Stocktaking Efficiency 2026 | Why Counting Faster Won’t Help

When a plant sets out to improve stocktaking efficiency, the first idea on the table is almost always hardware: handheld terminals, or RFID. Yet it is entirely common for a site to triple its counting speed and still report much the same inventory discrepancy rate the following period. This article breaks the true cost of a physical inventory count down into baht, using a model factory in Thailand, and shows where three quarters of that cost is actually buried. It then sets out a three-layer model for sequencing the work in the order that pays.

Why counting faster does not shrink inventory discrepancies

Stocktaking efficiency has a return-on-investment sequence. On the shop floor the counting itself feels like the hardest part, so that is where teams want to start. But once you price the cost structure out, the counting labour turns out to be a small slice of the total.

The sequence that works has three layers.

Layer 1 | Reach a state where you do not need to count

Location management, item tags on the physical stock, and recording receipts and issues at the point they happen. If the book inventory does not keep up with the physical stock, no amount of recounting stops the same discrepancy from reappearing next month. This layer is not a stocktaking initiative; it is a design decision about daily operations.

Layer 2 | Reduce the scope you have to count

ABC classification combined with cycle counting. Eliminating the day you shut the line down for a full wall-to-wall count is this layer’s job.

Layer 3 | Increase the speed at which you count

Handheld terminals and RFID. This layer shortens the read time per item.

The problem is that most sites skip Layers 1 and 2 and enter at Layer 3. The result: the plant can now count faster, the inventory discrepancy has not moved, and nobody can justify the investment.

Why does that happen? The next section breaks the cost of stocktaking into four components and prices each one.

Breaking stocktaking cost into four components | the Model Factory A calculation

To keep the discussion concrete, take Model Factory A: a Japanese-owned parts machining and assembly plant on Thailand’s Eastern Seaboard. Your own numbers will differ, but the shape of the cost mix tends to look similar across sites.

Assumptions

The 520 THB per person-day figure starts from the fact that Thailand’s minimum wage sits roughly in the 337 to 400 THB per day band depending on the province. We conservatively take 400 THB per day, the top of that band, and add statutory employer costs such as social security contributions and overtime premiums. The top band applies only to certain provinces and occupations, so some sites will be below this. Replace it with your own payroll data, since the applicable amount varies by location and job grade.

The four costs (annual)

The cost of a physical inventory count is not just the cost of counting. Split it into the cost of stopping, the cost of counting, the cost of correcting, and the cost of paying, and you have what you need for an investment decision.

Each line can be checked on a calculator.

• Cost of stopping: daily output value 1,200,000 THB × contribution margin 25% = 300,000 THB per count. Twice a year gives 600,000 THB.
• Cost of counting: 30 people × 2 days = 60 person-days. 60 person-days × 520 THB = 31,200 THB per count. Twice a year gives 62,400 THB.
• Cost of correcting: 3 people × 5 days = 15 person-days. 15 person-days × 1,200 THB = 18,000 THB per count. Twice a year gives 36,000 THB.
• Cost of paying: 97,920 THB per year, from the breakdown below.

The total is 600,000 + 62,400 + 36,000 + 97,920 = 796,320 THB per year (the shares do not add to exactly 100% because of rounding). The shares are 600,000 ÷ 796,320 = 75.3%, 62,400 ÷ 796,320 = 7.8%, 36,000 ÷ 796,320 = 4.5%, and 97,920 ÷ 796,320 = 12.3%.

Inside the cost of paying

At a Thai site, an inventory discrepancy does not stay an invisible cost. It surfaces as tax you actually pay (covered in detail in a later section).

• Discrepancy = 30,000,000 THB × 0.8% = 240,000 THB per count. The shortage side is 60% of that = 144,000 THB per count (288,000 THB a year)
• VAT on the deemed sale: 144,000 × 7% = 10,080 THB per count, or 20,160 THB a year
• Lost tax shield because the shortage is non-deductible: 144,000 × corporate income tax 20% = 28,800 THB per count, or 57,600 THB a year
• Penalties and surcharges: 20,160 THB a year allowed for conservatively (the mid-point of a range running from 20% on voluntary disclosure to 60% when raised by the authorities)

Total: 20,160 + 57,600 + 20,160 = 97,920 THB a year.

What this calculation tells you

This is the crux of the article.

What RFID or a handheld terminal can cut is the 62,400 THB cost of counting, just under 8% of the total. Even halving the reading work saves a little over 30,000 THB a year. That does not pay back an investment in the millions of baht.

