WIP Management in Thai Factories 2026 | Queues Set Lead Time

“We want to manage work-in-process properly. Where should we start?” This is one of the most common questions we hear at Japanese-owned factories in Thailand, and our answer is always the same. Start by counting how many pieces are sitting between one process and the next. WIP is inventory whose cost has already been incurred and which has not yet turned into revenue. In this article we follow a model case at an automotive parts machining plant in Chonburi Province, look at how the 24,000 pieces stacked between its 5 processes were counted, and see how far they were reduced.

Why WIP Quietly Erodes Factory Profit — You Cannot Reduce Inventory You Cannot See

Ask a plant manager “how much WIP do you have right now” and almost nobody can answer on the spot. What comes back is “I can give you the figure from the month-end physical count.” In other words, the plant knows how many pieces there were at the end of last month, not how many are on the floor today.

This is not a story about sloppy record keeping. If anything, the opposite is true. Month-end counts are done meticulously. The line is stopped, everyone available is pulled in, physical pieces are counted, the numbers are typed into Excel and reconciled with the books. It takes days. Precisely because it is done carefully, it can only be done once a month. And because it is done only once a month, nobody knows what happened in between.

What makes WIP awkward is that when it grows, nobody suffers. If raw material runs short the line stops and everyone notices immediately. If finished goods pile up unsold, sales gets involved. WIP, by contrast, simply sits between processes and blocks nobody’s work. If anything, having a comfortable stack of material ready for the next process feels reassuring on the shop floor.

From an accounting perspective, though, WIP is an asset into which material cost, labour cost and processing cost have already been poured. Cash has changed form and is now lying on the factory floor. Until it is sold and collected, that cash cannot be used. A plant with high WIP is not necessarily a plant with poor margins; it is a plant tying up more capital to produce the same profit.

The loss caused by excess WIP never appears in the financial statements under a line item called “loss from work-in-process.” It appears as separate symptoms instead — tight cash flow, a shortage of storage space, long lead times, large physical-count variances. Because one cause shows up as several symptoms, each symptom gets its own countermeasure and the root cause is never touched.

The starting point of WIP management is therefore not to look for reduction techniques. It is to know how many pieces exist, daily rather than monthly. The simple fact that you cannot reduce inventory you cannot see underlies everything else.

What WIP Actually Is — The Cost Is Already Spent, The Sale Has Not Happened

Let us settle the terminology first. Work-in-process is product that sits partway through the manufacturing route and is not yet finished. It has moved past the raw material stage, so it is not raw material; it has not passed inspection, so it is not finished goods. It is the awkward middle state that resists a clean name.

Daiko Xtech, in its explanation of work-in-process, describes it as product in the middle of manufacturing for which costs such as material and labour have already been incurred, but which cannot be sold because it is incomplete. That single definition contains every reason WIP management is difficult. The cost is spent, and the item cannot be sold. Spending is already fixed; recovery depends on how the route flows.

In English the term is WIP, for Work In Process or Work In Progress. The MachineMetrics WIP guide likewise defines it as goods that are partway through production and not yet complete, and treats it as something to be managed together with indicators such as cycle time, lead time and throughput. At Thai sites, local staff usually understand the term WIP directly, so it is worth establishing early that the Japanese term and the English WIP refer to the same thing. When Japanese managers say one word and local staff say another, the two groups can be looking at the same figure and still talk past each other.

A second point that regularly causes confusion in practice is the difference between WIP and semi-finished goods. Semi-finished goods have been processed far enough to be sold on their own, or are held as an established stocking unit. WIP cannot be sold on its own. Because the distinction affects accounting treatment, check which account your own operation uses. If the classification stays vague while you chase quantities, the accounting ledger and the shop floor figures will never reconcile.

WIP also has the property that its unit cost rises as it advances. A part that has only been machined and a part that has been heat treated and ground may look identical, yet they carry very different accumulated cost. The same single piece is worth more the further down the line it sits. That obvious fact matters a great deal once you start valuing WIP. If you look only at quantity and decide to attack the biggest pile, you may be attacking the place with the smallest financial effect.

Three Structural Reasons WIP Management Is Hard — Quantity, Cost, Dwell Time

The difficulty of WIP management has nothing to do with the ability or motivation of the people doing it. There are three structural reasons.

First, quantity is hard to capture. Raw material sits in a warehouse and finished goods sit in a warehouse. Warehouses have fixed locations and receipt and issue records. WIP is not in a warehouse. It is spread beside machines, on conveyors, in carts, in racks awaiting inspection, and in returnable crates parked along the aisle between processes. It also moves constantly. It flows to the next process while you are counting it, so capturing an accurate snapshot by hand is difficult in principle.

