broker — and that is where the reel ledger and the purchase route ledger join. Being able to answer a customer inquiry immediately with “of the parts on this board, zero came through a broker” is the single biggest payoff from carrying that column.

Translating the four IPC-1782 levels into your own customer requirements

IPC-1782 is the standard for traceability in the manufacturing and supply chain of electronic products. It defines four levels for materials and for processes, structured so that minimum requirements scale with risk. The levels map onto IPC’s product classes (Class 1, Class 2, Class 3) and onto sector-specific requirements.

There is one thing many factories misread about how to use this standard. Do not read it as a certification scheme. Its real use is as a measuring stick for translating the scattered, inconsistent requirement statements that arrive from customers into your own vocabulary.

In practice, customer specifications arrive worded like this. “Individual identification shall be possible.” “Traceable back to the manufacturing lot of the component.” “Affected scope shall be presentable within 24 hours of a defect occurring.” The wording and the granularity differ with every customer. Handle each model individually and the number of line-side designs grows with the number of customers.

So you assign each requirement to one of the four levels first. Once that assignment is done, your model list looks like this.

With that one sheet, “which models get which equipment” is decided. Start equipment selection without it and you end up equipping every model to the highest level, and the cost has no ceiling. Note that IPC-1782 is a paid standard, so for the specific requirement items at each level you should refer to the standard document itself. This article goes no further than three statements — there are four levels, requirement depth scales with risk, and the levels map to product classes.

Breaking the cost into five layers — three-year TCO for all models at once

From here we move to money. Everything below is a model case. The full arithmetic is shown so you can substitute your own actuals and recalculate.

The premises. A Japanese-owned EMS (electronics manufacturing services) plant in Chonburi Province, Thailand. Three SMT lines, AOI and ICT after reflow, depaneled before shipment. It produces 180 models, with an average of 240 major components per model. Monthly output is 60,000 boards, and with a five-up array that is 12,000 panels a month. Two shifts, 250 working days a year.

Labor cost is set as follows. The minimum wage in Chonburi Province is 400 baht per day. Divided by eight hours, that is 50 baht per hour. On top of that we apply roughly a 1.5 multiplier for social security, bonuses, meal allowances, transport and similar overheads, giving an effective operator cost of 75 baht per hour. Engineers are set at 220 baht per hour. That 1.5 multiplier is an assumption. Replace it with your own actuals and recalculate. The overhead ratio varies widely between factories, so substitute the real figure from your payroll ledger.

Benefit A — engineering hours spent investigating customer defects

Customer complaints and returned parts run at 26 cases a year. Each investigation takes an engineer 18 hours, spent manually cross-checking the reel ledger spreadsheet, the daily reports, and the AOI logs. If the four ledgers are joined on time, that work drops to 3 hours.

The saving per case is 18 minus 3, or 15 hours. 26 cases times 15 hours is 390 hours a year. 390 hours times 220 baht is 85,800 baht per year.

Benefit B — excessive sorting because containment scope cannot be narrowed

Cases requiring you to cut out a suspect scope run at 6 a year. Today you can only cut at lot granularity, so an average of 9,000 boards go into sorting. Narrow it to unit level and that average becomes 900 boards, one tenth. Sorting costs 12 baht per board (receiving, re-inspection, and repacking combined).

Today that is 6 cases times 9,000 boards times 12 baht, or 648,000 baht a year. Narrowed to unit level it is 6 cases times 900 boards times 12 baht, or 64,800 baht a year. The gap is 583,200 baht.

However, the link between unit ID and panel ID only holds for models already converted. That gap will not appear on every case, so we set a realization rate of 60%. 583,200 times 60% is 349,920 baht per year. That 60% is an assumption. Replace it with your own actuals. The rate moves a great deal depending on what share of models is converted and which models the complaints concentrate on.

Benefit C — MSD floor life overruns

Reels at MSL3 equivalent run at 420 a month. Of those, 34 a month exceed floor life and go to bake. The direct cost of one bake (equipment occupancy and setup) is 180 baht. On top of that, waiting for bake generates 9 extra changeovers a month at 1,200 baht each.

Today that is 34 times 180, or 6,120 baht, plus 9 times 1,200, or 10,800 baht. Together 16,920 baht a month, and 203,040 baht a year.

Record the opening time automatically and make remaining floor life visible, and we assume reels going to bake fall to 8 a month and extra changeovers to 2 a month. 8 times 180 is 1,440 baht, and 2 times 1,200 is 2,400 baht. Together 3,840 baht a month, and 46,080 baht a year. The saving is 203,040 minus 46,080, or 156,960 baht per year.

