English Views: 0 Author: Elsa Long Publish Time: 2026-11-13 Origin: Site
Three suppliers can quote the same machine model, the same board size and the same placement speed, and still deliver three very different lines. The difference is not in the specification sheet. It is in whether the machine arrives when promised, whether it runs when the shift starts, whether the part you need on a Friday night is on a shelf somewhere, and whether the person who answers the phone can diagnose a fault without being walked through the machine from the beginning. Evaluating those four things is what a JUKI mounter availability configuration service support review actually consists of, and none of them can be settled by comparing catalogues.
The mistake buyers make is to treat the three words in that review as one quality score. Availability becomes a feeling, configuration becomes a list of options with ticks next to them, and service support becomes the salesperson's mobile number. Separating them produces better questions and harder answers. The framework below splits availability into three separately measurable numbers, audits configuration against three documents instead of one, and turns service support into a four-step chain you can test before the order is placed. Machine references are drawn from the placement platforms this evaluation is written against, and every worked figure is labelled as an assumption so it can be replaced with your own data.
Availability sounds like one property and behaves like three. The first is whether the machine can be delivered when you need it, which is a supply question. The second is whether it runs when the shift starts, which is a reliability and maintenance question. The third is whether a failed part can be replaced in hours rather than weeks, which is a logistics question with a surprisingly large effect on the second. A supplier can be strong on one and weak on another, which is why a single availability score tells you almost nothing. A JUKI mounter availability configuration service support review that keeps the three separate will surface exactly that pattern.
Treating them separately changes what you ask and when you ask it. Delivery is settled before the order and confirmed at the order. Runtime is negotiated as a definition and then measured by you. Spares is a stocking decision that belongs in the order documents, because the part you need most is rarely the part the supplier keeps closest. Each has a different evidence type as well: a delivery claim is a commitment, a runtime claim is a statistic, and a spares claim is a location.
The table below is the whole framework in one place. The rest of this guide fills in how to test each row.
Availability | Question it answers | Where the evidence comes from | Decided by |
|---|---|---|---|
Delivery | Can the machine be on my floor in the month I need it? | Written confirmation with a date and a penalty, allocation status, FAT slot | Before and at order |
Runtime | What proportion of planned production time does it actually place boards? | Your own MES or stop log; MTBF and MTTR by subsystem from the vendor | Definition at order, measurement after |
Spares | How long is the line stopped waiting for a part? | Parts list with lead times and classes, depot locations, stock held locally | At order and at handover |
Delivery availability is measured in weeks from order to production-ready, not from order to shipment. Those are different numbers and the gap between them is where most project plans fail. A machine that ships in eight weeks and takes four more to install, program and prove is a twelve-week project, and any supplier quoting the first number while you plan against the second has given you a number that is technically true and practically useless.
The useful question is narrower: on what date can this specific configuration be on my floor with the options I need? Options change the answer. A standard machine may be in stock; the same machine with a wide-board option, an additional feeder cart and a specific nozzle set may need a build slot, and the build slot may be the constraint rather than the base machine. Ask which parts of the configuration are in stock, which are on the line and which are made to order, and ask for the date in writing against the full list.
Finally, ask what happens to the date if it slips. A date without a consequence is a forecast. A date with an agreed remedy — a revised delivery plan, a temporary loan unit, a discount, or a right to cancel — is a commitment. The difference shows up in how carefully the supplier checks the date before writing it down, which is information you get for free.
Runtime availability is the number everyone quotes and almost nobody defines. A machine that is powered on is not available; a machine that is powered on but waiting for a feeder, a program or an operator is idle. The standard way to make the number meaningful is to separate the reasons the line did not place boards: planned stops such as changeovers and preventive maintenance, unplanned stops such as faults and jams, and losses that belong to other stations. Availability, as a reliability measure, counts only the unplanned stops against planned production time.
Once the definition is fixed, the number becomes arithmetic on two quantities: mean time between failures and mean time to repair. Availability equals MTBF divided by MTBF plus MTTR. A machine with a 240-hour mean time between failures and a 4-hour mean time to repair has an availability of 240 divided by 244, or 98.4 percent. Change the repair time to 12 hours and the same machine drops to 95 percent. That sensitivity is why repair capability, including spares and support, matters more than the failure rate in most real evaluations.