By contrast, moving from a full wall-to-wall count to cycle counting removes the 600,000 THB cost of stopping, and that alone is three quarters of the total. Layer 1, real-time recording and location management, attacks the 36,000 THB cost of correcting plus the 97,920 THB cost of paying, or 133,920 THB. Layers 1 and 2 together put 733,920 THB, 92% of the total, within reach. Strictly speaking, cycle counting introduces its own routine counting hours. But because it does not stop the line, the net saving remains large.

That is why entering at Layer 3 fails to produce a return. It is the reason so many stocktaking efficiency projects end with “we got faster, but the discrepancy did not move.”

Note that this calculation rests on a 25% contribution margin, two counts a year, and a 0.8% discrepancy rate. Substituting your own numbers and recalculating is itself the first step toward efficiency. Bringing down the inventory value itself is a separate angle, covered in optimal inventory level management.

Layer 1 | Reach a state where you do not need to count: five causes of inventory discrepancy

To unpack “counting faster does not shrink discrepancies,” look at it from the cause side. Discrepancies found in a physical inventory count fall into roughly five categories.

The key point is that only categories 3 and 4 are found on count day. Categories 1, 2 and 5 are records-side problems, so counting faster or counting more often still leaves the same discrepancy next period. A handheld terminal improves detection of 3 and 4; it does not touch 1, 2 or 5.

Four moves that make up Layer 1

(1) Put location management on every SKU

Give every item an address down to rack, level and bay, and hold that location against the stock in the system. This is not only about cutting search time. Being able to state exactly what you counted is a precondition for the cycle counting described later. If you start counting only part of the stock without locations, you cannot describe the scope of what was not counted, and the approach fails both audit and tax scrutiny.

(2) Standardise item tags and pack-quantity labelling

Wrong unit or pack quantity (category 4) usually traces back to item tag formats that differ from process to process. Manage the conversion factors for pieces, boxes, reels and pallets centrally in the master, and always print a barcode and the pack quantity on the tag. Without that alignment, bringing in handhelds at Layer 3 simply reproduces the same errors in the quantity calculation after each read.

(3) Record receipts and issues where they happen

Collecting paperwork and keying it in later at the office is what breeds categories 1 and 2. Record receipts at the goods acceptance point and issues at the point of line feed. Only once this closes daily do you have a real cut-off. An inventory record without a working cut-off is one an auditor cannot rely on, even after you move to cycle counting.

(4) Build a route that records physical stock leaving

Scrap, returns and sample removals get treated as exceptions and pushed outside the standard flow, but in baht terms they are not negligible. Put approval and posting on a single track so that whenever physical stock moves, a record moves with it. Scrapping in particular carries procedural requirements for tax purposes in Thailand, covered in a later section.

All of this ties directly to how the inventory management system is designed. To judge whether your existing system can support it, the criteria in comparing inventory management systems are a useful reference.

How to tell Layer 1 is done

Three tests:

• For any item, someone can answer on the spot when asked for the system quantity and the location
• All of the previous day’s receipts and issues are posted by the following business morning
• The most recent discrepancies can be tallied by category, and the share of categories 1, 2 and 5 is known

The third one is where a surprising number of sites fall short. If you report “discrepancy rate 0.8%” without classifying the causes, you will never settle which layer deserves the investment.

Layer 2 | Reduce the scope you count: from wall-to-wall to cycle counting

The structural cost of a full wall-to-wall count

A full wall-to-wall count, where every item is counted at the same time, is the clearest way to demonstrate completeness of period-end inventory. Structurally, though, it generates three things at once.

• Production stops: you cannot count while stock is moving, so the line goes down. At Model Factory A, that alone is 600,000 THB a year.
• A large temporary headcount: 30 people pulled off their normal work for two days. The knock-on effect of that work stopping is not in the calculation, but in practice it is not negligible.
• A concentrated pile of after-the-fact investigation: six months of discrepancies land at once, so you investigate at exactly the moment when traceability is at its worst. Chase a document from six months ago and nobody remembers.

The third is easy to overlook and the most fundamental. A wall-to-wall count generates the largest volume of discrepancies at the moment they are hardest to investigate.

Conditions for moving to cycle counting

Cycle counting splits the item list and counts continuously through the year. But simply “splitting it up and counting” is not enough for audit or tax purposes. Six conditions are the minimum.

Conditions 1 and 2 are Layer 1 restated. You cannot skip Layer 1 and move to Layer 2 — that is a procedural requirement, not a matter of principle.

