Manufacturing Lead Time Reduction Is Information, Not Equipment | The Five Layers of Waiting

One Japanese-owned plant in Thailand replaced the machining centre it had believed was the bottleneck with a faster model. Cycle time did drop. Six months later, however, the number of days between receiving an order and shipping the product was almost exactly what it had been before. The equipment was faster, yet the time it took for a product to leave the factory had not moved. This is precisely why a serious look at manufacturing lead time reduction ends up being a conversation about information systems rather than hardware. This article breaks manufacturing lead time into five layers of waiting, and sets out which layer shrinks with which measure, and which measures do nothing at all.

Why Capital Investment in Equipment Does Not Shorten Lead Time

In most factories, the phrase “manufacturing lead time” is used almost interchangeably with “the time it takes to process the work”. But when you actually decompose the interval from order receipt to shipment and lay it out process by process, the time during which material or work-in-process is genuinely being worked on turns out to be a small fraction. The overwhelming majority is time in which nothing is happening at all. It is waiting. This is not a quirk of one industry or one plant size; it is a structure observed broadly in high-mix, low-volume manufacturing (see the discussions of stagnation between processes at Factory Advance and Zaiko Kanri 110ban, listed at the end of this article).

Once you see the structure, you can estimate what equipment investment will actually buy you. If processing accounts for only a small share of the total, then halving that processing time yields a saving of half of a small share. Equipment investment carries a large price tag and delivers a limited effect on the order-to-shipment calendar. Waiting time, by contrast, can move in response to a category of measure that costs an order of magnitude less: changing how information flows. That is the whole reason practitioners say lead time reduction starts with information, not with machines.

Waiting Is the Time Someone Spends Unable to Decide

So what is waiting, really? Stand in front of a stack of pallets sitting between two processes and observe what is not happening. It is not processing that has stopped. It is deciding. Nobody has determined what runs next. Nobody can start because it is unclear whether the material will arrive. The downstream process does not know the upstream process has finished. The shipping instruction cannot be issued because the inspection result has not been judged. In every case the work is not physically impossible. It is stopped because the person who has to decide does not have what they need in front of them.

Waiting time, in other words, is the sum of the time during which the person who must decide does not have the information they need, at the moment they need it. Adopt that definition and the direction of the countermeasure becomes obvious. Deliver the information that is not being delivered. A surprising volume of the waiting inside a factory shrinks on that alone. And the converse holds just as firmly: faster equipment, more headcount and more overtime leave the undelivered information exactly where it was, so that waiting does not shrink by a single minute.

A brief note on scope. Many plants set manufacturing cost as the KPI for improvement work. This article does not go into cost. The point worth holding onto is simply that lead time moves before cost does, which makes it the faster primary indicator when you are trying to run an improvement feedback loop.

Breaking Manufacturing Lead Time Into Five Layers of Waiting

This is the core of the article. Lamenting that lead time is “long” gives you nothing to act on. Break it into the following five layers, however, and you can see how many days are accumulating where. More importantly, you discover that the effective measure is completely different in each layer.

There is an important asymmetry across these five layers. Layers 1 and 5 are fundamentally about transmitting information, so they shrink comparatively cheaply with systems. Layer 3 is a planning-logic problem, which is where a production scheduling system (APS) does its work. Layer 2 involves commercial terms and import clearance, so it does not move on internal effort alone and takes time. And Layer 4 is a measurement problem first: without measurement you cannot even judge which of the other four layers is hurting you. What follows treats each layer in turn, separating symptom, cause, what shortens it and what does not, with one example of the form it actually takes in a Japanese-owned plant in Thailand.

Layer 1 | Waiting on Information, From Order to Confirmed Plan

The symptom is easy to recognise. Sales or headquarters has sent through an order or a forecast, but it does not appear in the production plan for several days, or in bad cases until the following week. The shop floor says the plan has not come down yet. The planner says a plan cannot be built until the information is complete. During this interval the factory is capable of running and is not running.

The cause is that order information reaches the planner through human hands. Sales updates a spreadsheet of orders, emails it across, and the planner re-keys it into the production plan. If any step in that chain sits overnight, that whole night is added to the lead time. On top of that, if the planning cycle itself is weekly, an order that arrives just after Monday morning’s plan is frozen may wait until the following Monday in the worst case. This cycle wait is also why the same product can show lead times that differ by several days depending purely on when the order happened to land.