Ask for both numbers by subsystem rather than one number for the machine, and ask what counts as a failure in the data you are given. Feeder faults, nozzle clogs, vision errors and conveyor stops have different owners and different fixes; an average across them hides exactly the information you need in order to decide where to hold spares.
Spares availability is the most concrete of the three and the easiest to verify. It is not a percentage; it is a list. For each part that can stop the line, you want to know whether it is held locally, regionally or only at the factory, and how many days each of those means in practice. A nozzle is a day. A head board is often a week. A conveyor motor or a vision module can be several weeks if it is not stocked anywhere closer than the factory.
The reason this belongs in the evaluation rather than in maintenance planning is that it changes the purchase. If a critical board is only available from the factory with a four-week lead time, the correct response is to buy one at order and keep it on the shelf, which is a capital decision made once. Discovered after the first failure, the same decision is made under pressure, at whatever price the situation allows, with the line stopped while you make it.
Ask for the parts list with lead time and price per line, then sort it by what happens to the line when that part fails. The top of the list is the shelf list. Everything below it is a support agreement question. A supplier that can produce this list quickly and accurately has a spares organization; one that cannot has a catalogue.
Delivery dates fail for boring reasons. A build slot moves because a larger order took it. A subassembly arrives late. A test fails at final inspection. Export documentation takes longer than the plan assumed. None of these are unusual, and none of them are visible in a quotation that says "lead time: 8 to 10 weeks". The purpose of asking harder questions is not to catch the supplier out; it is to find out which of these risks is live on your order while there is still time to plan around it.
There is also a structural reason to be careful. A machine order competes for capacity with every other order, and the buyer with a clear, documented schedule and a penalty clause tends to get the slot that a vaguer buyer loses. Being precise about dates is not adversarial. It is the behaviour that makes you predictable to plan around, which is exactly what a factory allocates capacity to.
Most projects carry at least three dates and treat them as one. The promise date is what the quotation or the sales conversation implies. The confirmation date is what appears on the order acknowledgement after engineering review. The shipment date is what the forwarder is told. Slippage usually happens between the first and the second, quietly, because the acknowledgement is read as a formality.
Read the acknowledgement against the promise line by line, including the configuration. If the acknowledgement lists a different option code, a substituted feeder type, or a different vision package, the date may be right and the machine may be wrong — which is worse, because it is discovered at installation. Any difference between quotation and acknowledgement should be resolved before the deposit, not after.
The shipment date then needs one more conversion: shipment plus transit plus customs plus installation equals the date you can produce. Keep all four in the project plan and treat only the last as the real one. When a supplier reports progress, ask which of the four dates has moved, because "on schedule" against a shipment date says nothing about the date your line starts.
A lead time is a range because it is a forecast, and a forecast is only as good as the assumptions behind it. Ask what the range assumes: a standard configuration, an existing build slot, cleared payment, and documentation the supplier prepares. Then ask which of those assumptions is currently true for your order. The answer usually narrows the range to a specific risk you can plan for.
Watch for the difference between a stock machine and a built machine. A platform that is held in stock locally can be delivered in weeks; the same platform with options that are fitted at the factory goes back into the build queue. The two are quoted with the same confidence and behave completely differently. What is actually orderable also varies by platform in a range: the LED placement platform in the current range and a general-purpose chip mounter will not share the same lead time, even from the same supplier, because they are not built on the same line.
One practical test: ask for the last three shipments of the same configuration and the date each one actually left. A supplier with the data will answer in an hour. A supplier without it will answer with a range, and the range is a guess.
Allocation is the word for capacity that has been reserved for you rather than promised to you. It is the difference between being in the queue and being in the plan. Ask directly whether a build slot has been allocated to this order, on what date it was allocated, and what would cause it to be reallocated. A confident, specific answer is a good sign. Vague reassurance is information too.
Ask also about the factory acceptance test slot, because that is frequently the real constraint. A machine can be built and still wait weeks for a test bay. If the acceptance test is part of your payment terms — and it should be — the test slot belongs in the project plan next to the shipment date.
