Inventory Discrepancy: Causes and Solutions for Thai Factories

“Every year at the annual stocktake, our book inventory and physical inventory don’t match. We write off the difference without knowing why, and the next year it doesn’t match again.” We hear this often from plant managers, production control managers and accounting staff at Japanese factories in Ayutthaya, Chonburi and Rayong, Thailand. This article sets out the causes of and solutions for inventory discrepancy as a practical procedure for stopping it from recurring on the shop floor.

Here is the conclusion first: inventory discrepancies are not “counting mistakes”. Almost all of them arise structurally from “defects in how transactions are recorded”. That is why simply resetting the numbers to the physical count at a once-a-year wall-to-wall stocktake leaves the same discrepancies appearing every year. There are three countermeasures.

• Measure discrepancies not by “quantity only” but by four elements: quantity, location, unit of measure and part number (define inventory record accuracy within your company)
• Classify discrepancies by cause and fix the recording flow for each cause
• Replace the once-a-year wall-to-wall stocktake with ABC-based cycle counting, so that discrepancies keep being found early, while they are still small

Note that all item counts, amounts, person-hours and accuracy figures relating to “Model Plant A” in this article are original model settings and assumptions created for this article. They are neither industry averages nor survey results. Please read them as a calculation template to be replaced with your own figures.

Why Inventory Discrepancies Are Likely to Increase in 2026

Thailand’s production recovery means more inventory transactions

According to the economic indicators site Trading Economics, Thailand’s Manufacturing Production Index (MPI) rose 4.44% year on year in August 2026, beating the forecast of 1%. It was the second consecutive month of growth. According to the same site, the Ministry of Industry has raised its 2026 production growth forecast from 0.0-0.5% to 2.75-3.55%.

What follows is this article’s own inference: when production recovers, the number of inventory transactions, such as receipts, issues, line feeding, transfers and shipments, increases. More transactions mean more opportunities for paper slips to be left unwritten, for later data entry to fall behind, and for transfers to be posted on only one side. A recording system that just about kept up in the slow season collapsing all at once when the busy season arrives is a common pattern. Now, while production volumes are recovering, is exactly when reviewing the recording system makes sense.

Inventory accuracy ranks near the top of “areas to improve”

“State of Inventory Management 2026”, published in July 2026 by inventory software company inFlow, is a survey of 400 full-time US workers conducted in March 2026. In that survey, 49.5% named inventory accuracy as the “area most in need of improvement”, and 84.8% said they manage inventory with spreadsheets.

This is a US vendor survey, and its figures do not apply directly to factories in Thailand. Even so, the picture of “inventory accuracy being recognized as a problem at many companies, while it is still managed in spreadsheets” will probably look familiar to many people at Japanese factories in Thailand.

Tax digitalization is also advancing

As background, according to tax news sites, on June 16, 2026 the Thai Cabinet approved extending measures that promote the adoption of electronic tax systems such as e-Tax Invoice and e-Withholding Tax (including a 200% deduction for expenses such as the acquisition of hardware and software) until the end of 2027. This is not directly related to inventory discrepancies, but it is worth noting as background that, on the tax side too, measures encouraging the digitalization of records are continuing.

What Is Inventory Discrepancy? Look at Book vs Physical Inventory Differences in “Absolute” Rather Than “Net” Terms

How differences between book and physical inventory arise

An inventory discrepancy is the difference between the inventory in the system or ledger (book inventory) and the physical inventory actually on the shelves. Book inventory is built up as “quantity received − quantity issued”. In other words, whenever a transaction record is missed, delayed or entered with the wrong quantity somewhere, book inventory drifts away from the physical stock by exactly that amount.

The important point here is that most discrepancies are not created on the day of the stocktake. They arise somewhere in day-to-day transactions and are merely “found” on the stocktake day. If you count only once a year, it can take up to a year from when a discrepancy arises to when it is found. Tracing afterwards which transaction a year ago went wrong is practically impossible. That is why many factories end up simply making an adjustment to “match the physical count” without knowing the reason.

