Views: 0 Author: Elsa Long Publish Time: 2026-11-16 Origin: Site
Cost model guide · Published by I.C.T · For SMT process engineers, line leads and equipment buyers
Two factories buy the same model and one reports payback in fourteen months while the other is still arguing about it in year three. The machine is not the difference. The difference is in what was attached to it, what the line runs, how often the line changes over, how much of the scheduled time it actually places boards, and what a defect costs when it reaches the customer. A JUKI SMT machine cost ROI calculation that stops at the quotation has captured perhaps a third of the money involved, and usually the third that matters least.
The model below separates cost into three clocks: what you pay once, what you pay every year, and what you pay every time the line changes over. It then builds the return from three levers — utilization, yield and changeover — against one denominator, the capex actually required to run your product mix. Prices vary by region, by configuration and by year, so no absolute figures are asserted here; the worked examples use index units and clearly labelled assumptions that you can replace with your own quotation. The platforms referenced throughout are drawn from the placement platforms this cost model is applied to, and every number you see in an example is a placeholder, not a price list.
Equipment buyers are trained by the sales process to compare a single number. The quotation arrives, the numbers are put in a row, the lowest is shortlisted. That works if the machines are otherwise identical, the products are identical and the factories run identically — which is never the case. A JUKI SMT machine cost ROI comparison that starts and ends with the quotation price is comparing an entry fee and calling it the cost of ownership.
The reason is not that capital cost is unimportant. It is large, it is paid early, and it is hard to reverse. The reason is that it is the only part of the cost that is fixed, and everything else — consumables, labour, uptime, changeover, scrap — is decided later, by choices that are still open when the order is signed. Those later numbers are usually several times larger over the life of the line, and they are the ones a buyer can still influence.
The table below is the whole model in one view. The sections that follow take one row at a time and show what to measure and what to ignore.
Clock | What it contains | When it is decided | How it is measured |
|---|---|---|---|
Once (capex) | Machine, feeders, nozzles, options, handling, software, installation, spares package | At order, and again at any upgrade | Total delivered and commissioned cost, not the machine price |
Every year (recurring) | Consumables, spares, maintenance labour, power, air, floor space, training | Continuously, by discipline | Cost per million placements, tracked quarterly |
Every event (changeover) | Lost output, labour, scrap at start-up, programming time | By product mix and configuration | Minutes and boards lost per changeover, times events per month |
Start with the product instead. A machine is a way of converting components into boards at a rate, a quality and a cost, and every one of those three depends on what the board is. The mix determines how many feeder positions are needed at once, how many different nozzles are touched in a shift, how often the line changes over, and how much of the platform's headline speed is usable. Begin there and the machine comparison becomes a consequence rather than a starting point.
This order also protects against a common failure: buying a platform sized for a product that never arrives. A machine chosen because it can place 40,000 components an hour looks like good value until the actual mix runs at a fifth of that rate and the changeover count is four times the plan. The machine is not underperforming. It was priced against a scenario that did not happen, and the cost model never had the product in it.
The practical version of this advice is to build a one-page product profile before the shortlist: board sizes, placement counts, component mix by package, monthly volumes, number of active products and how often they rotate. That page is the input to every number in this guide, and it costs an afternoon.
Capital cost is paid once and amortised. Recurring cost is paid every month, forever, and scales with volume rather than with the purchase. Event cost is paid every time the line stops to become a different line, and it scales with the number of products rather than with either of the other two. Mixing them produces the two mistakes buyers make most often: comparing machines on price alone, and underestimating the cost of a high-mix schedule.
Keeping them apart makes the arithmetic honest. A machine that costs more but consumes fewer nozzles per million placements has a different total than one that is cheaper and hungrier, and the crossover point can be calculated. A machine that is fast but needs eight hours to change over has a different total from one that is slower and changes over in two, and the difference depends entirely on how many changeovers your schedule contains.
Each clock also has a different owner. Capex belongs to finance and the capital committee. Recurring belongs to production and maintenance. Event cost belongs to planning and process engineering, because it is created by the product schedule. A cost model that assigns all three to the purchasing department will be accurate for one of them.