Finally, ask what the supplier does when a date slips through no fault of yours. The answer separates suppliers who plan capacity from suppliers who sell it. A revised schedule with a named owner and a new date is usable; an apology is not. Write whichever answer you get into the order file, because in three months it becomes the reference point for a conversation you would rather have with evidence.
A quoted uptime figure is a negotiation position until it has a definition attached. Two suppliers can both claim 98 percent and mean completely different things: one excludes changeovers, the other excludes any stop under fifteen minutes, a third excludes stops caused by material. None of them are lying; they are answering different questions. Before you compare numbers, agree what counts as planned, what counts as unplanned, and what counts as somebody else's problem.
The reason to spend time on this is that availability is the number your capacity plan depends on. Every hour of unplanned stop is an hour of output that has to come from overtime, from a second line, or from a customer conversation. Getting the definition right at the order stage costs an email. Getting it wrong means discovering in month four that the number you planned against was measured under conditions your factory does not have.
Write the definition into the order documents in plain language. Planned production time is the shift length minus planned breaks and planned maintenance windows. Unplanned stop time is any interval in which the line was scheduled to run and did not place boards because of a machine, feeder, nozzle, vision or conveyor condition. Changeover, material starvation, program loading and operator absence are recorded separately, not hidden inside availability and not excluded from the report.
Then fix the two inputs. Mean time between failures is total runtime divided by the number of unplanned stop events. Mean time to repair is total unplanned stop time divided by the number of events. Both are computable from a stop log that records four fields per event: start time, end time, station and cause. Anything more elaborate is optional; those four fields are not.
Agree also how the vendor's own figures will be reported. If the supplier quotes MTBF and MTTR, ask whether they are per machine or per line, and whether feeder faults are included. A feeder failure is a machine stop from the line's point of view even though it is not a machine failure from the platform's, and the difference matters when you are deciding what to stock.
Most lines already generate the data and throw it away. The machine controller logs errors, the MES logs board counts, and the shift report logs downtime in a free-text field that nobody can aggregate. The fix is not a new system; it is a fixed vocabulary. Define ten to fifteen stop codes that cover the faults you actually see, print them on the shift sheet, and require a code rather than a sentence.
Two habits make the data usable. First, record the station, not just the line, because a stop attributed to the printer and a stop attributed to the mounter have different owners. Second, record the intervention, not just the cause: was it cleared by the operator, by maintenance, by a remote session, or by replacing a part? That field tells you which stops your support chain is actually absorbing, which is the link between runtime availability and service support.
The maintenance side of the same data is covered in the maintenance schedule that stops uptime from drifting, which maps checks to the failure modes that produce these codes. Running the two together means a rising stop count can be traced to a degraded condition instead of being treated as bad luck.
Once a month of data exists, the arithmetic is simple and the result is useful. As an illustration only, assume a line scheduled for 20 hours a day, 25 days a month, which is 500 scheduled hours, and assume the stop log shows 14 unplanned events totalling 11 hours. Availability is 489 divided by 500, or 97.8 percent, MTBF is 489 divided by 14, or about 35 hours, and MTTR is 11 divided by 14, or about 47 minutes.
That combination tells you something specific: the failure rate is high and the repair time is short. The response is maintenance and consumable discipline, not spares. Reverse the numbers — four events totalling 20 hours — and the diagnosis flips: the machine rarely fails but recovers slowly, and the response is spares and support, not more frequent checks. Same availability figure, opposite action.
Then convert to output, because that is the language management uses. Assume the line produces 1,200 boards a shift at a contribution of USD 4 per board. Losing 11 hours a month out of 500 is 2.2 percent of capacity, which is roughly 26 boards a shift, about USD 104 a shift, or roughly USD 2,600 a month on a 25-day calendar. The figures are assumptions, but the method is not: it turns an availability percentage into a monthly number that can be compared with the cost of the fix.
Availability comparison needs a starting point, and the first month is the only time you can capture it cheaply. Record the as-installed cycle time, the placement offset on one critical reference, the feeder fault count per thousand placements, and the vision recognition failures per thousand reads. Those four numbers, taken while the machine is new and the program is golden, are the reference for every later argument about whether the machine has degraded or the product has become harder.