Benefit D — recurrence caused by missing splice records

Defects that recur because you cannot identify which reel was involved, and therefore cannot put a countermeasure in place, run at 11 a year. The loss per case (extra sorting, customer reporting, redoing the corrective action) is 28,000 baht. Once you can identify the reel, we assume this falls to 3 a year.

The reduction is 11 minus 3, or 8 cases. The saving is 8 cases times 28,000 baht, or 224,000 baht per year.

Total annual benefit

Adding the four together.

Now the cost side. A traceability investment is not one lump of “system cost.” It splits into five layers, each buying something different. Separating them out lets you judge later, when you narrow the scope, which layers shrink and which do not.

A word on each layer. Layer 1 buys the answer to “what carries the ID,” with the laser marker as the main item, and its unit count barely grows as models are added. Layer 2 buys “taking data automatically from equipment,” and scales with the number of lines. Layer 3 is the ledgers themselves, and database scale does not grow with model count. Layer 4 is “the part people record,” and scales with terminal count, which equals the number of recording points. Layer 5 is “producing the answer,” and depends on how many customer report formats you must support. General price ranges by layer, and where you can realistically cut, are also covered in our article on the cost and approach of building traceability.

Now the three-year TCO. Initial 3,194,000 baht, plus annual running of 321,000 baht times three years (= 963,000 baht), gives 4,157,000 baht.

The three-year benefit, meanwhile, is 816,680 baht times three years, or 2,450,040 baht. The gap is 1,706,960 baht out of pocket. Simple payback is 4,157,000 divided by 816,680, about 5.1 years. It does not pay back in three.

Narrowing to 32 models still does not pay back

“It only fails to pay back because you did it for every model at once — just narrow the scope.” That is the natural reaction. So let us actually narrow it and run the numbers.

We narrow the target to the 32 automotive and medical device models where the customer explicitly requires unit-level trace. Those 32 models account for 45% of board volume and 80% of customer complaints. Because the complaints are so concentrated, the benefit left behind should be small.

Layer by layer, here is what shrinks and what does not.

This is the most overlooked feature of the electronics traceability cost structure. Narrowing the scope only shrinks Layer 2 and Layer 4. Layers 1, 3 and 5 are costs you incur if you do this for even one model, so cutting the target to under a fifth barely moves the number. Initial cost fell from 3,194,000 baht to 2,450,000 baht — only 23%.

Three-year TCO is initial 2,450,000 baht plus annual running 229,000 baht times three years (= 687,000 baht), or 3,137,000 baht.

The benefit side is prorated by the target ratio. Complaints and containment cases concentrate in the models with customer requirements, so 80% applies; MSD is prorated at 45%, the share of board volume those 32 models represent.

The three-year benefit is 598,408 baht times three years, or 1,795,224 baht. It does not reach the three-year TCO of 3,137,000 baht. Simple payback is 3,137,000 divided by 598,408, about 5.2 years. Narrowing does not make it pay back. In fact payback came out slightly longer than the 5.1 years for all models at once, because the layers that do not shrink now carry more weight.

This is where most capital requests die. “Five-year payback” is grounds for rejection at a factory that sets a three-year hurdle for capital investment. And this is exactly where the temptation appears to inflate the benefit estimate until it fits inside three years. Do that and, one year after go-live, someone will say the benefits are not materializing and the budget will stop.

So measure it as an order requirement, not as cost reduction

Let us lead with the conclusion. Electronics traceability is an investment that does not pay back on cost reduction. And that is not because the estimate is bad — it is because the yardstick is wrong.

What this investment actually protects is neither sorting cost nor engineering hours. It is the orders for those 32 models themselves. If automotive and medical device customers are demanding unit-level trace, then failing to provide it means being dropped at the next model change. What you lose is not a cost saving. It is revenue.

So recalculate it this way.

• Set annual revenue for the 32 target models at 86,000,000 baht
• Set the gross margin rate at 14%, giving an annual gross profit of 12,040,000 baht
• Annualize the three-year TCO of 3,137,000 baht, giving 1,045,667 baht
• 1,045,667 divided by 12,040,000 is 8.7%

That 8.7% is the only number that belongs in the capital request. Written out as a sentence, it reads like this.

“If you judge the probability of losing the orders for these 32 models over the next three years, on the grounds of not supporting unit-level trace, to be higher than 8.7%, then this investment stands up.”

Whether you can make that conversion decides whether the request is approved. The question you are putting to the decision maker is no longer “will you approve an investment that pays back in five years?” but “how do you assess an 8.7% probability of losing the business over three years?” The second question is one sales and quality assurance already have an answer to. If trace requirements are already written into the customer specification, 8.7% is a low bar by any reading.

For the record, revenue and gross margin are assumptions. Replace them with the actuals for your own equivalent of those 32 models and recalculate. The formula is just three lines.