Errors arise structurally, not randomly

There is research showing that inventory record errors arise structurally. In a 2008 paper in the academic journal Management Science, DeHoratius and Raman examined about 370,000 inventory records across 37 stores of a single US retailer and reported that 65% were inaccurate. They also showed that the degree of inaccuracy was not random but could be explained by stock auditing practices (a mitigating factor) and by the complexity of the store environment and the distribution structure (aggravating factors).

This is a study of US retail, not figures from manufacturing or from factories in Thailand. The 65% figure cannot be used as a description of actual factory conditions. All we want to borrow here is the idea that “record errors do not happen by chance; they arise structurally from the characteristics of the system and the environment”. If they arise structurally, then unless the structure is fixed, the same discrepancies will keep appearing no matter how many times you take stock.

Assumptions for Model Plant A

The explanation below uses the following model plant (all of these are model settings for this article).

• A Japanese resin molding and assembly plant in Ayutthaya Province, Thailand
• 2,000 items under inventory control (raw materials, purchased parts, work in process and finished goods)
• A once-a-year wall-to-wall stocktake that stops the line for 2 days and takes 20 people × 8 hours × 2 days = 320 person-hours
• Daily receipts and issues are written on paper slips and later re-entered into Excel and the production management system

Looking only at the net amount makes the discrepancy look small

At its last annual stocktake, Model Plant A made the following adjustments (model settings for this article).

From an accounting perspective, what affects profit and loss is the net figure of −1,800,000 THB. But for the purpose of finding the causes of inventory discrepancy, looking only at the net figure is dangerous, because pluses and minuses cancel each other out and make the errors look smaller than they really are.

At Model Plant A, the actual size of the record errors, measured as the sum of absolute values, was 4,600,000 THB, more than twice the net figure. For example, if a part with part number A is mistakenly recorded as an issue of part number B, part A ends up with more physical stock than the books and part B with less. In net terms these offset each other and go unnoticed, but two record errors have occurred. Aggregate discrepancies in absolute terms, not net, and by item. This is the starting point for finding causes.

Regarding this type of recording error, Steven Bragg of the accounting resource site AccountingTools explains that if a transfer, for example between warehouses, is processed on only one side, the same inventory ends up appearing to be in two places, or nowhere at all. It is a typical example of an error that is invisible in net terms.

How to Measure Inventory Record Accuracy (IRA): Four Elements and Tolerances

Inventory record accuracy is “the percentage of items that were completely correct”

To reduce inventory discrepancies, you first need to agree within the company on a yardstick for “how accurate are we”. That yardstick is Inventory Record Accuracy (IRA).

AccountingTools calculates inventory accuracy as “number of completely accurate items ÷ number of items in the sample”, and explains that an item counts as accurate only when all four elements, quantity, location, unit of measure and part number, match. In the site’s example, 8 items were checked and the quantities all matched, but because there were errors in location codes and part numbers, accuracy was rated 0%. The idea is that “the quantity being right” does not by itself make a record accurate.

Tolerances and the “judge by the first count” rule

You also need to decide the range within which a quantity is considered “correct” (the tolerance). ERP product specifications are a useful reference here. The following are introduced not as recommendations for any specific product, but as examples of how to think about designing tolerances.

• Oracle’s product manual (Oracle Inventory User’s Guide) defines, for the difference between book inventory and the physical count, a quantity variance tolerance and a hit/miss tolerance, both as ±% of the system on-hand quantity. If the first count is within tolerance it is a “hit”; if outside, a “miss”. The manual notes that the hit/miss ratio is based on the first count, not on recounts.
• The same manual also has a value-based adjustment value tolerance, where the adjustment value is calculated as “(count quantity − system on-hand quantity) × current item cost”. Adjustments outside tolerance can be configured to require approval.
• Official documentation for Microsoft Dynamics 365 Supply Chain Management explains that deviation limits (maximum percentage and maximum quantity) can be set per worker, and counts outside the limits are held for review. It also explains a design in which the system does not display the expected quantity during counting (so-called blind counting) as a way to prevent deliberate miscounts.