Use the platform's own published figures as the outer boundary, not as the target. As one reference point in this class, the KE-3020VA and its published placement figures list an optimum rate of 20,900 CPH with 17,100 CPH under IPC9850 and 9,470 CPH for IC placement, laser recognition accuracy of plus or minus 0.05 mm, vision accuracy of plus or minus 0.03 mm, and up to 160 feeder inputs on 8 mm tape with electric double tape feeders. Those numbers describe the machine under defined conditions. Your baseline describes your line under yours, and the gap between them is the honest measure of what your product mix costs.
Spares is where availability stops being a statistic and becomes a purchase order. Every line has a small number of parts whose failure stops production and a much larger number whose failure degrades it. The shelf list is built from the first group, and it is usually shorter and cheaper than buyers expect, because most of what fails is either consumable or already held locally.
The mistake is to stock by price rather than by consequence. A cheap sensor that stops the conveyor is worth holding. An expensive board that produces an occasional warning is not, if the machine can run degraded until it arrives. Sorting by consequence rather than by cost is what turns a spares budget into an availability decision, and it is the sort that a supplier's parts list makes possible in an afternoon. It is also the block of a JUKI mounter availability configuration service support review that most often changes the purchase order rather than the opinion.
Start from the stop log and work backwards. Any part that has caused more than one unplanned stop in the last twelve months, and whose replacement takes more than four hours to obtain, belongs on the shelf. That rule is crude and it works, because it uses your own failure history rather than a generic recommendation list that assumes a different factory.
Add the parts whose failure is rare but whose recovery is long. A vision module, a head board or a conveyor drive motor may fail once in three years, but if the lead time is four weeks the line is down for four weeks, which no factory survives. These are the parts that justify holding stock even though the failure rate argues against it, and they are the reason the decision belongs at order stage, when it can be bundled into the machine purchase.
Exclude anything that is consumable, because consumables follow a different logic covered below. Exclude also anything that can be borrowed from an identical machine on the same site, unless both machines run the same product at the same time, in which case the argument for holding two returns immediately.
Holding stock costs money and running out costs more, so the decision needs both numbers. As an illustration only, assume a head board costs USD 3,000, assume holding cost including capital and obsolescence is 20 percent a year, which is USD 600 a year, and assume the line contributes USD 4 per board at 1,200 boards a shift. A four-week stockout therefore costs the contribution from roughly 20 shifts, which is about USD 96,000 — against USD 600 a year to hold the board.
The comparison is not close, and that is the point. For parts that stop the line, holding cost is almost never the deciding factor; the deciding factor is whether the part can be obtained at all within the production window. The arithmetic matters more for the middle group: parts that degrade output rather than stopping it, where the loss is measured in yield rather than hours.
Set a reorder point rather than a fixed stock level. Consumption per month multiplied by lead time in months, plus a buffer for the variability you have actually seen, gives a number that moves with usage instead of a number set once and forgotten. Review it quarterly against the stop log, because a part that has not been used in a year is either a candidate for removal or a signal that the failure it guards against has not happened yet.
Part class | Examples | Typical path to the line | Stock decision |
|---|---|---|---|
Wear consumable | Nozzles, filters, belts, lamps | Local stock, days | Reorder point based on monthly usage |
Line-stopping component | Feeder units, sensors, conveyor drives | Regional depot, days to two weeks | Hold one on site if lead time exceeds a shift |
Long-lead module | Head boards, vision modules, controllers | Factory, weeks | Buy one at order, keep in a controlled store |
Site-specific part | Tooling, fixtures, custom nozzles | Made to order | Duplicate the critical ones at order |
Nozzles, filters and belts are not spares in the availability sense; they are the cost of running the machine, and treating them as spares produces both stockouts and dead stock. The right measure is consumption per million placements, reviewed quarterly, with a reorder point derived from that rate and the delivery lead time rather than from a fixed quantity.
Consumption also tells you something about the process. A nozzle consumption rate that rises without a change in product mix usually means the cleaning interval or the vacuum level has drifted, not that the nozzles have become worse. Reading consumable usage as a process signal rather than a purchasing chore is one of the cheapest diagnostic habits a line can adopt, and it links directly back to runtime availability.