Condition 6 deserves particular attention. A change of method presumes agreement with your auditor beforehand. Push ahead on the site’s own judgement and you risk hearing, at period end, that the results cannot be relied on and a full count has to be redone. Thai Financial Reporting Standards (TFRS) are built on IFRS, but how physical count procedures are evaluated sits within the auditor’s judgement. The practical route is to put a policy paper in front of your audit firm while you are still considering the move, and get their view.

Designing ABC classification and count frequency

Cycle counting design starts from not counting every item at the same frequency. Classify items by value, turnover, and the line impact of a stockout, then vary the frequency.

ABC classification is value-based by default, but in practice it pays to treat “items that stop the line when they run out” as class A regardless of value. Even a low-unit-cost part that blocks shipment when missing carries high importance for inventory accuracy.

The commonly cited target range for inventory record accuracy (IRA) is 95 to 98%, with best practice put at above 99% for class A items. Start by measuring your current IRA, then set targets by class. A single company-wide target gets dragged around by the large number of class C items and blurs the focus of improvement.

One more point: blind counting, where the counter is not shown the system quantity, is regarded as fundamental to protecting accuracy. When the theoretical quantity is visible, there is an unconscious bias toward making the count agree with it. When specifying handheld terminals, whether the count entry screen can be configured to hide the book quantity is a requirement worth confirming.

A realistic transition path

Rather than abolishing the wall-to-wall count overnight, run both in parallel for a period. Typically, in year one you cut the wall-to-wall count from twice to once a year and run cycle counting alongside it, checking that the results agree. Being able to show your auditor a track record where cycle count results matched the wall-to-wall results makes the reliance decision far easier to obtain the following year.

Layer 3 | Increase counting speed: judging handheld terminals and RFID

Once Layers 1 and 2 are moving, Layer 3 investment starts to mean something. The order matters.

How to read the published case study

One published case of RFID-based stocktaking automation is Nagano FCL Components Co., Ltd., a manufacturer of electronic components and electronic application equipment. Counting roughly 30,000 chip reels reportedly took 10 people and about 2 days before, and 6 people and about 2 hours after. The company also reports that it was able to raise count frequency from four times a year to once a month.

The thing to notice is not the reduction in working hours. It is the move from quarterly to monthly. Triple the frequency and the interval between a discrepancy occurring and being detected shrinks from three months to one, which improves traceability of the cause dramatically. In substance, that site moved to something close to cycle counting.

In other words, Layer 3 investment earns its keep when the argument is not “we got faster so we need fewer people” but “we got faster so we can raise the frequency, which reaches into Layer 2.” Whether the business case makes that translation changes what the approval paper actually says. The operational design on the handheld side is covered in handheld terminal deployment, and the cost structure of RFID in RFID deployment cost.

Choosing between handheld terminals and RFID

The practical decision turns not on SKU count but on the number of pieces read in a single count. Where bulk reading without unpacking carries real value, as with 30,000 chip reels, RFID pays. Where stock can be counted at box level, barcodes and handhelds get you comfortably into Layer 2.

Three places where RFID cost jumps or the project fails

Three items routinely go missing from RFID estimates.

(1) Tag unit price × item count, and the labour of applying the tags

Even at a few baht per tag, tens of thousands of pieces add up. What gets overlooked more often is the labour cost of applying them. Tagging all existing stock as an initial exercise can amount to as much work as a full wall-to-wall count in its own right. After go-live, every receipt adds a tagging step, so that labour has to be built in as a running cost.

(2) Verifying the RF environment on site

Metal racking, stock containing liquids, and high-density storage all degrade RFID read accuracy. Failing to read is one problem; reading the next rack as well, the over-read, is just as serious. An over-read shows as a surplus at that location and a shortage at the neighbouring one, so it cancels out and never appears in the total-quantity discrepancy while quietly building up wrong location data. Assume the read range in the catalogue will not reproduce in your own racking.

(3) Reconciliation with the existing inventory system or WMS

Read does not equal posted. Unless the tag ID the reader picked up maps correctly to the item code, lot and location in the system, the count result cannot be reflected in the inventory record. Layer 1 master data work is a direct precondition here. Introduce RFID while item master units and pack quantities are still inconsistent and you end up with excellent read accuracy and quantities that still do not agree.

RFID therefore has to be run on the assumption of staged verification, from a limited area to the whole site. Verify the worst-case area first (metal racking, high-density storage), confirm the read rate holds there, and only then decide on full rollout. Do it the other way round and you discover the unreadable areas after tagging everything.

The Thailand-specific pressure point | inventory discrepancy surfaces as tax

From here the discussion is specific to manufacturers with operations in Thailand. The same inventory discrepancy is understood to be treated differently for accounting and tax purposes in Japan and in Thailand, and that difference is the basis for the cost of paying in the calculation above.