Second, cost accounting becomes complex. As noted, WIP unit cost varies with progress. With 5 processes there are at least 5 cost stages. In practice a degree-of-completion approach is used to allocate cost, but if you do not know how many pieces have reached which process, you have no basis for the allocation. Weak quantity capture automatically degrades costing accuracy. Daiko Xtech lists exactly this chain — difficulty of quantity capture, complexity of cost calculation, and the risk of excess inventory — as the three challenges of WIP management, and the first two feed each other.

Third, dwell time is hard to detect. As long as the total quantity between processes is unchanged, nobody notices that the contents have not turned over. Day-to-day management that watches only totals cannot show that first-in-first-out has broken down. Metal parts rust, resin degrades, electronic components become unusable after a design change. Dwell is a question of time, not volume, so management that watches volume alone cannot catch it.

Put the three side by side and a common element appears. None of them is solved by effort in the moment. Count more often, do costing more carefully, walk the floor more frequently — all of these help, but as long as people do the work there is a hard limit on frequency. WIP management turns into a systems discussion because it lies beyond the reach of exhortation.

Little’s Law | Lead Time Is Determined by WIP Quantity

There is one principle we always raise first when explaining WIP management. It is Little’s Law, a basic relationship used in production control and queueing theory, and it can be written as follows.

Lead time = WIP quantity divided by throughput (output per unit of time)

The formula is simple; its implication is powerful. If throughput is constant, lead time is directly proportional to WIP quantity. Halve WIP and lead time halves. Double WIP and lead time doubles. As long as throughput does not change, the proportionality holds.

What makes this useful in practice is that it clarifies the direction of causation. Most factories explain long lead times by saying there are too many processes, the machines are slow, or changeovers take time. None of that is wrong, but all of it is about throughput, and raising throughput requires capital investment. Little’s Law points to a second route: reduce WIP quantity. That route does not require buying equipment.

Take concrete figures. A line producing 2,000 pieces per day carries 24,000 pieces of WIP. Lead time is 24,000 divided by 2,000, or 12.0 days. If the same plant changes no equipment at all and brings WIP down to 15,000 pieces, lead time becomes 15,000 divided by 2,000, or 7.5 days — a reduction of 4.5 days. Not one machine was purchased. Only the balance between input and output changed.

One caution is important here. Little’s Law says that reducing WIP shortens lead time; it does not tell you how to reduce WIP. Actually reducing it requires throttling release to match downstream capacity, and that requires knowing how much each process is holding right now. The law tells you where to aim; execution requires visibility. The order is visibility first, reduction second.

There is a further implication. Reducing WIP does not only shorten lead time — it also speeds up the discovery of quality problems. In a plant with a lead time of 12.0 days, a defect created upstream takes an average of 12.0 days to reach final inspection, and during that time the plant keeps producing items with the same defect. If lead time falls to 7.5 days, the delay in discovery is 7.5 days. That shrinking time gap is why WIP reduction is said to help quality.

Four Reasons WIP Piles Up Between Processes

So why does WIP accumulate? Ipros, in its explanation of WIP inventory reduction, organises the obstacles into four causes. They match what we see on the floor closely, so let us take them in order.

First, insufficient response to demand variability. Because orders are hard to forecast, upstream processes are run ahead so that anything can be shipped at any time. Taken alone the decision is rational; failing to meet a sudden increase in demand costs customer trust. But if you buy insurance against forecast uncertainty in the form of WIP, the premium is paid every day in inventory value and floor space. And whatever you guessed wrong about turns straight into stagnant stock.

Second, inefficiency in production planning. In most cases the problem is not the absence of a plan but a plan that cannot reflect real-time information. A schedule set at the start of the week cannot absorb an equipment breakdown or a rework loop that occurs midweek. Release continues according to plan while downstream falls behind, and the gap becomes a pile between processes.

Third, weak coordination between processes. The upstream process pursues its own output target; the downstream process handles what its capacity allows. If neither side shares how many pieces have been handed over and how many are waiting, the upstream process has no reason to stop. In plants that evaluate each process on machine utilisation, the tendency is stronger still, because the behaviour that raises utilisation collides head-on with the behaviour that reduces WIP.

Fourth, rework caused by quality problems. When rejected items go back upstream, inventory between processes increases by that amount. Rework also leaves the normal flow, so it drops out of routine progress tracking. “That crate is waiting for rework” is knowledge held only in one person’s head — a scene visible in every factory. On the day that person is absent, the crate ceases to exist.

Of these four, the second and third are information problems and the first and fourth are operational problems. Until the information problems are solved, the operational ones cannot be touched, because deciding how much insurance to hold against demand variability requires knowing how much you hold today. Ipros lists AI demand forecasting, real-time management via IoT and MES, integrated production management systems, and stronger quality checks as countermeasures — all of which presuppose that current data is actually being captured.