These are product specifications, not industry-standard definitions. Still, three points are useful when creating internal rules: (1) set tolerances by item rank; (2) judge accuracy by the first count, not by “the result after recounting and reconciling”; and (3) make approval and recording of the cause mandatory for discrepancies outside tolerance. If items that matched only after a recount are also counted as “accurate”, problems in the recording system stop showing up in the numbers.

How to think about the target

There is no uniform industry standard for an IRA target. Consulting firm Oliver Wight says its inventory record accuracy course teaches how to maintain a level of 95% or higher through cycle counting, and the introduction to its 2019 white paper states that “world class inventory record accuracy today is a minimum of 99.5%”. This is a consulting firm’s claim, not a statistic. It is realistic to set your own target in stages, looking at item ranks and your current level.

Model Plant A’s trial count: measuring quantity alone overstates accuracy

As a cycle counting trial, Model Plant A sample-counted 200 items (all of the following are model settings for this article).

Of the 162 items whose quantity was within tolerance, 12 were physically on a different shelf, so their location was wrong. No unit of measure or part number errors were found among the remaining 150 items. Therefore 150 items matched on all four elements, and IRA was 75%.

Measured by quantity alone, accuracy was 81%; measured by the four elements, it was 75%, a gap of 6 points. Items where “the quantity is right but it isn’t on the shelf” steal time from the floor at picking, in the form of “the system says it’s there but we can’t find it”, and are sometimes judged to be out of stock, leading to unnecessary orders. If location errors are left out of the accuracy calculation, these losses disappear from the numbers. If you aim to improve inventory record accuracy, aligning the yardstick to the four elements first is essential.

Five Causes of Inventory Discrepancy

Classifying the 50 failed items by cause

At Model Plant A, for the 50 items (200 − 150) on which at least one of the four elements did not match, the team traced back the most recent transactions item by item and classified the causes (model settings for this article).

To clarify how the breakdown fits together: the 38 items (200 − 162) whose quantity was outside tolerance consist of “missing recording 16 + backflushing 12 + unit 6 + counting errors 4”, and the 12 items (162 − 150) whose quantity was correct but whose location was wrong are “wrong location 12”. Together they make 50 items.

92% of discrepancies came from the recording system

What this classification shows is that causes rooted in the recording system (missing recording 16 + backflushing 12 + location 12 + unit 6 = 46 items) account for 92% of the 50 items, while pure counting errors are only 4 items, or 8%.

In other words, no matter how carefully you recount on stocktake day, most discrepancies will not stop. Improving how you count can reduce only the 8% portion; the remaining 92% will keep being generated by the same system from the very next day. This is why the books fail to match in the same way every year.

Mr. Bragg of AccountingTools also lists types of inventory error including counting errors, unit of measure errors, part number errors, customer-owned inventory being counted as your own, consignment inventory being missed, cutoff problems, transfers posted on only one side, and incorrect scrap relief due to backflushing. His inventory reconciliation procedure likewise lists, after recounting and checking units and part numbers, looking for unrecorded transactions and verifying scrap and backflush records. And he writes that transactions that have occurred but that nobody has recorded yet are a major source of discrepancies. In other words, it is the recording flow, more than the count itself, that should be suspected.

Cause 1: Missing or late recording of receipts, issues and transfers

This was the most common cause at Model Plant A. In an operation where paper slips are written first and then entered into Excel or the production management system in the evening or the next day, discrepancies like the following occur routinely.

• A slip is forgotten for an urgent issue, or a slip is written but never entered
• Goods already received stay off the books while awaiting inspection, and are used in the meantime
• In a transfer from the warehouse to a process, only the issuing side is recorded and the receiving side is not
• Goods that arrive in the middle of a stocktake end up in only one of the books or the physical count (a cutoff problem)

In every case, the root cause is a gap between “the moment the goods moved” and “the moment it was recorded”.