The form this takes in a Japanese-owned plant in Thailand is most often the two-stage structure of forecast and firm order from the parent company. A forecast comes mid-month, the firm order comes near month end. The local plant waits for the firm order before arranging material, and the start of production slides back by exactly that much. Beyond that, arrangements above a certain value and any expedited response require headquarters approval, and that approval takes days to move through the Japanese internal sign-off chain. This is not a time zone problem. It is a design question about where decision rights sit. The thinking on how far the local site should be allowed to decide on its own is covered in how to divide roles between headquarters and the local site when rolling out systems to overseas plants, and it applies here without modification.

What shortens it is putting sales, headquarters and the plant in front of the same view of order and forecast information, and moving the planning cycle from weekly to daily. Daily does not mean rebuilding the plan from scratch every morning. Freeze the confirmed portion and refresh only the next few days each day, and the average cycle wait falls sharply. What does not shorten it is unambiguous: equipment renewal, extra headcount and overtime contribute nothing whatsoever to this layer. The information has not arrived, so adding people and machines that could act changes nothing about the fact that everyone is waiting.

Layer 2 | Waiting on Material, or Procurement

The symptom is a plan that exists while the material does not. The shortage is discovered on the day production was supposed to start, and the plan is rebuilt. Rebuilding it adds changeovers, which pushes back the start of some other product. In plants where this chain has become routine, most of the planner’s job has quietly become replanning.

There are two causes. The first is that requirements are fixed too late. When design changes or specifications land late, part ordering starts late, and material arrives late by the same margin. The second is that procurement lead time is never actually measured. In many factories the procurement lead time held in the master data is still the nominal figure a supplier quoted several years ago. Even where ocean freight congestion and customs clearance routinely push the real figure past the nominal one, the system keeps calculating reorder points from the stale number. The result is that orders which are “in time” on the system are not in time in reality.

The form this takes in a Japanese-owned plant in Thailand is the swelling of Layer 2 caused by the shift of procurement toward China. In plants where a substantial share of parts has been replaced by imports, the material wait has moved outside the factory, into transport and customs. No amount of improvement inside the four walls shrinks that layer. The structural change behind this is treated with figures later in the article.

What shortens it starts with measuring procurement lead time. Accumulate actual order and receipt dates by item and by supplier, and update the master data from the measured distribution rather than the nominal value. From there, fix requirements earlier and revisit reorder points. Inventory level design itself is covered in how to set optimal and safety stock levels, and the mechanics of ordering and delivery control in how to think about order and purchasing management systems, so please go there for formulas and product selection. What matters here is that Layer 2 involves external commercial terms and is therefore the slowest layer to move, which is exactly why it should not be the layer you start with. What does not shorten it is raising safety stock across the board. That merely converts waiting into inventory cost while leaving the underlying variability in lead time untouched.

Layer 3 | Waiting to Start, or Sequence and Lot Size

The symptom is work not starting even though the machine is free. Walk onto the floor and the machine is running, but what it is running is not the item due today. It is a different item that shares a setup and is therefore being batched through.

The cause is that the shop floor is quite correctly averse to the cost of a changeover. On processes where setup takes real time, running items of the same die or the same material together raises machine utilisation. If the shop floor’s KPI is utilisation, that judgement is entirely rational from where they stand. Seen from plant-wide lead time, however, every lot that got batched behind is simply waiting its turn. Utilisation and lead time are frequently opposing indicators, and if you leave the trade-off to the shop floor without deciding which one wins, utilisation wins.

The form this takes in a Japanese-owned plant in Thailand is a lot size inherited unchanged from the mass-production era in Japan. At start-up the plant ran large volumes of identical items, so a large lot was rational. Today the mix has shifted to high-mix, low-volume, yet the standard lot quantity still sits at the value first loaded into the master, and even a small order gets planned in large-lot units. The plant makes quantities nobody needs, and the next item waits for exactly that long.