• Annual gross profit = annual revenue of target models times gross margin rate
• Annualized cost = three-year TCO divided by three
• Threshold = annualized cost divided by annual gross profit

Factories with a gross margin below 14% get a higher threshold. At a 10% margin, annual gross profit becomes 8,600,000 baht and the threshold is about 12.2%. Even then, “a 12% chance of losing the business over three years” is a far easier question to answer than a five-year payback.

One more thing worth saying plainly. This is not about inflating figures. Keep the 598,408 baht per year of cost reduction exactly as it is. Write both into the request. “Cost reduction of 598,408 baht a year. That alone does not pay back. The substance of this investment is protecting 12,040,000 baht of gross profit, and the threshold is 8.7%.” Written that way, nobody comes back later saying the benefits failed to appear — because you never promised them.

Three issues specific to electronics factories in Thailand

Investment in PCB (printed circuit boards) and PCBA (assembled boards) in Thailand is climbing fast. In its notification of 22 April 2024 (sor.4/2567), the BOI added PCB manufacturing processes — lamination and plating — and the production of PCB raw materials to its list of promoted activities, granting corporate income tax exemption for up to eight years. According to JETRO, combined investment applications across the PCB and PCBA industries expanded 6.3-fold, from 15.9 billion baht (159 hundred million baht) in 2022 to roughly 100.9 billion baht (1,009 hundred million baht) in 2023.

Then in early January 2026 the BOI published Thailand’s first national semiconductor strategy, setting a goal of attracting 2.5 trillion baht in investment by 2050 (Japan Center for Economic Research, 8 April 2026).

More investment means, for an existing Japanese-owned EMS, both more competitors and more customers. There are three practical implications.

First, the more customers you have, the more your requirement levels scatter. Automotive and consumer work riding the same line becomes the normal state. Equipping every model to the highest level is financially unsustainable, and splitting the design per model is operationally unsustainable. Which is precisely why doing the IPC-1782 four-level assignment first is worth the effort. One line, with the depth of recording switched by model attribute.

Second, workforce turnover is fast. A ledger that collapses the moment the quality assurance lead changes will not survive in Thailand. The typical failure mode of personality-dependent spreadsheet operation is columns that keep multiplying. One owner adds a column because “we need this too,” and the next owner has no idea which columns are mandatory. So instead of trying harder at spreadsheets, you need a ledger with fixed columns. Design the change control for adding a column as part of the operation, not as an afterthought.

Third, operation in Thai is a given. Handheld terminal screens, error messages, corrective action entry fields. If those cannot be written in Thai, the fields will be left blank. Tell an operator to describe a defect in English and either nothing gets written or a single meaningless word does. When corrective action records turn out to be useless later, this one point is very often the cause. Enter in Thai, output in Japanese or English when needed. Design in that direction.

What to do in the first 90 days

Do not start from equipment selection. Start from an inventory of customer requirements. Here is a 90-day sequence paired with completion criteria.

There is a reason for that order. Running the reel and MSD ledgers first on days 31-50 works because those two deliver value on their own, with no unit IDs. Even if the unit ID discussion stalls, those two ledgers start turning. Start from unit IDs instead and three months disappear into marker selection and board design changes.

Five common failures on the floor

The same failures repeat, factory after factory. Check all five before you start.

• Deciding the identification method first. Choose marking versus label up front and you will find columns missing from the ledger later. The order is customer requirement, then ledger columns, then identification method.
• Forcing MSD into the inventory system. Floor life is cumulative time, not inventory quantity. Adding a column to the inventory table will not manage it. Stand up a separate table.
• Recording splices on handwritten paper. Splices happen dozens of times a day. Paper will always have gaps. Make it an operation that ends after two scans.
• Adding unit IDs but never linking AOI and ICT judgments. An ID alone, with no process judgment attached, will not narrow the containment scope. Design ID issuance and judgment linkage together.
• Starting with every model at once. All five layers hit maximum and it does not pay back in three years. Start with the models that have customer requirements and expand in the order requirements arrive.

Frequently asked questions

How far back does electronics traceability need to reach to be enough

The customer specification holds the answer. There is no general definition of “enough.” In practice there are three useful stages. Back to the lot (you can state the affected scope by incoming lot), back to the reel (you can separate before and after a splice), and back to the individual board (you can identify one shipped board). Automotive and medical device work typically needs reel to unit level; consumer work is often satisfied at lot level. Start by writing down, model by model, which of the three stages is being demanded.

For PCB serial number tracking, is marking or labeling better

Decide it on whether it survives downstream. If there is a cleaning process, if it passes through reflow twice, if a placed component covers the ID location, then a label will peel or become unreadable from heat. Marking is reliable but requires securing marking space, changing the board design, and investing in a laser marker. In this article’s model case that is the 950,000 baht of Layer 1. On small boards with no space to mark you will end up choosing a label, and in that case measure how far down the process it stays readable before you commit.