Cause 2: Backflushing and unrecorded BOM and scrap

Backflushing (automatic consumption) is a mechanism that, when production output of a finished item is posted, automatically records the consumption of components based on the BOM (bill of materials). It is a convenient mechanism that reduces data entry on the floor, but AccountingTools explains that because it relieves inventory on the assumption that the BOM’s standard component quantities are correct, if actual scrap differs from the standard, the wrong quantity is deducted from the inventory records.

Microsoft’s official documentation (Dynamics 365 Business Central) also lists, as a drawback of automatic flushing the possibility that scrap may not be recorded accurately, or may not even be noticed. Furthermore, according to the same documentation, with backflushing for the whole order that does not use links to individual operations, consumption is not posted until the production order is finished, and in the meantime the components still appear “in stock” to other demand.

In a resin molding plant, if rejects from setting up molding conditions, runners, purge material from color changes, and prototypes or samples handed out are not included in the standard yield in the BOM, book inventory simply remains higher than physical inventory. At plants where work in process stays in the process for a long time, timing gaps add to this.

Cause 3: Wrong location

These are cases where the quantity is correct but the goods are on a different shelf. They occur when there are no bin location labels, when temporary staging areas have become permanent, or when transfers are not recorded at location level. As mentioned above, this cause is invisible if you measure accuracy by quantity alone.

Cause 4: Unit conversion errors

These occur when a single item has multiple units, such as boxes as the purchasing unit, bags as the storage unit, and kg or pieces as the usage unit. AccountingTools gives the example of counting individual units while the unit of measure in the system was dozens. If the conversion factor in the item master is wrong, or if the floor enters only the number without checking the unit, discrepancies of an order of magnitude appear.

Cause 5: Counting errors

These are errors such as miscounting, counting a different item, or counting the same shelf twice. At Model Plant A this was the least common cause, at 4 of the 50 items. However, if counting errors are left unchecked, correct book records get “overwritten” with wrong physical counts, creating new discrepancies. They are kept in check with blind counting, where counters do not see the quantity, and with recounts for items outside tolerance.

Solutions by Cause of Inventory Discrepancy

Countermeasure for missing or late recording: record at the moment the goods move

The most effective step is to stop re-entering paper slips later and instead record on the spot, by scanning barcodes with a handheld terminal or smartphone at the moment the goods move. Scanning also reduces handwriting and transcription errors in part numbers and quantities at the same time. We cover barcode-based recording systems in detail in key points for introducing a barcode management system.

In addition, decide on the following operating rules.

• Scan incoming goods at the time of receipt, and keep goods awaiting inspection or on quality hold in dedicated locations to separate them
• Record transfers between warehouses and between processes as a single transaction covering both issue and receipt, eliminating one-sided postings
• Decide in advance how goods arriving during the stocktake period are treated (which period they belong to)

Countermeasure for backflushing and BOM: close the gap between standard and actual

There is no need to stop backflushing. The first step is to regularly review the BOM component quantities and yields (scrap rates) on which it relies, aligning them with actual results. Then make it a rule that scrap, defects, prototypes and samples handed out are recorded within the process each time they occur. If the system is set up so that consumption can be recorded by operation, the time lag during which stock appears to be “still there” until the order is finished also becomes shorter.

The accuracy of the BOM and item master directly affects not only inventory but also the accuracy of MRP requirements calculations. We discuss this in our article on MRP systems and master data accuracy. For how to think about recording work in process, work-in-process inventory management may also be useful.

Countermeasure for wrong locations: assign bin locations and record by location

Put location labels on every shelf, and scan receipts and issues at location level. If you create a temporary staging area for “putting things here for now”, register that as a location too. Location numbering and operating rules are explained concretely in warehouse location management.

Countermeasure for unit conversion: standardize units in the item master

Standardize the base unit and conversion factors in the item master, and print the unit on labels as well. Official documentation for Dynamics 365 Supply Chain Management includes an example of entering units separately, such as boxes and pieces, and having the system convert and total them (2 boxes and 5 pieces become 25 pieces at a conversion of 10 pieces per box). A design in which the floor does not do conversions and simply enters the units it sees is an effective way of thinking about reducing unit errors.