Quotations describe machines, not lines. The machine price typically excludes the conveyor sections that connect it to its neighbours, the loader and unloader, the stencil printer and its options, the inspection equipment, the software licences, the spare parts package, the freight, the installation and the training. None of that is hidden; it is simply outside the scope of the document you are comparing, and every supplier leaves out roughly the same things.
The consequence is that comparing machine prices compares the same partial picture, which is fine for choosing between machines and useless for deciding whether the project pays back. Build the delivered and commissioned cost for each option, including the items above, and the ranking can change. A cheaper machine that needs an extra conveyor section and a different feeder cart can be more expensive than the alternative before either one produces a board.
Ask for the excluded items in writing rather than inferring them. A supplier who lists the exclusions clearly is easier to compare than one who does not, and the list itself is a useful check on your own project budget, which usually has the same gaps.
Within a platform family, the purchase price moves for five reasons: the size class of the machine, the number and type of heads, the feeder positions and nozzle set, the options that are licensed rather than built, and the integration items that turn a machine into a line. Understanding which of the five you are actually paying for is what separates a justified configuration from a comfortable one.
The useful discipline is to attach every increment to a product requirement. A wider board class is justified by a panel you expect to run. An extra head is justified by a placement count that the current head count cannot reach inside the takt time. An extra feeder cart is justified by a changeover count that a single cart cannot cover. An increment without a requirement behind it is a cost that will never earn anything.
Cost driver | What it buys | Justified when | Common over-buy |
|---|---|---|---|
Board size class | Panel capacity and future product range | Roadmap includes a larger panel within the asset life | Buying two classes up for a product that never arrives |
Head count and rate | Placements per hour at a given mix | Placement count and takt time require it | Paying for optimum rate the mix cannot use |
Feeder positions | Parts online at the same time | BOM needs the positions to run without a cart swap | Maximum positions for a BOM that fits in half of them |
Nozzle set | Package coverage and pickup window | Smallest and largest parts in the mix require the bores | Full kit bought once, half of it never used |
Options and licences | Vision, traceability, reporting, software | A customer, audit or process requires the output | Licences bought for a capability nobody operates |
Board size is the single largest step in price within a family, and the least reversible. A machine sized for the panels you run today will be sized for them for its whole life, and panels tend to grow rather than shrink. The decision is not to buy the largest available; it is to know where your roadmap sits relative to the class boundaries and to pay for the class you will need within the depreciation period.
Make the judgement with a number rather than a feeling: the largest panel dimension you expect in five years, and the largest the machine handles. If the gap is one step, the decision is close and the rest of the configuration should decide it. If the gap is two steps, the machine will limit the business and no amount of programming will recover it. The selection logic for this trade-off is set out in the platform shortlist that sets the price bracket, and it belongs before the quotation, not after it.
One practical caution: the wide-board option is sometimes a handling option rather than a placement capability. Confirm what the extra width actually buys — board support, conveyor length, or genuine placement area — because paying for width that the head cannot reach is paying for steel.
Head count drives both price and the shape of the program. More heads place more components per pass, but only when the parts are distributed so that each head has work in the same region. A high head count on a board with a clustered layout produces idle heads and a rate well below the headline. The right question is not how fast the machine is but how much of that speed your board can absorb.
Estimate it before you buy: placement count per board, divided by the effective rate you expect after handling, fiducial reads and distribution losses, gives seconds per board. Compare that with the takt time your volume requires. If the machine has margin, the extra heads are buying flexibility for future products rather than capacity for this one, and that is a legitimate reason — as long as it is stated as one.
There is also a focus-versus-flexibility trade. A platform built around a specific product family can be very efficient at that family and poor outside it; a dedicated LED mounting platform is the obvious example, and a general-purpose chip mounter is the opposite bet. Neither is cheaper in the abstract. One is cheaper if your mix stays narrow, the other if it does not, and the cost model should say which assumption you are making.
Feeder positions are usually quoted as a maximum, and bought as a number. The maximum is a property of the bank; the number you need is a property of the BOM. Convert every part to an 8 mm equivalent using the tape width ladder, add the parts that must be online simultaneously, and compare the total with the usable positions. If the product needs more positions than one cart provides, the second cart is part of the capex, and so is the changeover time it removes.