Model Case | Counting the WIP in an Automotive Parts Plant in Chonburi

From here we work through a concrete model case. The following is an independent estimate constructed as a model case, not the figures of any real company. Rather than the amounts themselves, please look at the structure — where the queues form, and what moves once they become visible.

The premise is as follows. A Japanese-owned automotive parts machining plant in Chonburi Province, Thailand, with 180 employees. It runs an integrated line of 5 processes — machining, heat treatment, grinding, assembly and inspection. It operates 22 days per month and produces 2,000 finished pieces per day. WIP is tracked with handwritten paper tags and Excel, and a physical count is performed once a month. As Japanese-owned machining plants in Thailand go, this is an entirely ordinary configuration.

At one month-end physical count, WIP was counted by process and converted into value using the unit cost at that stage. The unit cost is the cost already invested by the time the item reaches that point.

That is 24,000 pieces in total, worth 4,320,000 THB. Until they saw this table, the management team did not know the total value of their WIP. The monthly trial balance carried a WIP account balance, but nobody had looked at where in the route, and in what form, that balance physically existed.

Apply Little’s Law first. Dividing 24,000 pieces by a throughput of 2,000 pieces per day gives a lead time of 12.0 days. That means an average of 12.0 days elapses between material release and completion. The sum of actual processing time across the 5 processes comes nowhere near that figure, so almost all of the remainder is waiting.

The figure of 12.0 days sits uncomfortably against shop floor intuition. Ask an operator how long processing takes and you get net processing time. Ask sales about the delivery lead time and you get a number with buffer built in. The waiting time in between belongs to nobody, so nobody can answer for it. Only when WIP is counted does that waiting time become a number.

There is a second thing to read from the table. The largest quantity is the 7,200 pieces waiting for heat treatment, but the largest value is the 1,144,000 THB waiting for assembly, because the unit costs differ. Whether you prioritise by quantity or by value changes where you start. Neither is inherently correct — look at quantity if you want to shorten lead time, and at value if you want to release working capital.

Where the Queue Was Concentrated — Between Machining and Heat Treatment

Reading the breakdown by process, the 7,200 pieces waiting for heat treatment stand out. Of the total 24,000 pieces, three tenths sit in a single location between machining and heat treatment. Investigating the reason revealed a structural cause.

Heat treatment is a batch process. A fixed volume is loaded into the furnace and processed at a set temperature for a set time. You cannot run a single piece; you either wait until the furnace is full or run it partly empty. Running it empty wastes fuel, so the floor naturally waits for a full load. That wait becomes the pile of machined WIP.

There is also a change in “granularity” across heat treatment. Machining flows one piece at a time, heat treatment flows in batches, and grinding returns to one piece at a time. Wherever continuous and batch processes alternate, inventory necessarily appears at the junction. This is not poor shop floor practice; it is an inevitable consequence of processes with different characteristics.

That said, the quantity that arises inevitably and the quantity that arises from weak management are different things. In this plant, the furnace processes 1,000 pieces per cycle and runs twice a day. Daily capacity is therefore 2,000 pieces, exactly matching the line throughput. The queue genuinely required by the process is the volume for the day’s two batches, roughly 2,000 pieces. What was actually there was 7,200 pieces — 3.6 days’ worth. The difference of 5,200 pieces is a candidate surplus arising not from the nature of the process but from release timing that is out of step with downstream capacity.

Why does that surplus arise? Because the machining process was behaving so as to maximise its own utilisation. Stop a machining centre and utilisation falls, and that number lands in the daily report. Let WIP pile up ahead of heat treatment and no number is recorded anywhere. When some indicators are measured and others are not, people act in line with the ones that are measured. This is not a matter for blame; it is a matter of indicator design.

The 5,600 pieces waiting for grinding have a similar background. Hardness after heat treatment can vary, and items cannot move to the next process until sample inspection passes. That inspection wait accumulates into the grinding queue. Here too, necessary waiting and surplus waiting are mixed together, and unless they are separated you cannot judge how much can be removed.

When you look at inventory between processes, the decisive question is whether you can explain why it is there, split between the nature of the process and the state of management. The portion you cannot explain is the portion you can remove. Cut into the portion you can explain, and the line stops next.

With Paper Tags You Do Not Notice the Queue Until Inventory Count Day

The model case plant used handwritten paper tags and Excel. A paper tag is attached to the returnable crate holding the parts, an operator writes the quantity and date each time it passes a process, and at the end of the day an administrator transcribes it into Excel. It is the standard method seen in many factories.

The problem with this method is not accuracy. What is written down is broadly correct. The problems are three.

First, aggregation lag. Figures written on the floor reach Excel the next day at the earliest, and several days later in busy periods. Only after aggregation can the balance by process be seen, so “how many are there now” is always a picture of the past. There is a time gap between the moment a queue starts and the moment it appears as a number.