Countermeasure for counting errors: blind counting and recounts

Blind counting, where counters are not shown the book quantity, prevents people from “writing down a number that matches the books”. Items outside tolerance are recounted by a different person. However, as mentioned above, accuracy is judged by the first count, and an item that matched on recount should not be treated as “accurate”.

Separating customer-owned and consignment inventory

These were not among Model Plant A’s 50 items, but AccountingTools also lists the error of counting customer-owned inventory as your own and errors in handling suppliers’ consignment inventory. At plants that hold customer-supplied materials or supplier-owned consignment stock, these need to be clearly separated by location or inventory category.

From an Annual Stocktake to Cycle Counting: Designing ABC Classes and Tolerances

What is cycle counting?

Oracle’s product manual defines cycle counting as “the periodic counting of individual items throughout the course of the year to ensure the accuracy of inventory quantities and values”. Instead of counting everything at once a year, you count items in rotation, a little every day.

In ERP systems, counting frequency is set by the user. In Oracle, you define cycle count classes by entering, for each ABC class, “how many times per year each item is counted”. SAP’s older help documentation (R/3 4.6C) also explains that materials can be divided into categories such as A, B, C and D, with a counting interval set for each category. SAP S/4HANA learning content describes cycle counting as a procedure in which selected materials are counted several times within a fiscal year, and assumes that all inventory is counted at least once a year.

In ABC analysis, value is often concentrated in a small number of items. In an illustrative example in SAP’s training material, 9 A materials accounted for 56% of consumption value. By counting high-consumption-value items frequently and low-value items less often, you concentrate limited labor where discrepancies have the greatest impact.

Model Plant A’s ABC cycle counting design

Model Plant A divided its 2,000 items as follows and drew up a plan to replace the once-a-year wall-to-wall stocktake. The item proportions below (A 10%, B 20%, C 70%), the annual counting frequencies and the tolerances are all settings for this article’s model plant, not industry standards.

Labor does not go down. Why replace it anyway

Let us calculate the labor (model settings for this article).

• Counts per year: 2,400 + 1,600 + 1,400 = 5,400 counts
• Divided by 240 working days, that is 22.5 counts per day
• At 4 minutes per count, that is 90 minutes per day (22.5 counts × 4 minutes)
• Per year, 5,400 counts × 4 minutes = 21,600 minutes = 360 person-hours

The once-a-year wall-to-wall stocktake took 320 person-hours, so cycle counting actually takes 40 person-hours more. The value of cycle counting is not a reduction in labor. We want to state this honestly. The value of making the switch lies in three things.

1. The line does not have to stop. There is no longer any need to stop the line for 2 days for a wall-to-wall stocktake. The 90 minutes of work per day can be spread across gaps between shipping and production.
2. Discrepancies are found sooner. Because A items are counted 12 times a year, a discrepancy is found within 1 month at most after it arises. With the once-a-year stocktake, it could take up to 1 year.
3. Causes are easier to identify. For A items, when a discrepancy is found you only need to trace back roughly the last month of transactions, making it easier to pinpoint which receipt, issue, transfer or backflush went wrong. Classification by cause, like the 50 items described above, can become a routine monthly task.

Ways to shorten the stocktake itself are covered separately in our article on reducing stocktake time. Please consider them in combination with cycle counting.

Designing tolerances and approvals

At Model Plant A, the tolerance for A items was set at zero quantity variance (exact match), and approval was made mandatory for any adjustment exceeding the value tolerance. B items are ±2% and C items ±5%. Judgment is based on the first count, and entering a cause code is mandatory for discrepancies outside tolerance. With this in place, “how many items failed for which cause” is aggregated automatically every month, and improvement priorities become visible in numbers.

Target after 90 days

Model Plant A set its target after 90 days as an IRA of 95% or higher for A items. As mentioned above, Oliver Wight also presents 95% or higher as one benchmark level. However, this is a target set for the model plant, and it is not a guarantee that it will be achieved simply by working on it. When it can be reached will vary from plant to plant depending on the current level and the breakdown of causes.