Nozzle sets follow the same logic at a smaller scale. A standard kit covers the common bores; the smallest and largest parts in your mix may need bores outside it, and buying them later costs more than buying them with the machine. Estimate the set from the package list rather than from the catalogue: one nozzle per package family, plus spares for the nozzles that wear fastest, which are usually the small ones running at high cycle counts.
The option bill is where cost accumulates without anyone deciding to spend. Vision packages, traceability links, reporting modules and software seats are licensed rather than built, so they appear as line items that are easy to tick and hard to remove. Attach each to an output somebody will actually use. If no customer, audit or process currently requires the report, the licence is a cost with no return, and it can usually be added later at a similar price.
A placement machine alone does not produce boards. The line around it — loader, printer, conveyor sections, buffers, reflow, inspection, unloader — is part of the capex and often a third or more of it. Buyers who compare machine prices and then build the line piecemeal frequently discover that the interfaces, not the machines, are where the budget went.
Price the line as a line. Ask for the handling equipment, the interface hardware and the software that connects them as one scope, so that the responsibility for a board that does not transfer cleanly sits with one party. Mixed scopes produce the most expensive kind of fault: the kind where two suppliers each demonstrate that their own machine is working.
Include the spares package in the same conversation, because it is capex with a return. A modest package bought at order is cheaper than the same parts bought in an emergency, and it converts a four-week lead time into a four-hour repair. The parts worth including are the ones whose failure stops the line, not the ones that are merely expensive.
Recurring cost is the number that decides whether a machine was a good buy, and it is the number almost nobody models before purchase. It has four components: consumables that wear with volume, spares that fail without warning, labour that maintains and operates, and utilities that run whenever the machine is on. Together they typically accumulate to a meaningful fraction of the original capital over five years, and the fraction varies enough between platforms to change a purchase decision.
The way to keep the model honest is to express each component per million placements rather than per month. A monthly figure scales with the order book and hides efficiency; a per-million figure is comparable across machines, across products and across years, and it is the number that shows whether a change in consumption is a process problem or a volume effect.
Nozzles, filters, belts, lamps and wipes are the small items that never appear in a capital request and quietly become a budget line. Their consumption is driven by placements, by component mix and by how well the machine is maintained. A nozzle that is cleaned on schedule lasts longer than one that is cleaned when someone notices, and the difference shows up as a consumption rate rather than as a failure.
As an illustration only, assume a line running three million placements a month, assume nozzle consumption of two units a month at USD 60 each, and assume filters and wipes at USD 80 a month. That is USD 200 a month, or about USD 67 per million placements. The absolute number will differ in your plant; the useful part is the ratio. If the same line later runs four million placements and consumption rises to USD 400 a month, the rate has risen from 67 to 100 per million, and that is a process signal rather than a volume effect.
Track consumption quarterly against placements, not against calendar time. A per-million trend that climbs is the earliest warning that cleaning intervals, vacuum levels or feeder presentation have drifted, and it usually precedes a placement defect by weeks. Consumables are the cheapest instrument on the line for detecting that drift, provided someone plots them.
Maintenance labour is real cost even when it is absorbed inside existing headcount, because the hours are not available for something else. Estimate it from the schedule: daily checks per shift, a weekly stop, a monthly stop and the unplanned interventions recorded in the stop log. Multiply by the loaded hourly rate and by the number of machines, and the figure is usually larger than the annual consumable spend.
Spares consumption is the part of maintenance that is lumpy. Most years it is small; the year a head board fails it is not. Model it as an average plus a reserve rather than as a monthly line, and fund the reserve from the capital project rather than from the maintenance budget, because the first failure usually happens before the maintenance budget has accumulated anything. The structure that keeps the checks and the spares aligned is set out in the maintenance schedule that keeps the recurring line from drifting, and its value in a cost model is that it converts an unpredictable failure cost into a planned one.
Do not model maintenance as a percentage of the machine price. Percentages borrowed from another factory embed another factory's labour rate, shift pattern and product mix, and they hide the two variables you can actually change: the interval and the discipline.