What shortens it is short-interval scheduling, meaning a detailed start sequence covering a few days to a week, together with changeover sequence optimisation. The aim is not to eliminate changeovers but to sequence work so that fewer of them are needed. Experienced planners can do this by feel up to a point, but the combinations explode as process and item counts grow, and this is where a production scheduling system (APS) finally comes into its own. Targets such as a 50 percent reduction in setup time are cited as achievable outcomes of planning system deployment (Yachiyo Solutions). What does not shorten it is overtime and weekend shifts. Attacking a sequencing problem with additional hours leaves the queue order for the waiting item unchanged while raising total operating hours.

Layer 4 | Stagnation Between Processes, or Work-in-Process

The symptom is visible to anyone who walks the floor. Pallets and trolleys line the aisles between processes, and the downstream process only learns at the following morning’s briefing that the upstream process has finished. The two are a few metres apart physically, yet the information takes half a day to a full day to travel.

The cause comes down to results not being recorded at the point they occur. In most plants, shop floor data is written on paper day sheets, carried to the office at the end of the shift, and keyed in the next morning. Under that routine, the picture of the factory held in the system is permanently half a day to a day out of date. The planner is making today’s decisions from yesterday’s picture while the shop floor is making a different decision from the physical goods in front of it. Of course the two diverge.

The form this takes in a Japanese-owned plant in Thailand is day sheet entry concentrated on a single Thai administrative staff member, so that the plant’s entire shop floor data feed stops when that person is away. The dependency on one individual is a problem in itself, but the more serious consequence is that this routine makes it structurally impossible to measure lead time process by process. If results exist only at daily granularity, how many hours a given item waited between process A and process B is unknowable, permanently.

What shortens it is moving shop floor data collection to the moment the work happens. Scan a barcode with a handheld terminal at the point a process completes, or take the signal from the machine. Selection of the mechanism is covered in shop floor data collection with handheld terminals and barcodes, so this article will not go into it, but the essential requirement is a state in which completion is transmitted downstream the instant it happens. Achieve that and WIP visibility and downstream notification follow automatically. What does not shorten it is standing up at the morning briefing and calling for closer coordination. Coordination is not poor. The information that coordination requires is not being generated at all.

Layer 5 | Waiting to Ship, or Consolidation and Closing

The symptom is finished goods sitting in the despatch area for days without moving. Manufacturing is complete but shipping has not happened. This layer is easily overlooked by teams focused on improvement inside the production area, yet when it is actually measured it often accounts for a surprising number of days.

There are two main causes. One is waiting for transport consolidation. Where the plant aligns to two truck departures a week, anything finished the day after a departure waits several days for the next one. The other is waiting on paperwork, because inspection certificates and traceability documents are produced by hand, so more time passes between the product being finished and the documents being complete.

The form this takes in a Japanese-owned plant in Thailand is the inspector typing measurement values into a spreadsheet by hand, producing separate Japanese and English versions of the certificate, and the goods waiting for a Japanese manager’s signature before they can ship. If that manager is travelling, the product waits in the warehouse for exactly that long. A process intended to assure quality has been converted directly into lead time.

What shortens it is digitising inspection records and revisiting shipping units. If measured values are captured digitally at the point of measurement, producing the certificate becomes an output step. Approval, likewise, stops being a wait once authority is defined clearly and signatures move to electronic approval. On shipping units, compare the transport cost and lead time trade-off quantitatively once, then judge whether more frequent departures are worth paying for. What does not shorten it is skipping inspection. That trades lead time for a different risk, which is not improvement.

Layer Versus Measure | Which Measures Actually Work Where

With the five layers laid out, the next step is to confirm which measure maps onto which layer. Decide on investment while that mapping is vague and you will spend money on measures that cannot touch the layer that is hurting you. The table below sets out how strongly four representative measures act on each layer.

The column that deserves the closest attention is the one on the far right, which is filled almost entirely with “no effect”. Equipment renewal, extra headcount and overtime are the three responses a factory reaches for first when it is missing delivery dates, and they barely touch any of the five layers. Layer 3 reads “almost no effect” because adding machines does in theory raise parallelism, but that is buying your way around a sequencing problem, and unless the changeover sequence is optimised the same queue will form in front of the new machine too.

The second thing to read out of the table is how often the word “required” appears in the data collection column. Measured procurement lead times in Layer 2 and standard times in Layer 3 simply do not exist unless results are being captured. Investment in Layer 4, then, does not merely shorten Layer 4. It is a precondition for shortening the others. From that structure a rational investment sequence follows: Layer 4 to measure, Layer 1 to cycle faster, Layer 3 to re-sequence, then Layers 2 and 5.