Without unit-level trace, how much wider does the containment scope get

Ten times, in the model case. Cutting only at lot granularity, today’s average is 9,000 boards into sorting; narrowed to unit level it becomes 900. At a sorting cost of 12 baht per board, that is 108,000 baht versus 10,800 baht per case. Across 6 cases a year it is a gap of 583,200 baht. To estimate this for your own plant, pull the actual sorted quantities from the last year of containment cases and re-divide them by “how many boards would this have been if we could have narrowed to unit level.”

Does MSD floor life management need a dedicated system

It does not have to be a dedicated system, but you cannot build it inside the inventory system. The primary keys differ. Inventory holds quantity by part number and lot; floor life holds cumulative time per individual reel keyed on opening time. Three reels from the same lot each hold a different remaining time. What you need is a separate table keyed on reel ID and a terminal operation that stamps the opening, return and bake times. At small scale a simple ledger app is enough. What matters is designing it so nobody writes the opening time by hand.

Do we need to get IPC-1782 certified

This article does not assume certification. IPC-1782 defines four levels for materials and processes and scales requirement depth to risk. Its practical use is translating inconsistent customer requirement statements onto a common measuring stick and deciding how deep the records must go for each model. If a customer explicitly demands conformity to a specific standard, follow that requirement statement itself. The specific requirement items are set out in the paid standard document.

How much does EMS factory traceability cost

In this article’s model case, all models at once came to 3,194,000 baht initial and 321,000 baht annual running, for a three-year TCO of 4,157,000 baht. Narrowed to the 32 models with customer requirements it was 2,450,000 baht initial and 229,000 baht annual running, for a three-year TCO of 3,137,000 baht. It varies greatly with line count, model count, and whether data can be pulled from existing equipment. To size it yourself, classify each of the five layers against three questions — does it scale with line count, does it scale with model count, is it required even for a single model — and the narrowed figure becomes readable.

Can this be added to our existing production management system

Layer 3, the ledgers, and Layer 5, the inquiry function, can sometimes be absorbed into an existing system. The MSD ledger is the exception, because its primary key differs and it needs its own table. Layer 2, equipment data acquisition, cannot be estimated without checking the physical equipment, since what data can be extracted depends on the model and generation of your mounters and AOI. As a sequence, fix the column definitions of the four ledgers on paper first, then check which tables of the existing system can hold those columns. With the column definitions locked, the same design document works whether you extend the existing system or build new.

Summary

Designing electronics traceability is not about drawing a lineage of lots. It is about holding four ledgers keyed on time.

• A reel links to a board only through the time of the splice
• MSD needs a cumulative-time ledger keyed on the opening time
• Unit-level boards branch one-to-many from the panel, and if the parent-child link is not kept at that instant there is no way back
• Purchase route means carrying a column for who you bought from, not for the number on the part

Cost splits into five layers, and narrowing the target does not reduce Layers 1, 3 and 5. In the model case, all models at once gave a three-year TCO of 4,157,000 baht against a three-year benefit of 2,450,040 baht; narrowed to 32 models it gave a three-year TCO of 3,137,000 baht against a three-year benefit of 1,795,224 baht. Neither pays back.

So change the yardstick. The annualized 1,045,667 baht is 8.7% of the 12,040,000 baht annual gross profit from those 32 models. If you judge the probability of losing that business over three years, on the grounds of not supporting unit-level trace, to be higher than 8.7%, the investment stands up. Measured as cost reduction it fails; measured as a requirement it passes. Nothing was inflated — the yardstick was simply the wrong one.

If you are considering electronics traceability

You do not need to start by talking about systems. You can start by taking inventory together of which of the four ledgers your plant is currently missing. Does the reel ledger have all its columns? Where is the MSD opening time recorded? Can unit IDs still be read downstream? Is there a supply category column? Answering those four questions alone makes it clear what you should buy next.

TOMAS TECH is based in Bangkok, Thailand, supporting IT and OT adoption on the production floor for Japanese-owned manufacturers. Sometimes the inventory concludes that modifying your existing system is all that is needed. Please feel free to get in touch through our contact page.

References

• ERAI “2024 Annual Report”
• IPC “IPC-1782B Standard for Manufacturing and Supply Chain Traceability of Electronic Products”
• Piek Training “IPC-1782 Standard for Manufacturing and Supply Chain Traceability of Electronic Products”
• Neotel “IPC/JEDEC J-STD-033 A Practical Guide to Moisture Sensitive Device Storage”
• JETRO “Thailand strengthens investment promotion for digital and printed circuit board industries” (11 March 2025)
• Japan Center for Economic Research “Thailand’s semiconductor strategy gets underway” (8 April 2026)