A 12-Item Checklist of Inventory Discrepancy Countermeasures

To make it usable on the floor as is, the countermeasures so far are summarized in 12 items.

There is no need to start everything at once. The shortest route is to measure the current state correctly with items 1 and 2, look at the breakdown of causes, and then begin with the items that address the top causes.

Issues Specific to Inventory Discrepancy at Factories in Thailand

Tax: the law sets a deadline for recording inventory

According to the text of the Revenue Code published by Thailand’s Revenue Department, Section 87 requires VAT-registered operators to prepare an output tax report and an input tax report, and additionally requires operators who sell goods to prepare a “report of goods and raw materials”. Entries must be made within 3 business days of the date on which goods or services are acquired or disposed of. Section 87/3 requires these reports and related documents to be retained for at least 5 years.

Here we limit ourselves to introducing the fact that “the law imposes an obligation and a deadline for recording inventory”. The specific methods and formats for recording are to follow criteria set by the Director-General, so please be sure to confirm individually with your accounting firm or tax advisor whether your own operations meet the requirements. That said, at plants where paper slips are batched and entered several days later, it is probably worth checking once whether the timing of entries is consistent with the legal deadline.

ERP vendor Microsoft also provides, as a Thailand-specific feature of Dynamics 365 Finance, a stock card report summarizing inventory movements, quantities and costs by item and warehouse, described as reports that must be produced when the government requests them. This is vendor documentation, not a primary source setting out legal requirements, but it is a useful reference for keeping inventory movements explainable by item and by warehouse.

Thai accounting standards also include TAS 2 (Inventories), which deals with inventories. For the valuation and accounting treatment of inventories, and for how annual physical counts are treated in audits, please check with your auditor or accounting firm.

BOI: company inventory, physical inventory and the BOI balance don’t match

At plants that receive import duty exemptions on raw materials under BOI (Thailand Board of Investment) incentives, there is a third figure, the BOI raw material balance, in addition to inventory in the company system and physical inventory.

According to an explanatory document published by the BOI in January 2020, management of BOI raw materials (the raw material master list, MML) is carried out on RMTS, an online system operated by the Investor Club Association (IC). The document recommends downloading the MML from RMTS every 6 months and reconciling BOI inventory with the company’s physical inventory. It also states that to write off BOI inventory for exported goods, export documents must be submitted to the BOI within 1 year of the export date.

If the company’s inventory records are off, they cannot be reconciled with the BOI balance either. Put the other way around, reducing internal inventory discrepancies is also a precondition for making BOI balance management easier. Note, however, that this document dates from 2020. Please be sure to confirm current operations and procedures with the BOI and the Investor Club Association.

Multilingual shop floors: Thai screens, labels and cause codes

Even if you change the recording system, it will not take hold unless the Thai staff on the floor can use it. Prepare handheld terminal screens, location labels and cause code options in Thai. The key point, especially for cause codes, is to make them selectable options rather than free text. If options such as “missing recording”, “unrecorded scrap”, “wrong location”, “wrong unit” and “counting error” are defined in advance, discrepancies can be aggregated by cause each month, and Japanese managers and Thai leaders can discuss improvements using the same numbers.

Before transaction volumes grow with the production recovery

As mentioned at the beginning, Thailand’s Manufacturing Production Index rose year on year in August 2026, and the Ministry of Industry has raised its annual production outlook. As production volumes increase, so does the number of receipt and issue transactions. Weak points in the recording system are the first to break down when volumes rise. Because it is hard to tackle this once things get busy, we recommend reviewing your recording flows before production fully recovers.

A 90-Day Plan for Implementing Inventory Discrepancy Countermeasures

Using Model Plant A as an example, here is how to proceed over 90 days. The period divisions and workloads are rough guides; please adjust them to your plant’s size and current situation.