Second, dwell time is not recorded. A paper tag records when a crate passed a process, but not how long it has been sitting where it is. As long as the total matches, a crate that has not moved for months in the back of a storage area changes nothing in the aggregate, so the time dimension drops out entirely.

Third, transcription effort caps the frequency of visibility. In this plant, writing tags by hand and transcribing them into Excel consumed a total of 4.5 hours per day. That is exactly why capturing the full count more often than the monthly physical inventory was not realistic. Raising the frequency would have required more people.

Fabrico, in its guidance on reducing work in process, makes the point bluntly — where WIP is tracked on paper tags, its existence goes unnoticed until the day of the inventory count. A paper tag records only the “point” at which an item passed a process. How many days the item waited between one point and the next is recorded nowhere. Dwell shows up not as a point but as the length between points, so this format cannot capture it in principle.

Excel itself is not the villain. But the moment the object being managed is something that keeps moving, Excel stops being suitable, because transcription effort sets the update frequency and update frequency sets the speed of decisions. We have set out how to judge where that limit lies in the limits of Excel-based management and when to switch, which is worth reading alongside this article.

Paper tags carry one more side effect: physical count variance. In this model case, the gap between the counted quantity and the book quantity was 3.2%, or 768 pieces. Missed transcriptions, writing errors, lost tags and double-counted rework all contribute. Each individual case is small, but the accumulated quantity degrades costing accuracy.

What a Shop Floor Tracking System Changes — Three Ways to Capture Data Automatically

The essence of a shop floor tracking system is replacing the part that people wrote and people transcribed with a mechanism that machines read. SmartMat Cloud, in its explanation of factory visibility systems, describes using IoT sensors and the cloud to quantify utilisation, inventory and quality. Factory Advance, writing on real-time cost management, describes operators entering results on the spot using tablets, smartphones, QR codes and barcode readers. Both point the same way — capturing data without passing through human memory and handwriting. The means divide into three broad options.

The first option is barcodes and QR codes. Codes are printed on returnable crates or tags and scanned with a handheld terminal as the item passes each process. Implementation cost is lowest and the existing tag operation can be retained. On the other hand, scanning depends on human action, so scans get missed. The practical trick is to build the scan into the entrance and exit of each process so that the item cannot move on without being read.

The second option is RFID. IC tags are attached to returnable crates and gate antennas at process entrances and exits detect passage automatically. No human action is required, so missed reads are eliminated in principle. In plants handling metal parts, radio reflection can be an issue, so verifying tag type and mounting position in advance is essential. If tags are attached to reusable crates, the tag cost is divided across many cycles and the cost per part becomes small.

The third option is sensors and cameras. Weight sensors estimate the number of pieces in a crate, ceiling cameras capture the occupancy of a storage area, and machine signals count actual output. Nothing has to be attached to the parts themselves, which suits processes handling large volumes of small components. Individual traceability is not possible, but for knowing how much has accumulated in a given location it is entirely sufficient.

Which of the three to choose depends on the nature of the process. If per-item traceability is required, RFID or barcodes; if only the total in a location matters, sensors or cameras. In practice, mixing methods by process is realistic, and there is no need to standardise the whole route on one approach.

Whichever method is chosen, the outcome is the same. The when, the where and the how many are all recorded without human intervention, and the aggregation lag disappears. Only at this point does dwell time become calculable. Once you know how many days a given crate has stayed in the same place, you can identify where first-in-first-out has broken down. That information was impossible to obtain under a paper operation.

The connection to cost accounting follows from the same data. Once process-passage records accumulate automatically, you know daily how many pieces have reached which process, so the base data for degree-of-completion allocation is already there. The month-end exercise of stopping the floor to count is replaced by the accumulation of daily records.

Model Case After Implementation | Changes in WIP Value and Lead Time

The model case plant fitted RFID tags to its returnable crates and installed gates at process entrances and exits for automatic reading. Alongside this, it changed its operating rule so that machining release was throttled to match the furnace operating plan for heat treatment. The state 6 months later is shown below. To repeat, these are independent estimates and not the figures of a real company.

Placing the main indicators side by side before and after gives the following.

Quantity fell from 24,000 pieces to 15,000 pieces, a reduction of 9,000 pieces, or 37.5%. Inventory value fell from 4,320,000 THB to 2,780,000 THB, a reduction of 1,540,000 THB. That reduction rate is 35.6%, slightly below the quantity reduction rate. The reason is that most of the reduction occurred in the low-unit-cost early processes. The machining queue fell by 1,500 pieces and the heat treatment queue by 3,000 pieces, a combined 4,500 pieces, while the inspection queue at a unit cost of 380 THB fell by only 800 pieces. That structure is why the quantity and value reduction rates diverge.