Utilities are small per machine and large per line. A placement machine draws modest power compared with a reflow oven, but the oven runs continuously, the compressed air system runs continuously, and the air conditioning runs to remove the heat both produce. Model the line, not the mounter, and model it at the duty cycle you actually run rather than at full load.
Floor space is the component buyers forget entirely, because it is already paid for. It still has an opportunity cost: the square metres occupied by a line that runs at 60 percent utilization could hold a line that runs at 85. In a plant that is short of space, that opportunity cost is real and occasionally larger than the machine's depreciation. The calculation is simple — space occupied times the internal rate per square metre — and it belongs in the model even when the answer is uncomfortable.
As an illustration only, assume a placement line drawing an average of 15 kW over 500 scheduled hours a month at USD 0.12 per kWh, which is about USD 900 a month, and assume air and cooling add USD 200. That is roughly USD 1,100 a month, or about USD 367 per million placements at three million a month. The figures are placeholders; the ratio is the point, because it shows how quickly utilities become significant when utilization is low and the same energy serves fewer boards.
Every hour of unplanned stop has a cost, and the cost is not the machine's hourly rate. It is the contribution from the boards that were not built, plus the labour that was paid while the line waited, plus the schedule disruption that pushes other orders. Model it explicitly, because a machine with a slightly higher purchase price and a materially better uptime can be the cheaper asset over five years.
As an illustration only, assume a line that contributes USD 4 per board and builds 1,200 boards a shift, assume 20 shifts a month, and assume 11 unplanned hours a month out of 500 scheduled hours. Those 11 hours are about 26 boards a shift, roughly USD 104 per shift and about USD 2,080 a month. Over five years that is more than USD 120,000 — a number large enough to change which machine is worth buying, and one that never appears in a quotation.
This is why the availability review belongs in the cost model rather than in the technical annex. The availability and support review that decides how much of the year you actually get converts a support promise into a monthly figure, and that figure belongs on the same page as the depreciation.
Changeover cost is invisible in a single-machine comparison and dominant in a high-mix plant. It is paid every time the line stops being one product and becomes another: the stop itself, the labour to swap feeders and load programs, the first-article checks, and the boards scrapped while the process settles. None of those is large individually. Multiplied by the number of changeovers a month, they can exceed the annual consumable spend.
The number of changeovers is a property of the schedule, not of the machine, but the cost per changeover is very much a property of the machine and its configuration. That split matters: it means the buyer can influence the cost per event at the order stage, and the planner can influence the number of events. Neither can fix the other's side.
Take the product list and count how many times the line must switch in a typical month. A plant running three products in long campaigns may change over six times; a plant running forty products in weekly batches may change over eighty. The same machine is a good purchase in the first case and a questionable one in the second, and no specification sheet will tell you which you are.
Count properly, including the switches that are not full changeovers: a feeder swap mid-batch, a nozzle change for one large part, a program reload after an engineering change. Those smaller events cost less each and are frequently more numerous than the planned ones, and they are the ones that erode a schedule without appearing in a planning report.
Then attach a target to the count. If your schedule contains eighty changeovers a month and each costs 45 minutes, you are spending 60 hours a month not producing — six percent of a 1,000-hour month before any unplanned stop. That number, not the machine price, is what justifies a second feeder cart or a configuration that reduces the swap.
Cost per changeover has four parts: lost output during the stop, labour during the stop, scrap from the first boards, and the schedule disruption that follows. Only the first is easy to calculate, and it is usually the largest. Lost output equals the stop time multiplied by the effective rate and by the contribution per board.
As an illustration only, assume a 45-minute changeover, an effective rate of 1,100 boards a shift over 8 hours, which is about 138 boards an hour, and a contribution of USD 4 per board. Lost output is 45 minutes at 138 boards an hour, about 103 boards, or roughly USD 414. Add two hours of two people at USD 25 an hour, which is USD 100, and add the first-article scrap, assume 5 boards at USD 12 material cost, which is USD 60. The event costs about USD 574.
Multiply by eighty events a month and the annual figure is roughly USD 550,000 of contribution at risk, which is a different order of magnitude from the machine price discussion. The assumption set will not match your plant, and it does not need to: the point is that changeover cost is calculated from the schedule, and it is the line item most likely to be left out of a purchase decision.