Days 0-30: measure and classify

What you should do in the first 30 days is not implement improvements, but measure the current state correctly. Simply re-aggregating the adjustments from the last annual stocktake by item, in absolute rather than net terms, will start to show which groups of items have large discrepancies. Then run a trial count of around 200 items against the four elements, and for the items that fail, trace back the most recent transactions and classify the causes.

Days 31-60: fix the top 2 causes first

Once causes are classified, fix the recording flows for the top 2 causes by number of items. At Model Plant A, that would be missing recording (16 items) and backflushing and BOM (12 items), or alternatively wrong location (12 items), which is easy to tackle right away. Trying to fix every cause at once puts a heavy load on the floor and tends to leave everything half done.

Days 61-90: make cycle counting part of daily work

In the final 30 days, move cycle counting from trial to full operation. Build the daily count (22.5 counts, about 90 minutes, at Model Plant A) into the responsible staff’s daily work, and make approval and cause codes mandatory for discrepancies outside tolerance. Once a month, aggregate and review discrepancies by cause, and decide which recording flow to fix next. Once this cycle is running, discrepancies change from “something found in one batch once a year” to “something found and fixed every month”.

If the work involves moving from paper and Excel to a system, please also refer to how to implement an inventory management system.

Frequently Asked Questions (FAQ)

What are the main causes of inventory discrepancy?

The main cause is defects in the recording system rather than counting mistakes. Specifically, causes can be divided into five: missing or late recording of receipts, issues and transfers; backflushing and unrecorded BOM and scrap; wrong locations; unit conversion errors; and counting errors. At this article’s Model Plant A, 46 of the 50 failed items (92%) were attributable to the recording system, and counting errors accounted for 4 items (8%) (model settings for this article).

How much inventory variance at stocktake is acceptable?

There is no uniform industry standard. It is something each company decides internally by item rank. Oracle’s product manual defines the quantity variance tolerance as ±% of the system on-hand quantity, and adjustments exceeding a value-based tolerance can be set to require approval. At this article’s Model Plant A, A items require an exact match, B items ±2% and C items ±5%. These are model settings, not recommended values.

How is inventory record accuracy (IRA) calculated?

It is calculated as “number of completely accurate items ÷ number of items counted”. As explained by AccountingTools, an item is “accurate” only when all four elements, quantity, location, unit of measure and part number, match. Judging by quantity alone overstates accuracy. At Model Plant A, accuracy was 81% by quantity alone and 75% by the four elements.

What is cycle counting? Does it make the annual stocktake unnecessary?

Cycle counting is a method of counting items in rotation, a little at a time, throughout the year. Using ABC analysis, higher-value items are counted more frequently. Whether the once-a-year wall-to-wall stocktake can be replaced by cycle counting depends on financial audit and tax requirements. Please be sure to confirm individually with your auditor, accounting firm and tax advisor whether your company may stop its wall-to-wall stocktake.

Do I need a system to address inventory discrepancies? Can’t it be done in Excel?

The issue is less the tool itself than the timing and granularity of recording. If you can record at the moment the goods move, at location level and in the correct unit, the approach can work. However, with paper slips re-entered into Excel later, timing slips, and recording down to locations and cause codes tends to become difficult. Note also that in inventory software company inFlow’s US survey (2026), 84.8% said they manage inventory with spreadsheets, so managing inventory in spreadsheets is not unusual in itself.

What should I do when the BOI raw material balance and company inventory don’t match?

The precondition is first to reduce the discrepancy between company system inventory and physical inventory. On that basis, reconcile regularly with the BOI balance. The BOI’s 2020 explanatory document describes downloading the MML from RMTS every 6 months for reconciliation, and submitting export documents within 1 year of the export date to write off BOI inventory. Please be sure to confirm current procedures with the BOI and the Investor Club Association.

Summary

Inventory discrepancies do not arise from miscounts on stocktake day; they arise structurally from defects in day-to-day recording systems. That is why simply resetting to the physical count at a once-a-year wall-to-wall stocktake leaves the same discrepancies appearing every year.