Why Starting With APS Ends in Failure

This is the section that matters most in this article. Most factories that begin evaluating systems in order to shorten lead time put a production scheduling system (APS) at the top of the candidate list. The appeal is understandable. In a demo, the Gantt chart visualises load across every process and a sequence that maximises on-time delivery is generated automatically. Replacing two full days of spreadsheet planning with a few minutes of computation is a compelling pitch. And APS genuinely is a powerful tool when used on the right foundations. The question is what happens when it is dropped into a plant where those foundations are absent.

The Accuracy of an APS Output Never Exceeds the Accuracy of Its Input

To build a schedule, an APS needs the duration of each process, which is to say standard times. Which item, on which machine, with how much setup time and how much run time. Without those numbers the APS can compute nothing. And the number of factories that hold standard times derived from actual measurement is not large.

When an APS project starts in a plant without standard times, one moment always arrives. The vendor asks for standard times by item and by process to be loaded into the master. Nobody has measured values, so the person responsible fills the fields with whatever is at hand: the assumed hours used in a quotation, a theoretical figure back-calculated from a drawing, or the recollection of a veteran. There is no bad faith involved. There is nothing else to fill them with.

Standard times entered this way carry the same family of distortions every time. First, setup time is either absent or estimated far too low, because quoted hours were built around the processing itself. Second, the rework, re-machining and material-wait interruptions that actually occur regularly are not represented. Third, differences in operator proficiency are not reflected. The same process takes different amounts of time for a veteran and for a newcomer, and in plants in Thailand with meaningful staff turnover that gap cannot be waved away.

What the First Three Months Actually Feel Like

Run an APS in that state and the schedule on day one is beautiful. The Gantt chart is packed without gaps and late orders display as zero. The project team is certain it has succeeded.

The drift starts on day two. A process the previous plan had finishing by late morning is, once setup is counted, still running in the late afternoon. Everything queued behind it on that machine slides back. The planner enters the results the next morning and reruns the calculation. Now the system, trying to recover the delay, returns a schedule that has substantially re-sequenced the original order. From the shop floor’s perspective, the order they were told to run yesterday is a completely different order today.

By day three, the shop floor starts asking whether this sequence is really right. The planner cannot answer. To explain why the system produced that sequence, you have to assume the standard times are correct, and the planner is the person most doubtful about exactly that.

By the end of the first week, the shop floor starts protecting itself. The system’s schedule is printed and pinned up, but the actual running order is decided by the supervisor on a whiteboard. The operation is now identical to what it was before the project began. The only difference is one added task, morning data entry.

By the end of the first month, the planner’s job has changed character. Instead of doing what they should be doing, which is deciding how work should flow, they spend most of their hours manually correcting the system’s output so it matches reality. This is the point at which the team finds it is spending more time than it did when planning was done in a spreadsheet.

By the third month, nobody looks at it. Someone mentions the APS schedule in a meeting and the room goes subtly quiet. The licence fee continues to be paid, and at the following year’s budget review renewal is dropped on the grounds that no effect can be measured. The worst part is what the failure teaches the organisation, namely that installing systems does not change the shop floor. The next time someone proposes a genuinely necessary investment, that memory is produced as the argument against it.

The Cause of Failure Is Sequence, Not Product Selection

It is worth stressing that none of this is a judgement on any APS product. Whichever product you choose, the ending is the same if standard times are missing. Indeed, the point that scheduler accuracy is bounded by input data quality is one the vendors themselves make repeatedly (Asprova, Totec Sangyo among others). Product comparison as a discipline is handled separately in comparing and choosing production management systems, so this article will not repeat it. The claim here is a single one: APS is a Layer 3 tool, and it does not function in a plant where Layer 4 is not in place.

What Changes If You Fix Layer 4 First

Consider the reverse sequence. Start with Layer 4, meaning shop floor data collection. Start with a small number of high-load processes, and put in a mechanism that records process start and completion at the point of work. There is no need to cover every process at once. Accumulate a few weeks to a few months of data and you acquire things that did not previously exist.