Three configuration choices reduce changeover cost, and all three are cheaper at order than retrofitted. The first is a second feeder cart, which converts a swap into a trolley exchange and removes the reloading from the critical path. The second is enough feeder positions to keep the common parts loaded across products, so that only the product-specific positions change. The third is a nozzle set that covers the whole mix, so that a nozzle change is not part of every switch.
The fourth is not a configuration choice but a platform one: a machine whose board width class covers your largest panel without a changeover between products. A chip-shooter platform sized for 400 to 450 mm panels removes a class of handling adjustment for most consumer and industrial boards, and that adjustment is often the hidden fifteen minutes in a changeover that was planned as thirty.
Price these against the changeover cost, not against the machine price. A second cart that costs a fraction of a percent of the project and removes twenty minutes from eighty events a month pays back in weeks. That comparison only exists if the changeover count was modelled in the first place.
The return on a placement line is contribution divided by the capital committed to produce it. That is the whole formula, and everything else is detail about how to measure the two halves. The detail matters because contribution is not revenue: it is what remains after material, direct labour and variable overhead, and it is the only part of a board's price that the line can improve by running better. A JUKI SMT machine cost ROI model that uses revenue inflates the return by an order of magnitude and makes every project look attractive.
The denominator is the delivered and commissioned cost of the equipment needed to run the mix, not the price of the machine. The numerator has three levers: how much of the scheduled time the line actually produces, how many of the boards it produces are good, and how much time is lost converting between products. Those three are where every realistic improvement lives, and each of them can be estimated before purchase from data you already have.
Start from the selling price of the board and subtract material cost, direct labour and variable overhead. What remains is contribution per board, and it is the number that a stopped hour destroys. If the plant does not calculate it, the exercise is still worth doing once per product family, because without it every improvement argument becomes circular.
As an illustration only, assume a board that sells for USD 18, carries USD 11 of material, USD 2.50 of direct labour and USD 1.50 of variable overhead, leaving USD 3.00 of contribution. Assume the line builds 1,200 boards a shift over 20 shifts a month, which is 24,000 boards, or USD 72,000 of monthly contribution at full schedule. Those figures are placeholders, but the structure is the point: the line earns contribution, not revenue, and the difference decides whether a project looks viable.
Be careful with the denominator's scope as well. If the project requires a new conveyor section and a second feeder cart, they belong in the capex used for the return calculation. If it requires floor space that could hold a different line, that opportunity cost belongs in the model as a note even when it is not in the spreadsheet, because it will be in the decision.
Utilization is the share of scheduled time the line spends producing good boards, and in most plants it ranges wider than any other variable in the model. A line at 70 percent and a line at 90 percent of the same schedule produce very different returns from identical capital, which is why two factories with the same machine report different paybacks without either being wrong.
Improve it in the order that costs least. First, reduce unplanned stops through maintenance discipline and spares. Second, reduce planned losses by shortening changeovers. Third, reduce scheduling losses by batching products so that the line is not waiting for material or for a decision. Only fourth does adding capacity make sense, and by then the question is different.
The ceiling is set by the platform, and it is worth knowing what that ceiling looks like. The KE-3010A and its published rate and accuracy figures list an optimum of 23,500 CPH with 18,500 CPH under IPC9850 and 9,000 CPH for IC placement, laser accuracy of plus or minus 0.05 mm and vision accuracy of plus or minus 0.04 mm, on board sizes up to 610 by 560 mm. Those are published figures under defined conditions. Your utilization determines how much of that capability reaches a shipped board, and it is the number to attack before considering a faster machine.
Yield is the second lever and the one most often treated as a quality cost rather than a return lever. At 98 percent first-pass yield, two boards in a hundred are reworked or scrapped. At 99 percent, one is. The difference sounds small until it is multiplied by volume, and it compounds because rework consumes the same capacity that could be producing.