Keep the consumable stock visible on the floor, not in a locked store with a requisition form. The cost of a slightly higher stock level is trivial next to the cost of a stopped line waiting for someone with a key. The control that matters is the reorder point, not the lock.
Configuration is the part of the purchase that looks settled and is not. A quotation lists options; a contract references the quotation; the machine arrives with a configuration that is usually close to both and occasionally different in one line that matters. The differences are rarely dishonest. They come from substitutions made during build, from an option that turned out to be incompatible with another, or from a term that meant something different to the two sides.
The defence is dull and effective: three documents, one comparison. The quoted configuration, the contracted configuration and the as-built configuration should be the same list, in the same order, with the same units. When they are, the acceptance conversation is a checklist. When they are not, it becomes an argument about what was meant, conducted weeks after the people who knew have moved on.
The quotation is a proposal and should be read as one: it establishes intent and price, and it is frequently written quickly. The contract or order acknowledgement is the binding list, and it is the document that will be produced when a disagreement arises. The as-built record is what you create yourself, at installation, by counting and photographing what is actually on the floor. Only the third one describes the machine you will run.
Build the comparison as a single table with a row per configuration item and three columns. Do it before the deposit for the first two columns, and at installation for the third. Any row where the columns differ is either a resolved substitution — in which case write down who agreed it and when — or an open issue. This sounds like paperwork for its own sake until the first time a feeder count is two carts short, at which point it is the only document that settles the discussion in ten minutes.
Keep the table in the handover file next to the installation records, and update it whenever a change is agreed. Configuration drifts after installation as well: a feeder cart is added, a nozzle set is replaced with a different bore, an option is enabled. A living table is worth more than a perfect one that stops being maintained.
Some items change without anyone intending to mislead, and those are the ones worth checking by hand. The first is feeder count and width mix: a headline number of positions can be delivered with a different mix of 8 mm and wider bodies, and the effective capacity for your BOM depends on the mix, not the total. The second is the nozzle set, where a standard kit may cover the common bores and leave your smallest or largest part needing a purchase later.
The third is head configuration, which determines what the machine can place at what speed and is occasionally substituted within a family. The fourth is board handling: rail width range, long-board capability and the conveyor interface to neighbouring machines, all of which have options attached. The fifth is software and vision: recognition packages, traceability links and reporting modules that are licensed rather than built, and therefore easy to leave off an acknowledgement.
Check these five physically at installation and record what you find. Every other item on the list can be verified from documents; these five need a person with a torch and a parts list, and they are the five that determine whether the machine can build the board you bought it for.
A feeder count is a capacity statement only when it is expressed as positions of specific widths. Eight millimetre tape occupies one position; 24 mm tape occupies several, depending on the bank design; a wide body may block the position beside it. A machine quoted at 160 positions is not a machine that runs 160 different 8 mm parts alongside 20 wide parts without arithmetic.
Do that arithmetic against your BOM before the order. Convert every part to an 8 mm equivalent using the width ladder, add the parts that must be online at the same time, and compare the total with the usable positions. If the product mix needs two carts to cover one build, the changeover cost belongs in the business case, not in a footnote. The relationship between feeder and nozzle choices and what a configuration can actually produce is set out in how feeder and nozzle options change what a configuration is worth, and it is worth reading before the option list is signed rather than after.
Ask the supplier to state the usable position count for your specific mix, not the maximum. A supplier who does that calculation with you is demonstrating that they understand the machine; one who repeats the maximum is quoting a datasheet.
Board size is the option buyers most often regret. A machine sized for the current panel will be sized for the current panel for the life of the machine, and panels grow. The cost of the next size up is paid once; the cost of discovering the limit is a second machine, or a panel redesign, or an outsourced run for the one product that does not fit.
The decision is not simply "buy bigger". Larger platforms cost more, occupy more floor and may have a different feeder capacity or speed balance. What matters is knowing the class boundaries and where your roadmap sits relative to them. A 400 to 450 mm board class covers a large share of consumer and industrial work; a chip shooter built for the 400 to 450 mm board class is a different proposition from a wide-board platform, and the right answer depends on the largest panel you expect to run in five years, not the one you run today.