At this article’s Model Plant A, accuracy of 81% measured by quantity alone fell to 75% when measured by the four elements. Of the 50 failed items, 92% were attributable to the recording system, and counting errors were only 8%. And while replacing the once-a-year wall-to-wall stocktake (320 person-hours) with ABC-based cycle counting (5,400 counts and 360 person-hours per year) does not reduce labor, it makes it possible to find discrepancies in A items within 1 month at most without stopping the line, and makes them easier to trace to a cause and fix.

Start by re-splitting the discrepancies from your last stocktake by cause, in absolute rather than net terms. Once you can see the breakdown of discrepancies, which recording flow to fix first becomes clear naturally.

At TOMAS TECH, we are happy to help from the very first steps, such as re-splitting the breakdown of your last stocktake discrepancies by cause or designing a trial count of 200 items. If you are struggling because you cannot tell why your book and physical inventory don’t match, please feel free to contact us through our contact form.

References

• AccountingTools (Inventory accuracy formula): https://www.accountingtools.com/articles/inventory-accuracy-formula
• AccountingTools (Types of inventory errors): https://www.accountingtools.com/articles/types-of-inventory-errors
• AccountingTools (How do I reconcile inventory): https://www.accountingtools.com/articles/how-do-i-reconcile-inventory.html
• Oracle Inventory User’s Guide Release 12.1 (Cycle Counting): https://docs.oracle.com/cd/E18727_01/doc.121/e13450/T291651T292438.htm
• Microsoft Learn (Dynamics 365 Supply Chain Management, Cycle counting): https://learn.microsoft.com/en-us/dynamics365/supply-chain/warehousing/cycle-counting
• Microsoft Learn (Dynamics 365 Business Central, Flush components according to operation output): https://learn.microsoft.com/en-us/dynamics365/business-central/production-how-to-flush-components-according-to-operation-output
• Microsoft Learn (Dynamics 365 Finance, Thailand stock card reports): https://learn.microsoft.com/en-us/dynamics365/finance/localizations/thailand/apac-tha-stock-card-reports
• SAP Learning (Applying Cycle Counting): https://learning.sap.com/courses/inventory-management-and-physical-inventory-in-sap-s-4hana/applying-cycle-counting-optional
• SAP Help (R/3 4.6C, Cycle Counting): https://help.sap.com/saphelp_46c/helpdata/en/4d/2b8e3d43ad11d189410000e829fbbd/content.htm
• Oliver Wight (6 Steps to 95% Inventory Record Accuracy): https://www.oliverwight-americas.com/whitepapers/laying-the-foundation-6-steps-to-95-inventory-record-accuracy-guaranteed
• Oliver Wight (Inventory Record Accuracy Course): https://www.oliverwight-americas.com/private-courses/inventory-record-accuracy-course/
• DeHoratius & Raman (Management Science, 2008): https://ideas.repec.org/a/inm/ormnsc/v54y2008i4p627-641.html
• Thailand Revenue Department (Revenue Code, Sections 87 to 90): https://rd.go.th/5209.html
• Thailand Board of Investment (BOI), Well Management of BOI Incentives on Importing Machinery and Raw Materials (2020): https://www.boi.go.th/upload/content/4.%20Well%20Management%20of%20BOI%20Incentives%20on%20Importing%20Machinery%20and%20Raw%20Materials_5e270c2b4f2b2.pdf
• RSM Thailand (revisions to Thai accounting standards): https://rsm.global/thailand/node/117
• inFlow Inventory (State of Inventory Management 2026): https://www.inflowinventory.com/blog/state-of-inventory-management-2026/
• Trading Economics (Thailand Industrial Production): https://tradingeconomics.com/thailand/industrial-production/news/587981
• Orbitax (Thailand Extends Incentives): https://orbitax.com/news/country/article/Thailand-Extends-Incentives-to-62440
• Fiscal Requirements (Thailand’s e-invoicing remains voluntary): https://www.fiscal-requirements.com/news/5729-thailands-e-invoicing-remains-voluntary-key-20262027-updates-and-tax-incentives