As an illustration only, assume 24,000 boards a month and a contribution of USD 3.00 per board. Moving first-pass yield from 98 to 99 percent recovers 240 boards a month, which is USD 720 of contribution, and it frees the rework hours those boards consumed. Over a year, with the rework labour included, the effect is usually larger than the annual consumable budget. The example is an assumption; the direction is not.
Placement is only one contributor to yield, which is why the yield lever should not be attributed to the machine alone. Solder paste printing, reflow profile, handling and inspection all participate. What the placement machine controls is placement accuracy, pickup reliability and the stability of both over a shift, and those are measurable from the first article and the capability run rather than inferred from a specification.
Payback is capex divided by monthly contribution, and it is useful precisely because it is crude. Simple payback ignores the time value of money, the residual value and the years after payback, so it should be paired with a discounted calculation for anything large. But it is transparent, it is comparable across options, and it forces the two assumptions that matter — capex scope and monthly contribution — into the open.
Then test the assumptions instead of defending them. Move each lever by a plausible amount, one at a time, and see how much the payback changes. The lever that moves it most is the one to spend management attention on, and it is frequently not the one the project team has been arguing about.
Lever moved | Change assumed | Effect on monthly contribution | Effect on payback |
|---|---|---|---|
Utilization | 70% to 80% of schedule | About +14% output | Shortens materially |
First-pass yield | 98% to 99% | About +1% good boards, plus rework hours freed | Shortens modestly |
Changeover time | 45 to 30 minutes, 80 events | 20 hours a month recovered | Shortens modestly to materially |
Consumables | 10% lower consumption | Small direct saving | Barely moves |
The pattern in that table is consistent across most plants: utilization dominates, changeover is second, yield is third, and consumables are a rounding error in the return calculation although they are the easiest to measure. Spend the improvement budget in the order the table implies, not in the order the complaints arrive.
Two quotations for comparable machines are rarely comparable documents. One includes a feeder cart, the other lists it as an option; one prices installation, the other leaves it to the buyer; one includes two years of warranty, the other one. Comparing the totals without normalising the scope produces a decision that is defensible in the meeting and wrong in the factory.
The fix is a normalised metric that both documents can be converted into, plus a fixed list of lines that every quote must show. Neither takes long, and both make the conversation about evidence rather than about preference. They also make it much harder for a supplier to win on an omission, which is a service to everyone involved including the supplier who did not omit anything.
Convert each quotation into cost per placed component at your own volumes. Take the delivered and commissioned cost, add the present value of the recurring cost over the asset life, divide by the total placements the line will make in that period at your utilization. The result is a single number that reflects price, consumption and utilization together, and it is the fairest comparison available without a full financial model.
As an illustration only, assume option A at 100 index units with a recurring cost of 8 units a year, option B at 118 units with a recurring cost of 5 units a year, a five-year life and 300 million placements over that period. Option A totals 140 units and costs about 0.47 units per million placements; option B totals 143 units and costs about 0.48. They are effectively tied, which is a useful finding: the more expensive machine justifies itself only if its lower running cost is real and measured, not assumed.
Change the volume and the ranking can move, which is why the calculation belongs in a spreadsheet with the volume as an input rather than in a slide with a conclusion on it. High-volume plants can justify a more expensive platform; low-volume plants usually cannot, because the recurring advantage has too few placements to amortise over.
Ask for five lines explicitly. The delivered and commissioned price, including freight, installation and training. The full configuration as an annex, with feeder count and width mix, nozzle set, head configuration, board handling range and software licences. The recurring cost estimate per year for consumables and recommended spares. The support commitment with response times and coverage hours. And the acceptance criteria that will be used to sign the machine off.
Those five lines turn a quotation into a comparable document. They also expose the assumptions each supplier is making, because a quote that lists a nozzle set will differ visibly from one that assumes a standard kit, and that difference is worth more in the evaluation than a discount of the same value.
Ask for the same five lines from every supplier in the same order. Uniformity is not bureaucracy here; it is what makes the normalised metric possible without a week of follow-up questions. A supplier who resists structuring the quote this way is telling you how the delivery will be managed.
A cheaper machine becomes the more expensive decision in three situations, and all three are predictable. The first is a configuration gap: the machine is cheaper because it lacks feeder positions, nozzles or handling that the product needs, and closing the gap after delivery costs more than buying it in the first place. The second is a utilization gap: the machine cannot reach the rate the schedule needs, so a second machine or overtime covers the shortfall.