Write the answer down as a number: the maximum board dimension the machine will handle, and the maximum you expect to need. If the gap is small, the decision is close and price should decide. If the gap is large, the decision is already made and the only question is how to pay for it.
Service support is usually evaluated as a feeling: the salesperson was responsive, the references were positive, the website has a support page. None of that predicts what happens at 2 a.m. on a Saturday when the head throws an error nobody has seen. What predicts it is the structure of the response chain and whether each step has a defined owner, a defined time and a defined escalation.
The chain has four steps: first response, remote diagnosis, on-site attendance, and escalation to the factory. Each can be tested before you buy, at low cost, by sending a real question through the channel you would actually use. The results are more informative than any reference call, because they are your own measurement rather than someone else's impression. Support is the last block of a JUKI mounter availability configuration service support review and the only one you can test at zero cost before spending anything.
First response is the time between your message and a human acknowledging it with a name and a next action. It is not the time to a fix. Ask what the committed first response time is, in and out of working hours, and whether it applies to the channel you will actually use — a sales line that answers in minutes is not the same as a technical line that answers in a day.
Remote diagnosis is where most modern support is won or lost. A supplier who can connect to the machine, read the logs and identify the fault without travel resolves a large share of issues in hours. Ask what remote access requires, whether it is enabled by default, who authorizes a session, and what data leaves the plant. Ask also how a session is documented, because a support history is the strongest evidence of a working chain.
The practical test is a real one: before the order, send a technical question that requires looking something up — a parameter, a compatibility, a spare part number. Measure how long the answer takes and whether it is right. That single data point tells you more about step one and two than a page of service promises.
On-site attendance is expensive and slow, and the number that matters is not the promised arrival but the promised arrival path: who decides to travel, from where, with what visa or site-access requirement, and carrying what parts. A supplier with an engineer in the region and a stocked depot behaves very differently from one who travels from the factory, even when both quote the same attendance time.
Escalation is the step buyers forget to ask about. Every fault that cannot be fixed on site has to go somewhere: an application engineer, a design team, a factory specialist. Ask who that person is, how they are reached, and what the committed response is when a case is escalated. Ask what happens on the second escalation, because that is where slow cases actually sit.
A useful clause is a named escalation contact with a time commitment, written into the agreement. It changes behaviour, because a named engineer with a clock on them answers differently from a general support inbox.
Run the test while the purchase is live, not after. Send two questions through the normal channel at different times of day, one simple and one that requires a document. Record the response time, the accuracy, and whether the answer came with a reference number or a case ID. A support organization that issues case numbers has a system; one that answers from a personal mailbox does not.
Ask for two references with the same platform and a comparable product mix, and ask them a specific question rather than a general one: how long did the last unplanned stop take to resolve, and what was the cause? General satisfaction is easy to give; a number requires memory, and memory is what you want.
The same test applies to the ramp-up phase, where support quality matters most. The handover sequence in the handover checklist that decides how much support you will need puts installation, programming and operator training into three separate acceptances; a supplier who supports that structure is a supplier whose support you can schedule instead of summon.
Support clauses fail when they are written as adjectives. "Prompt response" means nothing at 2 a.m. Replace it with a table of commitments: first response time in working hours, first response outside working hours, remote session availability, on-site attendance time by region, escalation response time, and the hours of coverage per week. Then add the remedy when a commitment is missed, because a commitment without a remedy is a target.
Coverage hours deserve their own line. If your plant runs three shifts or works weekends, a support agreement written around a five-day week has a predictable gap, and the gap is where the expensive stops happen. Ask explicitly what the out-of-hours path is and whether it is staffed or on call.