Note: what follows reflects practitioner understanding based on commentary published by accounting firms, not a verbatim citation of legislation. Whether it applies depends on your own circumstances, so always confirm individually with your own auditor and tax adviser.

Inventory shrinkage is generally understood to be non-deductible

In Japan there is scope to treat inventory reductions identified in a physical count as an inventory shrinkage expense for tax purposes. In Thailand, by contrast, a shortfall of physical stock against book stock (shrinkage or loss) is generally described as not deductible for corporate income tax purposes.

Practitioners also commonly describe the shortfall as deemed to have been sold, requiring VAT (7%) to be declared and paid. In Thailand, then, an inventory discrepancy is not an invisible cost. It surfaces as tax you actually pay.

Returning to Model Factory A: against the 144,000 THB shortage side per count, the calculation produced 10,080 THB of VAT per count and 28,800 THB of lost tax shield from non-deductibility per count. Over two counts a year that is 57,600 + 20,160 = 77,760 THB. Adding the 20,160 THB allowance for penalties and surcharges gives the 97,920 THB figure.

Scrapping inventory is understood to carry procedural requirements

One cause of inventory discrepancy is scrap that was never processed (category 5). In Thailand, scrapping inventory is described as something that cannot proceed on the shop floor’s judgement alone. The distinction is commonly explained as follows.

Revenue Department regulation ป.79/2541 is frequently cited as the basis. In addition, a police report is said to be required before a loss can be deducted.

Companies with BOI promotion and companies inside IEAT industrial estates are understood to face separate procedures under their respective rules (attendance of an accountant, preparation of a scrapping report, and so on). Because tax privileges and bonded frameworks are tied directly to the inventory books, requirements are tighter than for an ordinary company, and that is worth knowing in advance.

The level of penalties

For voluntary VAT disclosure, the penalty levels cited are 20% within two years of registration, 40% beyond two years, and 60% where the authorities raise the point. What matters in practice is that something costing 20 to 40% on voluntary disclosure becomes 60% once it is raised for you — a 1.5 to 3 times spread.

A structure that finds discrepancies early enough to deal with them voluntarily translates directly into a difference in tax cost. A site that surfaces six months of discrepancies at once in a half-yearly wall-to-wall count and a site that sees them monthly through cycle counting are not in the same position here. Cycle counting is also a tax risk management tool, and that framing works well with senior management.

Bringing this section together

At a Thai site, stocktaking efficiency is a productivity question and a tax cost question at the same time. That said, the treatments above reflect understanding based on practitioner commentary, and application varies with your own circumstances (industry, BOI status, the history of past tax audits, and so on). Before you set a policy, always confirm individually with your own auditor and tax adviser. The calculation in this article is intended as the starting point for that discussion, not a substitute for it.

Sequencing the three layers over 12 months

How do you actually start? Dropping the priorities into a 12-month set of phases gives the following.

Keeping to this order is about more than return on investment. Layers 1 and 2 can largely be delivered within operational design and system configuration — that is, without major capital expenditure you reach the 92% of total cost that those two layers address. Layer 3 investment is easier to get approved once the foundation exists to measure its effect.

One more benefit: as inventory accuracy improves, outbound quality improves with it. Location management and standardised item tags are also the basis of preventing shipping errors, so the case is more persuasive when presented as an investment in warehouse operations as a whole rather than in stocktaking alone.

Frequently asked questions

What are the conditions for switching to cycle counting?

Six, as a minimum. (1) Location management on every SKU, (2) receipts and issues recorded the same day so a cut-off holds, (3) count frequency per item documented as policy, (4) a defined flow for investigating, approving and correcting discrepancies, (5) the ability to show at period end that every item was counted at least once in the past 12 months, and (6) the auditor concluding they can rely on the cycle count results. Number 6 in particular requires discussion in advance and is not something the site can decide alone.

How should we estimate the cost of RFID stocktaking?

Estimating on tag unit price × item count alone will miss the mark. In practice you build up five items: (1) tag unit price × item count, (2) the labour of tagging existing stock plus the ongoing labour of tagging at receipt, (3) hardware such as readers, (4) integration development with the existing inventory system or WMS, and (5) the effort of verifying the RF environment. Items (2) and (5) are the ones most often missing from a first estimate. Start with staged verification in one area, obtain actual read rates and cost per unit, and decide on full rollout from there.

Is it wrong to aim only at shortening stocktaking time?