First, the real duration of each process becomes visible as a distribution, not just an average but a spread. That distribution is the evidence base for the standard times you will eventually feed an APS. Second, dwell time between processes becomes visible. How many hours accumulate between which pair of processes, and therefore how much of the lead time Layer 4 is actually consuming, becomes a number. Third, the frequency of rework and interruption becomes visible, which tells you whether to fold it into standard time or manage it separately.

Deploy an APS from that position and the very first schedule is roughly right. When it drifts, the reason can be traced back through the results. The shop floor accumulates the experience that the system is broadly correct, and the probability that a plan is followed rises. A KPI such as a plan achievement rate of 95 percent or better only becomes a realistic target when this sequence has been respected (Yachiyo Solutions).

That the presence or absence of measurement determines the quality of decisions can also be read from data in a different context. Among plants surveyed on unplanned equipment stoppages, the share answering that they had no countermeasure was 6.7 percent at plants running a maintenance system versus 47.7 percent at plants running on paper, a very large gap (n=500, October 2025 survey, Yachiyo Solutions). The same survey found 68.6 percent citing faster improvement cycles as a challenge. Where data is not being captured, teams cannot tell what a countermeasure should even be aimed at. Exactly the same thing happens in production planning.

If APS Has to Come First Anyway

In reality, headquarters sometimes has already decided that an APS is going in. Even then, it is worth putting three conditions on the table. First, limit the scope to a subset of lines that includes the bottleneck process rather than the whole plant. Second, measure the standard times on that line for at least several weeks before go-live. Third, set the post-go-live evaluation metric as the proportion of work that flowed as planned, and refuse to evaluate on how the Gantt chart looks or on how much planning time was saved. Holding to those three alone sharply reduces the odds of the three-month collapse.

How to Measure | Look at Variability, Not the Average

The most common measurement error in lead time reduction work is tracking average lead time alone. The average did fall by several days. The improvement succeeded. Yet on-time delivery over the same period has not improved at all. This outcome is not unusual.

The reason is in the distribution. Orders that missed their promised date are not sitting near the average. They are in the right-hand tail, the small number of cases that took exceptionally long. Even if the lead time of the great majority of orders falls, the number of late deliveries will not drop while that tail remains intact. If anything, a lower average widens the gap to the tail and strengthens the impression of a plant that occasionally takes an alarmingly long time. What the customer experiences is not an average. It is the individual outcome of whether their own order was late.

What should be measured, therefore, is the average plus the variability, meaning the spread of the distribution, and how far the long tail extends. In practice, a workable routine is to sort every order in a month by lead time descending, take the top decile only, and stratify the causes. Waiting on material. Waiting on headquarters approval. Stagnation ahead of process C. Assign each to whichever of the five layers it stalled in. Keep it up for a few months and it becomes clear which layer your long-duration orders concentrate in. That is the layer to invest in.

The Minimum Data Needed to Measure by Layer

To sort orders into five layers, the timestamps that mark the boundaries of each layer have to be recorded. At minimum, capture the following points in time.

If reading that list prompts the realisation that process start and completion times are not captured at your plant, that realisation is itself the reason to begin with Layer 4. Put the other way round, once these timestamps exist you do not need an expensive analytics tool. Stratification can be done in a spreadsheet.

Setting the KPIs

For target setting, published practice offers usable reference points. Levels put forward as improvement targets include bringing manufacturing lead time to 70 to 80 percent of its current value within three to six months, cutting setup time by 50 percent, and holding plan achievement at 95 percent or better (Yachiyo Solutions). What matters more than the numbers themselves is bounding the period and watching several KPIs at once. Chase lead time alone and inventory rises or quality slips. Watching plan achievement alongside it keeps the goal on work flowing as planned rather than on compression at any cost.

Three Waits Specific to Japanese-Owned Plants in Thailand

Everything above applies to manufacturing generally. Japanese-owned plants in Thailand carry three additional waits that either do not exist in a domestic Japanese plant or are minor there. Miss these three and you end up transplanting a Japanese improvement method intact, only to find it produces nothing.

Wait 1 | Forecasts and Approvals From Headquarters in Japan, Which Inflate Layer 1

The forecast and approval wait touched on under Layer 1 is not an exception in Japanese-owned plants in Thailand. It is the standard configuration. Here the point is to look at it from the countermeasure side. Forecasts are monthly; the firm order follows. On top of that, decisions about capital investment, expedited response and additional headcount require headquarters approval, and that approval travels the sign-off chain in Japan.