Step | What you are testing | Pass evidence | Red flag |
|---|---|---|---|
First response | Speed of human acknowledgement | Named reply with a case number and next action | Answer only from a sales address |
Remote diagnosis | Ability to read the machine without travel | Documented session and identified cause | Remote access treated as an extra |
On-site attendance | Realistic arrival path and parts carried | Engineer based in region, depot nearby | Travel from factory quoted as local |
Escalation | Where unresolved cases actually go | Named engineer, committed time, second level defined | Generic inbox with no owner |
An evaluation becomes useful when it produces a number you can defend and a document you can enforce. The number is a weighted score across the four blocks — delivery, runtime, spares and support — with the weights chosen to reflect your plant rather than a generic template. The document is the set of clauses that carry the score into the order, because an evaluation that ends in a meeting has no effect on what arrives on the floor.
The starting point is the model itself. Availability, configuration and support are all properties of a specific platform in a specific configuration, so the evaluation only makes sense once the shortlist exists; the selection criteria this evaluation has to sit on top of decide which machines you are scoring. Scoring three machines you have not shortlisted properly wastes the most expensive part of the exercise, which is the supplier time you consume getting answers.
Weights should reflect the cost of being wrong in each block, not the enthusiasm of the evaluator. A plant that already holds a spare line weights delivery lower than a plant building its first line. A plant running three shifts weights support hours higher than one running a single shift. The weights are a statement about your operation, and they should be written down before the scores are known, or they will be reverse-engineered afterwards.
As an illustration only, assume a first line being added to an existing factory: delivery 20 percent, runtime 30 percent, spares 20 percent and support 30 percent. Score each block out of five against the evidence you collected, not against impressions. Two suppliers at 3.8 and 4.1 are effectively tied and the decision should go to price or to the reference call. A gap of more than half a point on a weighted score usually traces to one block, and that block is where the negotiation should focus.
Do not carry the score to two decimal places, and do not treat it as objective. Its value is that it makes the reasoning explicit and forces every claim to point at evidence. A score built from two reference calls and a written support table is worth more than a page of adjectives, even though it is still a judgement.
Four clauses carry most of the weight. The first fixes the configuration as an annex in the contract, in the same format as your three-column table, so that substitutions require written agreement. The second fixes the delivery date with a remedy, and names the acceptance test slot as part of the schedule. The third fixes the spares list with lead times and states which items ship with the machine. The fourth fixes the support table with response times, coverage hours and a named escalation contact.
Write each clause with a number and a document behind it. "Machine to be delivered in accordance with the agreed configuration" is weak; "configuration per Annex A, table 1, including 160 usable 8 mm feeder positions on electric double tape feeders" is enforceable. The difference is not legal drafting skill. It is whether the number was established during the evaluation, which is why the evaluation and the contract should be written by the same people.
Add one sentence that most orders miss: what happens if a clause is not met. A revised plan with a date, a discount, a loan unit, or a right to reject are all acceptable remedies. The point is that the remedy exists before the problem does, when both sides are still reasonable.
The evaluation does not end at acceptance; it is verified in the first quarter. Measure the same four baseline numbers captured at installation — cycle time, placement offset, feeder faults and vision failures — at day 30 and day 90, and compare them with the stop log. A drift inside tolerance is a trend to watch; a drift outside it is a support case with evidence attached, which is a very different conversation from a complaint.
Use day 30 for the mechanical and program review described in the handover documents, and day 90 for the commercial one: did the support chain respond as agreed, did the spares list prove right, and did delivery promises hold for the second order if there is one. Record the answers next to the original scores. A supplier whose score is confirmed after 90 days of real production has earned a different kind of conversation than one whose score has to be quietly revised.
Escalate early and in writing. The first ninety days are when a supplier is most motivated to fix things and when your own team still remembers the details. A case raised in week three with a log attached is resolved in days; the same case raised in month eight is a negotiation. The habit of writing it down while it is small is the cheapest form of support insurance available.
Score your shortlist against evidence, not catalogues
Most availability and support failures are visible before the order, in the answers a supplier gives when asked for a date, a parts list and a response time. If those answers are not yet in writing, the evaluation is still open, and this is the cheapest moment to close it.
Send us the board data, the panel sizes you expect and the shift pattern you run, and we will work through the three availability numbers, the configuration comparison and the support commitments for the platform you are considering. Bring the JUKI mounter availability configuration service support questions you are already asking; a scored shortlist is faster to review than a blank page.