Lead time moved exactly as Little’s Law predicts. 15,000 divided by 2,000 gives 7.5 days, a reduction of 4.5 days from the previous 12.0 days. The reduction rate is 37.5%, precisely matching the quantity reduction rate. Since throughput did not change, that is the only possible outcome. The match is not coincidence but confirmation that the law holds, so always perform this check on real projects. If the numbers do not agree, some premise has shifted.

The physical count variance rate improved from 3.2% to 0.4%, in quantity terms from 768 pieces to 60 pieces. This came from eliminating manual transcription, and it is less an achievement of the reduction programme than an automatic by-product of changing how data is captured. Smaller variance improves costing accuracy and brings the monthly trial balance into line with shop floor reality.

WIP storage area fell from 480 square metres to 300 square metres, a reduction of 180 square metres, or 37.5% — the same rate as the quantity reduction. The freed floor space was used to install a new machining cell. In other words, capacity could be added without expanding the site. Floor area is hard to convert into money, but in the sense that it defers the decision to put up another building, it matters a great deal in practice.

Tag transcription effort fell from 4.5 hours per day to 0.5 hours, a reduction of 4.0 hours. The remaining 0.5 hours covers checking read errors and handling exceptions. Automation never drives exceptions to zero, and a headcount plan that fails to allow for this leaves the floor unable to cope.

Four Routes by Which WIP Reduction Improves Cash Flow

What actually improves for the company when WIP falls? Four routes explain it. Biznet, writing on lead time reduction, sets out how reducing WIP inventory leads to better cash flow, more efficient use of storage space and shorter lead times, while the ISDI Imaoka System Dynamics Institute addresses why shorter lead times translate into corporate earnings. Let us take them in order.

The first route is the release of working capital. A fall of 1,540,000 THB in WIP inventory value means that much cash has been released from inventory form. It does not appear as profit in the income statement, but the asset composition on the balance sheet changes and cash flow becomes easier. In plants whose terms require paying for material before collecting receivables, this difference is cash flow headroom itself.

The second route is lower storage cost. Beyond floor space, the number of returnable crates, the number of transport moves, forklift running time and the effort of managing storage areas all fall. In the model case, 180 square metres were freed. You can convert this to a rental equivalent, but in practice the space is more often used to house new equipment, in which case its value exceeds the rent.

The third route is lower quality risk. Shorter dwell reduces the risk of rust, degradation, contamination and obsolescence from design changes. In addition, as noted earlier, the time before an upstream defect is discovered downstream shortens, so the volume of defective product produced falls. A lead time falling from 12.0 days to 7.5 days means defects are found up to 4.5 days sooner.

The fourth route is more order opportunities from shorter lead times. When the delivery date you can quote gets shorter, short-lead-time orders become winnable. In automotive parts, the ability to follow a customer’s production plan changes can be a condition of continued business, so this capability competes on something other than price. We have set out how delivery delays arise and what can be done upstream of them in preventing delivery delays through process control.

Of the four routes, the first and second are easiest to explain to management, and the third and fourth are what the shop floor feels most directly. When seeking project approval, lead with the first; when seeking shop floor cooperation, lead with the third and fourth. The same initiative lands differently depending on who is listening.

Four Steps to Roll Out Process Progress Visibility

Attempting process progress visibility all at once always stalls partway. Break it into stages. Here are four steps.

The first step is identifying the physical unit. Decide what constitutes one unit for counting and give that unit an identifier. Crate, pallet, lot or individual piece. The finer the unit, the more information you get and the heavier the operational load. For most machining plants, the returnable crate is a realistic starting point. What matters at this stage is consistency of granularity. If units differ by process, quantities will fail to reconcile later.

The second step is automatic data capture. For the chosen unit, record passage automatically using barcodes, RFID or sensors. There is no need to cover every process at once. Starting with the process gap holding the largest queue is the efficient choice. In the model case, that means starting on either side of heat treatment. Even one automated location produces daily figures there, which gives you something to show internally.

The third step is displaying process progress. Present the collected data on separate screens for the shop floor and for management. The floor needs to know how many pieces are waiting at its own process; management needs to know how many pieces sit where across the whole route. Trying to serve both with one screen produces something with too much information for either. For the floor, the rule is one large display visible from the aisle.

The fourth step is comparison against a standard. Set a target WIP level for each process gap and compare it against actual quantities. The target level should come from concrete grounds such as the downstream daily processing volume, or one batch for a batch process. Without that reference, nobody can judge whether a displayed number is high or low. The single biggest reason visibility projects end as “we displayed it” is the absence of this fourth step.

Of the four steps, the first and second are implementation work and the third and fourth are operational design. Results appear only once you reach the fourth. Plenty of factories stop at the third — the screen exists, but nobody has decided who does what when they look at the number. Include in your implementation plan the question of who sets the fourth-step reference values and how.