The easy misreading is to treat this as a time zone problem. Thailand and Japan are two hours apart and working hours overlap almost completely. The problem is not the time difference. It is the hierarchy of decision rights. If the range the local site is permitted to decide has been designed narrowly, every judgement outside that range becomes a wait. The remedy is not better communication tools but a revision of the authority rules. Draw the boundary by value or by scope of impact, state the range of local decision rights explicitly, and settle everything inside that range locally. On the system side, make it visible from both sites where an item requiring approval is currently sitting. That alone eliminates the wait in which days pass while nobody knows who to ask.

Wait 2 | The Procurement Shift Toward China, Which Inflates Layer 2

The environment surrounding manufacturing in Thailand has changed structurally over the past decade. According to JETRO’s analysis, China’s share of Thailand’s trade expanded from 4.7 percent in 2000 to 19.1 percent in 2024. Japan’s share of direct investment into Thailand, meanwhile, fell from 37.6 percent in 2014 to 8.6 percent in 2024. Manufacturing’s share of Thai GDP has also declined from its 2010 peak of 30.9 percent to 24.3 percent in 2024, and manufacturing growth has stayed negative, at minus 2.7 percent in 2023 and minus 0.5 percent in 2024.

What these figures mean on the shop floor is, first, that parts sourcing has moved from domestic suppliers to overseas suppliers centred on China, which makes procurement lead time longer and less stable. Second, the local layer of Japanese suppliers has thinned, so parts that used to arrive the next day because the supplier was close no longer do. Layer 2, in short, is structurally inflating.

There are two directions of response. One is to update master data with measured procurement lead times and align reorder points to reality. The other is to accept that Layer 2 is long and shorten the other layers to absorb it. Shorten the internal lead time from material receipt to shipment and the plant becomes more tolerant of variability on the supply side. Working to shorten Layer 2 and designing to withstand the length of Layer 2 are different projects, and the latter is the one you actually control.

Wait 3 | Wage Growth Has Broken the Practice of Adding People to Protect Delivery Dates

The lead time countermeasure used longest in Japanese-owned plants in Thailand is, in truth, sheer manpower. When a delivery date looks at risk, add people, work overtime, come in at the weekend. In an era when labour costs were comparatively low, that was a rational choice.

That premise is eroding. Thailand’s minimum wage has applied at a level of 400 baht per day from 1 July 2025 across all industries in Bangkok among other areas. Wage growth at Japanese companies is running from 3.8 percent in 2023 to 4.58 percent in 2024, with 4.64 percent projected for 2025 (Tokyo Consulting Group, Kurasu Asia). If increases in the 4 percent range continue, the cost of protecting delivery dates with manpower becomes impossible to ignore within a few years. And as shown above, extra headcount and overtime have almost no effect on any of the five layers. So the only cost that is rising is the cost of the measures that do not work.

Put the three together and the direction for a Japanese-owned plant in Thailand becomes clear. Shrink Layer 1 through authority design and systems. Accept that Layer 2 will not shrink and design to withstand it. And convert what has been absorbed by manpower into measurement in Layer 4 and re-sequencing in Layer 3.

Cost and Investment Sequence | Which Layer, in Which Order

This article deliberately avoids quoting specific figures when discussing cost. The order of magnitude changes with the number of processes, the number of sites, the scope of integration with existing systems and whether multilingual support is required, and a figure quoted without its preconditions leads to bad decisions. Cost ranges for process management systems are set out with their preconditions in how to estimate the cost of a process management system, so please go there for a sense of market pricing. What is offered here is not a price tag but the order of investment and the relative scale of each layer.

The basis for this order is what the article has already shown, namely that Layer 4 is the precondition for the others. Skip Layer 4 and invest in Layer 3 and you get the failure described in the middle of this article. Skip Layer 4 and invest in Layer 2 and you end up designing reorder points on procurement lead times nobody has measured. Respecting the sequence delivers more than reducing the amount spent.

How to Frame the Return

When pushing an investment through internally, many people try to explain the return through inventory reduction. Less WIP means less working capital. That argument has a weakness. Inventory reduction is a one-off effect and, in accounting terms, an exchange of assets, which makes it hard to rebut when a senior executive points out that it is not profit.