The third is a changeover gap, and it is the one buyers miss. A platform that is cheap but slow to reconfigure is a bargain in a plant running long campaigns and a trap in a plant running forty products. The cost does not appear as a machine cost at all; it appears as capacity the plant does not have, month after month.
The counter-example is worth stating too, because focus is a legitimate strategy. A universal platform that combines strip placement and soldering can remove a whole process step for a narrow product family, and removing a step is usually cheaper than optimising one. The test is not whether the machine is specialised but whether your mix stays inside its specialisation for the life of the asset.
The model is not finished when the order is placed. Three things change the return afterwards: the surprises of the first year, the financing structure, and whether anyone revisits the assumptions. The first is avoidable with better questions before purchase, the second is a treasury decision that is often made by default, and the third is a discipline that most plants intend to have and few maintain.
Treating the JUKI SMT machine cost ROI model as a living document rather than a purchase justification is what makes it useful. The figures that were assumptions become measurements in the first quarter, and the model either confirms the decision or explains why not. Both outcomes are more valuable than a spreadsheet that was correct in the meeting and has not been opened since.
Most first-year cost surprises were visible in the quotation, the configuration annex or the site survey, and were simply not converted into money. The four that recur are: a configuration item that was excluded and had to be bought afterwards, a consumable consumption rate higher than assumed because the mix was different, a changeover time longer than planned because the feeder bank was smaller than the schedule needed, and an integration issue between two suppliers that cost engineering hours nobody budgeted.
Each has a question that would have caught it. What is excluded from this price? What consumption rate should we plan per million placements, and what does it depend on? How many changeovers a month can this configuration support before it becomes the constraint? Who is responsible for the interface between these two scopes? Four questions, asked before the order, would have converted four surprises into line items.
None of them requires technical expertise beyond reading the documents carefully. They require the discipline to translate every omission into a number, which is the same discipline that makes the original model worth building.
The financing structure does not change the total cost much; it changes when the cost is paid and who carries the residual risk. Buying puts the full capital on the balance sheet and keeps the residual value. Leasing converts capital into an operating cost and usually bundles service, at a higher total price. Buying used lowers the capital but transfers condition risk and shortens the supported life.
Compare them on the same basis: total cost over the period you intend to keep the asset, including the residual value at the end. A lease that looks expensive per month can be cheaper over three years if it includes service you would otherwise buy, and a used machine that looks cheap can be more expensive over five if the spares are no longer stocked. The comparison is only meaningful over a stated holding period, because the ranking changes with it.
One structural point is worth stating: the financing choice should follow the utilization assumption, not replace it. A lease does not make a line that runs at 60 percent utilization into a good investment; it only changes the shape of a loss. Fix the utilization question first, then choose the structure that fits the cash flow.
Once a year, replace the assumptions with measurements and see what the model says. Placements made, consumables consumed, unplanned hours, changeover count and duration, first-pass yield, and the support interventions actually used. Six numbers, all of which the plant already records if the earlier disciplines are in place.
The review has three possible outcomes and all are useful. The model confirms the decision, in which case the numbers become the baseline for the next purchase. The model shows a gap in one lever, in which case there is a specific project with a specific return. Or the model shows that the original assumptions were wrong in a way that cannot be corrected, in which case the honest conclusion is that the asset is being used outside its design envelope and the next decision should account for it.
The transfer of knowledge matters here too. A line that changes hands between engineers loses its history unless the handover was documented at the start; the handover checklist that captures the baselines at installation is also the document that makes an annual cost review possible, because it establishes what normal looked like before anything drifted.
Build the cost model before the quotation is compared
Most of the money in a placement line decision sits outside the machine price: the recurring year, the changeover events, the unplanned hours and the configuration items that were never quoted. If those are not in the model yet, the comparison is being made on the smallest of the three numbers.
Send us the board data, the monthly volume and the current schedule pattern, and we will work through the three cost clocks with you — capex scope, recurring cost per million placements, and the changeover arithmetic — against the platform you are considering. If you already have a JUKI SMT machine cost ROI model, send it as it stands; reviewing real numbers is faster than building a new sheet, and it usually finds one assumption worth changing.