As a guide to sequencing, our experience supporting these projects is that steps one and two get up and running quickly if you confine them to the process gaps with the largest queues, but the reference values in step four cannot be fixed until enough automatically captured data has accumulated. Do not try to decide everything at once; leave the reference values until last. If you fix the timeline first and work backwards, the grounds for the reference values will inevitably be thin.

How to Think About the Cost and Payback of a WIP Management System

Now to cost. In the model case, the system carried an initial cost of 1,200,000 THB and annual maintenance of 180,000 THB. The initial cost covers RFID tags and gate antennas, reading terminals, server and software, installation work, and operational design and training on the floor. This too is an independent estimate, and it varies widely with configuration and scale.

When considering payback, separate one-off effects from recurring effects. Mixing them leads to bad decisions.

The one-off effect is the release of working capital. WIP inventory value fell by 1,540,000 THB, so that much cash returns to hand once. That 1,540,000 THB exceeds the initial cost of 1,200,000 THB. However, the same amount does not return the following year, so it must not be counted as an annual benefit. Its proper role is as a source of funds for the initial investment.

There are two recurring effects. One is effort reduction. Handwriting and transcribing tags fell by 4.0 hours per day, which across 22 operating days is 88 hours per month and 1,056 hours per year. Placing the hourly cost of manufacturing indirect staff at 150 THB per hour gives 158,400 THB per year. The other is lower inventory holding cost. Setting the holding cost rate — the combination of cost of capital, storage and obsolescence — at 12% per year, the 1,540,000 THB reduction yields 184,800 THB per year.

Adding the two gives an annual recurring effect of 343,200 THB. Subtracting the 180,000 THB maintenance cost leaves a recurring net of 163,200 THB per year.

Laid out as numbers, the investment takes the shape of “the release of working capital covers the initial cost in year one, and from year two onward a net of 163,200 THB per year accumulates.” It is not a spectacular return. Investment cases for WIP management systems usually settle at about this level.

This is therefore an investment that is hard to approve on the numbers alone. In practice you explain it alongside the effects that resist conversion into money. What can be placed in the 180 square metres that were freed? Are there orders that become winnable now that lead time is 4.5 days shorter? Do you stop the line for fewer days to take physical counts? These are hard to put in a calculation, yet they carry real value for both the floor and management.

One way to hold cost down is to avoid covering every process at once. As noted above, starting with the process gap holding the largest queue lets you split the initial cost while confirming the effect. For the model case configuration, a proposal covering only the two locations either side of heat treatment would be the sensible place to begin.

Do Not Cut WIP Too Far — How to Set the Right Level

Having spent this long on reduction, it should be said clearly that WIP is not something to drive to zero. Inventory between processes serves the purpose of absorbing upstream variability. Remove it and the downstream process stops the moment the upstream process pauses.

How to set the appropriate level divides into three cases according to the nature of the process.

Ahead of a batch process, one batch is the absolute floor. For the heat treatment in the model case, the furnace processes 1,000 pieces per cycle, so at least 1,000 pieces must be available ahead of it. The practical floor sits somewhat higher, however. The furnace runs twice a day, so 2,000 pieces are held for the day’s requirement. The heat treatment queue settling at 4,200 pieces after implementation reflects that daily requirement plus an allowance to absorb equipment trouble and changeover variability on the machining side. Of the 5,200 pieces of candidate surplus, only 3,000 pieces were removed and the rest was deliberately retained.

Ahead of a process with variable capacity, hold an amount matching the variability. Measure the frequency of equipment trouble and the spread of changeover times from actual records, and hold enough to absorb that variation. Deciding by feel — “let us keep a bit extra to be safe” — always produces excess. With actual data, the judgement can be made numerically. The benefit of visibility extends beyond reduction to justifying the quantity you genuinely need.

For products with large customer demand swings, holding stock as finished goods is often better. Holding it as WIP preserves flexibility because the final part number is not yet fixed, but the capital tied up differs greatly depending on whether you hold at a low-cost early process or a high-cost late one. If part number differentiation occurs downstream, holding immediately before that differentiation is the rational choice.

There are also warning signs of cutting too deep. Process stoppages become more frequent, changeovers increase and eat into effective running time, and rush jobs are inserted more often. All of these tend to appear immediately after a WIP reduction. If they show up, put some of the removed quantity back. The appropriate WIP level is not decided once and left alone; it is adjusted continuously against actual results.

With visibility in place, that adjustment can also be done numerically. Line up the time a process stopped against the WIP balance immediately before it, and you can see the level below which stoppages occur. Under a paper operation, that causal relationship could not be traced after the fact.

Why WIP Management Matters at Thai Sites in 2026

There are circumstances specific to Japanese-owned factories in Thailand. Behind the rising priority of WIP management in 2026 lies a people problem.