If the shortlist is still open, that is the better starting point: we can begin with the board and the volume, narrow the platform class, and build the evaluation around a machine that actually fits instead of scoring three that do not. Either way, the output is a comparison you can attach to an order rather than a set of impressions that disappears with the sales meeting.
These are the questions that come up while a shortlist is being scored, usually from a process engineer who has three quotations on the desk and no way to tell which set of promises is worth more. They are answered in the order the decisions are made: how to compare claims, what to stock, how much support is actually worth paying for, and how the answers change for a used machine. Where an answer depends on your product mix or your shift pattern, it says so and gives the measurement that settles it instead of a number borrowed from another factory. Nothing here is a performance promise. What follows can tell you what evidence to ask for and how to read the answer; it cannot tell you what your line will achieve.
Force both claims into the same definition before comparing them. Ask each supplier what counts as planned production time, what counts as an unplanned stop, and whether stops caused by material, program or operator are included. Most quoted percentages differ because the definitions differ, not because the machines do.
Then ask for the two inputs rather than the result: mean time between failures and mean time to repair, ideally by subsystem. Two suppliers quoting the same availability can have very different combinations, and the combination tells you whether your risk is failure frequency or recovery time. That distinction decides whether you spend money on maintenance or on spares.
No, and the two can move in opposite directions. A platform tuned for maximum placement rate may run at higher acceleration and shorter pickup windows, which makes it more sensitive to nozzle condition and feeder presentation. That does not make it a worse machine; it makes it a machine whose availability depends more heavily on consumable discipline.
Treat the quoted rate as a capability boundary and availability as a separate property. The rate tells you what the machine can do under defined conditions; availability tells you how much of the scheduled time it actually gets to do it. Both belong in the evaluation, and they should be scored in different blocks.
Start from your own stop log, then add the long-lead items that would stop the line for weeks. In practice that means one of each line-stopping component whose lead time exceeds a shift, one of each long-lead module regardless of failure rate, and a consumable stock sized by monthly usage rather than by a fixed quantity.
The list is usually shorter than buyers expect and more expensive per item than they expect. For a single line, a practical starting set is one feeder of the most-used width, one nozzle of each bore in the program, the sensors that gate board transfer, and one of the two or three modules whose lead time you cannot accept. Review it after 90 days against actual consumption.
Less for day-to-day faults and more for the events you cannot handle internally: a controller failure, a software corruption, a fault that repeats after two repairs. Those are rare and expensive, which is exactly the profile that a support agreement is for. On a mature line the value shifts from response speed to escalation depth.
Adjust the weighting rather than dropping the block. A mature line with in-house maintenance capability can accept slower first response if the escalation path is strong, and the support clause should be written to reflect that: longer first-response time in exchange for a named escalation engineer and a committed escalation response.
Yes, and it is the cheapest part of the evaluation. Send two real technical questions through the channel you would actually use, at different times of day, and record how long each took and whether the answer was correct. Ask whether responses carry a case number. Ask what remote access requires and who authorizes a session.
Then ask two references a specific question rather than a general one. "How long did your last unplanned stop take to resolve and what caused it" produces a number and a story; "are you happy with the service" produces a politeness. The difference between the two answers is the difference between evidence and impression.
The three availability numbers still apply, but the evidence is harder to get and the weights shift. Delivery is usually short, runtime is unknown without a condition assessment, and spares availability becomes the dominant risk because a used platform may be out of production and its boards no longer stocked.
Configuration comparison matters more, not less: a used machine arrives with whatever the previous owner ordered, and options cannot be added later at catalogue prices. Capture the full parameter set and the as-built configuration before anything is changed. A JUKI mounter availability configuration service support review written for new equipment needs a condition-assessment annex for a used one, not a different framework.
Ask for the parts list with lead times, sorted by what happens to the line when each part fails. It is one document, it takes a competent supplier an hour, and it reveals almost everything: whether they have a spares organization, whether they understand your production risk, and whether the availability number they quoted is supported by anything.
A supplier who produces the list quickly and discusses which items you should hold has done this before. A supplier who cannot produce it, or who answers with a generic recommended-spares package, has told you how the first serious failure will go. Both answers are useful, and only one of them is expensive to discover later.