We will also flag which of the three numbers your project depends on most. In most plants it is utilization, which is decided by maintenance and changeover discipline rather than by the machine, and that changes where the money should go in the first year.
These are the questions that come up when a cost case is being built for a capital request, usually a week before the committee meets and usually from someone who has been given a quotation and a spreadsheet template. They are answered in the order the numbers get built: what a realistic payback looks like, what belongs in the denominator, how to estimate costs you cannot yet measure, and how financing and used equipment change the picture. Where an answer depends on volume, mix or utilization, it says so and gives the measurement that settles it instead of a figure borrowed from another factory. No price is asserted here and no performance is promised; what follows is a method, and the numbers you put into it are yours.
It depends almost entirely on utilization and contribution per board, which is why any single figure quoted across factories is unreliable. A line running two shifts at high utilization on boards with meaningful contribution can pay back quickly; the same machine on single-shift, low-volume, low-contribution work can take several times longer.
Build the number rather than borrowing it: monthly contribution at your schedule, divided into the delivered and commissioned cost. Then test it at 70 and 90 percent utilization. If the payback only works at the higher figure, the project is a utilization project, and the plan should say how that utilization will be achieved rather than assuming it.
The whole line. A placement machine cannot produce a board without a printer, a conveyor, a reflow oven and some form of inspection, and the capital required to make the machine productive is the denominator that matters. Comparing mounter prices is a valid step for choosing between mounters, but it is not an ROI calculation.
Include handling, software, installation, training and the initial spares package as well. If the project requires floor space that could hold something else, note that opportunity cost even if the spreadsheet does not carry it. The number that decides the project is the capital required to ship boards, not the capital required to own a machine.
Ask for a consumption estimate per million placements and the assumptions behind it, then sanity-check it against the component mix. Nozzle consumption tracks the small packages running at high cycle counts; filter and belt consumption tracks hours rather than placements. A supplier who has installed similar lines should be able to give a range and explain what moves it.
Then treat the first quarter as calibration. Record actual consumption against actual placements, and replace the estimate with the measurement at the first quarterly review. The estimate only has to be close enough to keep the project viable; the measurement is what the second year's budget should use.
No. A faster machine improves the return only when the current machine is the constraint and the schedule can feed it. If the line is limited by changeovers, by material availability or by demand, extra placement rate adds capital without adding output, and the payback gets worse rather than better.
Test the constraint first: measure where the scheduled hours actually go. If unplanned stops and changeovers consume more time than the rate shortfall would recover, the faster machine is solving the wrong problem. Rate is a capability; utilization is what converts it into money.
As a capital item with a return, not as a maintenance expense. The parts worth holding are the ones whose failure stops the line and whose lead time exceeds a shift; buying them with the machine is cheaper than buying them in an emergency and converts a multi-week stop into a repair.
In the model, put the initial package in capex and put the expected annual replenishment in recurring cost. Then compare the package price with the cost of the downtime it prevents. For line-stopping parts the comparison is usually not close, which is why the conversation belongs in the purchase rather than in the maintenance budget.
Often in capital terms and sometimes not in total terms. A used machine lowers the denominator, which improves the payback ratio, but it can raise the recurring line through shorter supported life, scarcer spares and more unplanned stops. The two effects have to be modelled together rather than compared separately.
The decisive question is spares availability: if a board that stops the line is no longer stocked, the used machine carries a risk that no discount compensates for. If the platform is still supported and the condition is verified, the used route can be the better financial decision. A JUKI SMT machine cost ROI comparison for used equipment needs a condition assessment and a spares check before the price is discussed.
Cost per million placements, combining consumables, spares and maintenance labour, tracked quarterly against volume. It is the one figure that normalises everything else, and a rising trend is the earliest signal that the line is drifting away from the assumptions the project was approved on.
Track it next to utilization and changeover count, because those two explain most movements in it. When the cost per million rises, the question is whether consumption increased, utilization fell, or changeovers became more frequent — and each of those has a different owner and a different fix.