Allied Corporation, in its overview of the Thai working environment, points out that population ageing has begun to have a serious impact on labour supply and that the shortage of skilled workers has become a bottleneck for manufacturing and services. Operations built on the assumption of available labour no longer hold as automatically as they once did.

This connects directly to WIP management because handwriting and transcribing tags is precisely the kind of work that assumes available labour. The model case plant spent 4.5 hours per day on it. Moreover, the work generates no added value — the shape of the product does not change. In a period when hiring becomes harder, continuing to devote people to non-value-adding work steadily erodes competitiveness.

A second circumstance particular to Thailand is the speed of staff turnover. In a labour market where changing employers is common, it is not unusual for shop floor staff to turn over every two to three years. In plants where the location of WIP depends on individual memory, capture accuracy degrades every time a person changes. Information such as “that crate is waiting for rework” fails to be handed over, and a queue quietly begins in the corner of a storage area. This is a genuinely common sight.

The same applies to the rotation of Japanese expatriate managers. With a handover every three to five years, the intuition for how much WIP in which location counts as normal is reset each time. If reference values are registered in a system they can be handed over; what lives in someone’s head cannot.

Then there is the question of floor area. Space used for WIP storage is a fixed cost. Being able to free 180 square metres, as in the model case, also means being able to defer the decision to add a building in order to expand output. The more constrained a site is for expansion within its industrial estate, the more this point bears on capital investment decisions.

Finally, there is preparation for customer requirements. Within the scope of the enquiries we receive, requests for records of process passage across automotive supply chains have become more frequent than before. A mechanism that automatically records the passage of WIP through processes serves that requirement directly, not only WIP reduction. One investment satisfying two purposes is also a stronger structure for an internal approval request.

Five Common Failure Patterns

Here are five failures we have actually seen at Japanese-owned factories in Thailand.

Setting the reduction target before anything else. This is the pattern of declaring “we will cut WIP by a third” and only then starting on visibility. With the target set first, the floor tries to reduce before it counts. The result is that necessary buffer stock is cut too, the line stops, and the reaction leaves the plant holding more than before. The order is count, separate the causes, then reduce.

Covering every process at once. A plan to fit RFID across all 5 processes inflates the initial cost, so either the approval fails or, if it passes, the implementation drags on and exhausts the floor. Starting at the process gap with the largest queue and expanding after confirming results is faster in the end.

Building the screen and calling it done. This is the case where the project is treated as complete the moment the process progress dashboard is finished. Without the fourth step above — reference values for comparison against a standard — nobody acts on the displayed numbers. Include in the implementation plan who sets those reference values and how.

Leaving utilisation-based evaluation untouched. Continuing to evaluate each process on machine utilisation while instructing the floor to cut WIP hands people two contradictory orders. Keeping the upstream process running raises utilisation and increases WIP. If you take on WIP reduction, the evaluation indicators need to be revisited at the same time.

Excluding rework from the scope of management. Items awaiting rework leave the normal flow and are easily missed by identification and recording. Yet rework is also where dwell tends to be longest. When you assign identifiers, always decide how rework will be handled. A design that leaves this undecided will always drift out of balance on quantity once it goes live.

Frequently Asked Questions

What exactly does WIP management manage?

It is the activity of knowing, for unfinished items partway through production, how many are at which process, how much cost has been invested in them, and how long they have been sitting there — and of keeping that at an appropriate level. Because WIP has already incurred costs such as material and labour while being unsaleable, quantity capture, cost allocation and dwell detection have to be handled together. Most factories confirm quantities through a month-end physical count, but that alone reveals nothing about what happened mid-month. The practice of WIP management begins by replacing the monthly count with daily capture.

How far can WIP inventory be reduced?

There is no single answer, because the necessary floor varies with the nature of each process. At least one batch is required ahead of a batch process, and an amount able to absorb variability is required ahead of a process with frequent equipment trouble or variable changeovers. In the model case, WIP fell from 24,000 pieces to 15,000 pieces, a reduction of 37.5%, and that was possible because visibility allowed the plant to separate the portion explained by process characteristics from the surplus caused by release timing. Cut the surplus you cannot explain and keep the portion you can. Whether you can draw that line determines the size of the reduction.

How much does a shop floor tracking system cost?

It varies widely with configuration and scale, so there is no general market figure. The model case assumes RFID-based automatic reading on an integrated line of 5 processes, estimated at an initial cost of 1,200,000 THB and annual maintenance of 180,000 THB. As a decision framework, separate one-off from recurring effects. In the model case, the working capital release of 1,540,000 THB exceeded the initial cost of 1,200,000 THB, and the recurring net was 163,200 THB per year. If you want to hold costs down, starting with only the process gaps holding the largest queues is an effective approach.

Where should we start with process progress visibility?