There is nothing wrong with the objective, but the return is limited. In the Model Factory A calculation, stocktaking labour (the cost of counting) is 62,400 THB a year, just 7.8% of the total. Even halving the reading work saves a little over 30,000 THB a year. Lost profit from line stoppage (the cost of stopping), by contrast, is 600,000 THB a year, 75.3% of the total. Only when you convert the time saving into higher frequency, move to cycle counting and eliminate the stoppage day does the investment come into payback range.

How are inventory discrepancies treated in Thailand?

A shortfall of physical stock against book stock is generally described as not deductible for corporate income tax purposes, and the shortfall is commonly explained as deemed sold, requiring VAT (7%) to be declared and paid. For scrapping, storable inventory is said to require internal approval, the attendance of an accountant and the attendance of a Revenue Department officer, with notification to the Revenue Department at least 30 days in advance. All of this reflects understanding based on practitioner commentary, and application varies with your own circumstances. Always confirm individually with your own auditor and tax adviser.

How do we identify the cause of an inventory discrepancy?

Tallying discrepancies by value or rate tells you nothing about cause. Classify them into the five categories: receipts unposted, issues unposted, wrong location, wrong unit or pack quantity, and physical stock leaving unrecorded. Of these, only wrong location and wrong unit or pack quantity are detectable on count day; the other three are records-side problems. Once you know the mix by category, you can decide with numbers whether to invest in Layer 1 (real-time recording) or Layer 3 (faster reading).

What should we target for inventory record accuracy (IRA)?

The commonly cited target range is 95 to 98%, with best practice put at above 99% for class A items. A single company-wide target, however, gets dragged around by the large number of class C items and blurs the focus of improvement. Set separate targets by ABC class and raise accuracy starting with class A. For measurement, blind counting, where the system’s theoretical quantity is not shown to the counter, is regarded as fundamental.

Should we look at handheld terminals or RFID first?

The deciding axis is not SKU count but the number of pieces read in a single count. Where stock can be counted at box level, barcodes and handhelds reach cycle counting comfortably. Where you want to bulk-read tens of thousands of pieces without unpacking, such as chip reels, RFID shows its value. In either case, consistent item master units and pack quantities plus location management are preconditions. Deploy on top of an untidy master and you get high read accuracy with quantities that still do not agree.

Summary

Stocktaking efficiency looks like a question of counting speed, but what actually works runs in the opposite order.

• Layer 1 | Reach a state where you do not need to count: location management, standardised item tags and pack-quantity master, real-time recording of receipts and issues. Of the five categories of inventory discrepancy, only two are found on count day. The other three are records-side problems that changing how you count will not solve.
• Layer 2 | Reduce the scope you count: ABC classification and cycle counting. At Model Factory A, the 600,000 THB of lost profit from line stoppage is 75.3% of total stocktaking cost. This is the only layer that can remove it. The transition, though, requires six conditions and advance agreement from the auditor.
• Layer 3 | Increase counting speed: handheld terminals and RFID. What can be cut is the 62,400 THB that represents 7.8% of the total. The value of the investment lies less in reducing headcount than in raising frequency, which reaches into Layer 2.

And at a Thai site, an inventory discrepancy is understood to appear as tax actually paid, through non-deductibility and VAT on a deemed sale. At Model Factory A that is 97,920 THB a year, 12.3% of the total. A structure that finds discrepancies early enough to handle them voluntarily is, in itself, a way of managing tax cost.

The first thing to tackle is not the investment decision but the baseline. Calculate the four costs with your own numbers, and classify your recent discrepancies into the five categories. Once you have those two, which layer to invest in decides itself.

If you are at the stage of wanting to establish that baseline with your own numbers, or to judge whether you are ready to move to cycle counting, we are happy to talk. TOMAS TECH implements production and inventory management systems for Japanese-owned plants in Thailand and designs how they run on the floor, and we typically start by mapping your current stocktaking process and the state of your inventory records, then working out with you which of the three layers to start from. Organising the information before any decision on deployment is a perfectly good place to begin, so please get in touch through our contact form.

References

• SATO Case Study | Nagano FCL Components Co., Ltd. (RFID stocktaking)
• Toshiba Tec | Improving stocktaking and inventory management with RFID solutions
• Scroll 360 | What is cycle counting
• Asprova MRP Glossary | Cycle counting
• NetSuite | Inventory Cycle Counting 101
• Kaipro | Tax treatment of disposal, retirement and sale of fixed assets in Thailand
• JETRO | Thailand tax system (corporate income tax 20%, VAT 7%)
• IFRS Foundation | Thailand jurisdiction profile (TFRS)
• Thai Law Online | Minimum Wage in Thailand