What this article recommends instead is framing the return as order opportunities won through better on-time delivery. A plant that meets its dates can accept short-lead-time work and sudden volume increases. A plant whose dates are unpredictable quotes conservatively long lead times and loses to competitors. Ask the sales team which enquiries were lost and how many, and concrete examples will come back. Build the argument from there: if the variability in lead time narrows, the plant can quote shorter dates and still meet them. The monetary precision is lower, but for an executive audience it is a far more legible story than swapping one asset for another.

How to Run the First 90 Days

Before sketching a grand system architecture, decide what happens in 90 days. What follows maps the five-layer decomposition and the investment sequence directly onto a calendar.

The thing to hold onto through these 90 days is not to over-reach on scope. Take every process and every item at once and the 90 days will be consumed by data preparation alone. One line is enough. If you can decompose five layers on one line, the method transfers to the others. Push forward with a wide scope and vague data and you will still not know which layer is at fault at the end of it, and you will end up choosing where to invest on instinct.

Five Common Failures

Finally, here are five failure patterns observed repeatedly in lead time reduction work. All five are explainable within the framework above.

Failure 1, upgrading the bottleneck machine and calling it done. As long as processing is a small fraction of the total, the effect of equipment investment is limited. The structure that produces waiting before and after the upgraded process is unchanged, so the usual result is that the pallets have simply moved to a different spot.

Failure 2, tracking average lead time only. A falling average does not raise on-time delivery while long-duration orders remain. What the customer evaluates is not the average but whether their own order arrived as promised.

Failure 3, skipping data collection and installing an APS. As set out at length in the middle of this article, running an APS without standard times produces schedules that do not match the shop floor and stop being used within months. What is lost is not only the licence fee but the organisational lesson that systems are not useful.

Failure 4, raising safety stock to make shortages disappear. The Layer 2 wait vanishes on the surface, but the variability in procurement lead time that caused it is still there. The bill is simply being paid in inventory cost and cash flow, and lead time itself has not moved.

Failure 5, taking every process and every item at once. The wider the scope, the heavier the data preparation load, and the project exhausts itself before any effect appears. Decomposing five layers on a single line and rolling out the resulting method is faster.

Frequently Asked Questions

What exactly is a lead time reduction system?

It is not the name of a specific product category. It refers to information systems in general that shorten the waiting which makes up manufacturing lead time. In the framework used here, that covers mechanisms for shop floor data collection and visibility (Layer 4), for sharing order information and shortening the planning cycle (Layer 1), production scheduling software that optimises start sequence (Layer 3), procurement management (Layer 2) and digitised inspection records (Layer 5). What matters is measuring which layer your lead time accumulates in before you choose anything. Select a product without identifying the layer and you will invest in a layer that was never the constraint.

How do you calculate manufacturing lead time?

The most practical method is to record, for each individual order, the timestamp at which the order was registered internally and the timestamp at which the product shipped, then analyse the resulting set. Look at variability and at how far the long tail extends, not only at the mean. To stratify further, record plan confirmation time, material receipt time, start and completion of each process, and inspection completion, then decompose the total into intervals. In plants where process start and completion times are not recorded, this decomposition is impossible, which makes establishing shop floor data collection the starting point.

Will installing a production scheduling system shorten lead time?

Conditionally, yes. A production scheduling system (APS) acts on Layer 3, meaning the problem of start sequence and lot size. For it to act, standard times for each process must be in place and derived from measurement. Deploy it without standard times and the common outcome is schedules that are produced but do not match shop floor reality, and that stop being used within months. The correct sequence is to establish Layer 4 data collection first, hold measured data as standard times, and then deploy, in which case a target such as a plan achievement rate of 95 percent or better also becomes realistic.

Where should we start when moving production planning off spreadsheets?

The important thing is not to make replacing the spreadsheet the objective. The reason spreadsheets are impossible to give up in most plants is not that the planning logic is complex. It is that shop floor results exist nowhere except the spreadsheet. The starting point, therefore, is introducing a mechanism that records process start and completion at the point of work, on one line first. Once results flow into a system, the input data for planning assembles itself and the re-keying that used to happen in a spreadsheet naturally falls away. Replace only the planning tool while results stay on paper and the re-keying simply moves to a different tool.