Start by deciding the unit you will count. Crate, pallet or lot. If units differ by process, quantities will fail to reconcile later, so aligning granularity is the first task. Next, record passage of that unit automatically using barcodes, RFID or sensors. There is no need to cover every process — one location at the process gap with the largest queue is enough. Even a single automated location produces daily figures, which become the basis for judging the next investment. Building screens comes after that.

How should we use Little’s Law on the shop floor?

Use it to set targets. Lead time equals WIP quantity divided by throughput, so as long as throughput is unchanged, lead time and WIP quantity are proportional. In the model case, WIP fell from 24,000 pieces to 15,000 pieces, a reduction of 37.5%, and lead time fell from 12.0 days to 7.5 days, also 37.5%. That agreement is the arithmetic check that the law holds. Always run this calculation after a reduction on a real project. If the numbers do not agree, either throughput has changed or some process is still being undercounted.

Summary

To bring the main points together.

WIP does not fall because the shop floor is not trying hard enough. It does not fall because nobody knows, on a daily basis, how many pieces sit between one process and the next. A monthly physical count tells you nothing about what happened mid-month. WIP is an asset whose cost is already incurred and which has not yet turned into revenue, and because its growth blocks nobody’s work, it is easily left alone.

By Little’s Law, lead time equals WIP quantity divided by throughput. Without adding equipment, reducing WIP shortens lead time proportionally. But the law shows only where to aim; the route to get there has to be prepared separately. Put visibility of the real quantities between processes in place first, and start reduction after that.

In the model case estimate, the automotive parts machining plant in Chonburi Province held 24,000 pieces of WIP worth 4,320,000 THB. Against a throughput of 2,000 pieces per day, lead time was 12.0 days. After introducing RFID-based automatic reading and adjusting release volumes, WIP came to 15,000 pieces and 2,780,000 THB — a reduction of 37.5% by quantity and 35.6% by value. Lead time shortened by 4.5 days to 7.5 days, and the reduction rate of 37.5% matches the quantity reduction rate. The physical count variance rate improved from 3.2% to 0.4%, WIP storage area from 480 square metres to 300 square metres, and tag transcription effort from 4.5 hours per day to 0.5 hours. These are independent estimates rather than the figures of a real company, so please look at the structure — where queues form and how they are resolved — rather than at the amounts themselves.

On cost, against an initial 1,200,000 THB and annual maintenance of 180,000 THB, the working capital release of 1,540,000 THB arises as a one-off effect, while effort reduction of 158,400 THB and lower inventory holding cost of 184,800 THB combine into a recurring 343,200 THB, leaving a recurring net of 163,200 THB per year after maintenance. On the numbers alone it is an unglamorous investment. Judge it together with the effects that resist calculation — what the freed floor area can be used for, which short-lead-time orders become winnable, and how many days the line no longer has to stop for counting.

In practice the work proceeds in four steps: identifying the physical unit, capturing data automatically, displaying process progress, and comparing against a standard. Results appear only once you reach the fourth. Do not treat the project as finished when the screen is built, and do not aim for zero WIP — keep the portion that the nature of the process explains. Those two points are the conditions for making a reduction stick.

As a first move, take the process gap where you feel the queue is largest and count the actual quantity there for a single day. Calculate how many days’ worth that one figure represents against throughput, and you will have the material you need to judge whether visibility is worth investing in.

It is perfectly fine to be at the stage where you simply do not know how many pieces sit between processes. TOMAS TECH implements and operates production management systems for Japanese-owned factories in Thailand, and we are happy to start from mapping out your current processes and where the queues actually are. Consultation is welcome even at an early exploration stage with no implementation assumed, so if you would like to turn your current situation into numbers, please get in touch through our contact page.

References

• Daiko Xtech — What work-in-process is, why managing it matters, and how a production control system improves it, daiko+ production control column, in Japanese
• Ipros — Understanding WIP inventory reduction through four challenges, for manufacturing professionals, manufacturing portal article, in Japanese
• MachineMetrics — What is Work in Progress WIP in Manufacturing A Complete Guide, definition of WIP and management indicators
• Fabrico — 5 Strategies to Reduce Work in Process WIP in Manufacturing, paper tags and the limits of visibility
• SmartMat Cloud — Factory visibility systems, main system types, how to choose and implementation examples, quantification via IoT sensors, in Japanese
• Factory Advance by e-Port Inc. — Real-time cost management in manufacturing, knowing current profit without waiting for the monthly close, result entry via QR codes and barcode readers, in Japanese
• Biznet — Know-how on lead time reduction for manufacturers, with points to watch, WIP reduction and cash flow, in Japanese
• ISDI Imaoka System Dynamics Institute — Why manufacturers shorten lead times and how to calculate inventory time, lead time reduction and corporate earnings, in Japanese
• Allied Corporation — The Thai working environment in 2026, strategies for attracting skilled talent and the digital transition, labour supply and the shortage of skilled workers, in Japanese