If we reduce work-in-process, will lead time shorten?

Be careful about the direction of causality. Work-in-process is a consequence of long lead time, not its cause. WIP piles up because stagnation is occurring between processes, so forcing WIP down while the information delay causing the stagnation remains merely stalls the upstream process instead. The correct sequence is to measure dwell time between processes first, then create a state in which the downstream process knows immediately that the upstream process has finished. WIP falls as a result. WIP volume is better used as an indicator confirming that improvement is progressing than as the target of improvement itself.

What should we watch out for specifically at a plant in Thailand?

Three things. First, waiting on forecasts and approvals from headquarters in Japan inflates Layer 1, and this is a question of how decision rights are designed rather than of time zones. Writing down the range of local decision authority is the countermeasure. Second, the shift of procurement toward China has structurally lengthened Layer 2, and because internal effort will not shorten it, the design has to shorten other layers on the assumption that it stays long. Third, wage growth. The minimum wage has applied at 400 baht per day from 1 July 2025 across all industries in Bangkok among other areas, and wage growth at Japanese companies is running in the 4 percent range annually. Protecting delivery dates through added headcount and overtime is both weak in effect and rising in cost.

Summary

Most of manufacturing lead time is not processing but waiting, and waiting is the time during which the person who must decide does not have the information they need, at the moment they need it. That is why faster equipment, more people and more overtime do not shorten it. It shortens only when the information arrives.

That waiting decomposes into five layers. Waiting on information, waiting on material, waiting to start, stagnation between processes, and waiting to ship. A different measure works in each, and Layer 4 shop floor data collection is additionally the precondition for analysing all the others. The investment sequence therefore runs: measure (Layer 4), cycle faster (Layer 1), re-sequence (Layer 3), then procurement and shipping (Layers 2 and 5). Skip that sequence and start with a production scheduling system, and you will get schedules that do not match the shop floor and that nobody looks at within three months.

Only one question is needed for tomorrow’s meeting. Where, across the five layers, is our lead time accumulating, and how many days are sitting in each. If you cannot answer that with numbers, making it answerable is the first investment.

It is entirely fine if you have not yet worked out which layer your lead time is accumulating in. TOMAS TECH supports Japanese-owned manufacturers in Thailand and across ASEAN on the ground, from shop floor data collection through to building the planning mechanism. We are happy to start by mapping out your current process flow together, so please get in touch via our contact page.

Reference Sources

• Lead time reduction methods and how to set KPIs (Yachiyo Solutions) https://yachiyo-sol.com/library/leadtime-tanshuku/
• Lead time reduction in manufacturing (Yachiyo Solutions) https://yachiyo-sol.com/library/lead-time-tanshuku-seizogyo/
• Special report on structural change in Thai manufacturing (JETRO) https://www.jetro.go.jp/biz/areareports/special/2025/1001/88b00134e345c346.html
• Stagnation between processes and the share of processing time (Factory Advance) https://factoryadvance.jp/blog/1295/
• The waste of stagnation (Zaiko Kanri 110ban) https://shikumika.com/column/%E5%81%9C%E6%BB%9E%E3%81%AE%E7%84%A1%E9%A7%84/
• APS and lead time (Totec Sangyo) https://www.totec-sangyo.jp/blog/aps-leadtime/
• Production schedulers and planning accuracy (Asprova) https://www.asprova.jp/column/a2/productionscheduler-planningaccuracy/
• Lead time reduction in practice (Saiteki Works) https://saiteki.works/blog/cont-shorten-lead-time/
• Latest trends in Thai labour affairs and labour management (Tokyo Consulting Group) https://kuno-cpa.co.jp/thailand_blog/%E3%80%902026%E5%B9%B4%E6%9C%80%E6%96%B0%E7%89%88%E3%82%BF%E3%82%A4%E5%8A%B4%E5%8B%99%E3%83%BB%E3%82%BF%E3%82%A4%E5%8A%B4%E5%8B%99%E7%AE%A1%E7%90%86%E3%81%AE%E6%9C%80%E6%96%B0%E5%8B%95%E5%90%91/
• Thailand minimum wage guide 2025 (Kurasu Asia) https://kyujin.careerlink.asia/blog/thailand-minimum-wage-guide-2025/