English
Views: 0 Author: Vinci Zhang Publish Time: 2026-09-28 Origin: Site
Almost nobody buys an SMT production line twice in five years. That is exactly the problem.
You get one shot at the decision, you are learning the market while you learn your own product, and every supplier you talk to is more experienced at this than you are. The mistakes in this article are not exotic. They are the same six or seven errors that show up again and again in factories on every continent.
Some of them cost money on day one. The worse ones cost money quietly for years, in changeover time, in yield, in spares you cannot get, in a line that never quite reaches the output the quotation promised.
This article walks through six mistakes in the order buyers usually make them, explains what each one costs, and gives you the specific action that prevents it. Section 8 pulls all of it into a checklist you can take into your next supplier meeting.
The pattern is remarkably consistent. A company decides to buy a line. Someone collects three quotations. The lowest one wins. Eighteen months later the production manager is explaining why the line still cannot hit its target output.
None of that is a story about dishonesty. It is a story about information asymmetry. The supplier has sold two hundred lines. The buyer is buying their first.
There is a number that makes the asymmetry concrete. A line that underperforms its target output by 15% for three years is not a 15% problem. If the line was supposed to generate 2,000,000 of gross margin a year across its five-year life, that shortfall is roughly 450,000 in lost margin. That is more than the entire price difference between a good line and a mediocre one. Yet buyers negotiate hard on purchase price and almost never negotiate on output verification.
There is a second reason, and it is more subtle. SMT lines get bought by committees. Engineering wants capability. Finance wants a low price. Production wants reliability. Procurement wants a clean comparison. Each of those goals is reasonable, and each one pulls the specification in a different direction.
Look closely at projects that go badly, and you usually find one of these:
The specification was written from a brochure instead of from the product.
Price was compared across different scopes.
Line balance was never calculated, so one machine became a permanent bottleneck.
Installation and training were assumed rather than specified.
Service and spares were treated as paperwork rather than as capacity.
Nobody visited a running reference line.
Every one of those is preventable. None of them requires special knowledge. They require the discipline to ask the right questions before signing, and the willingness to slow down for two weeks instead of arguing about the consequences for two years.
One practical starting point: study how complete SMT production lines are actually structured before you start comparing individual machines. Understanding what belongs in a line makes every later conversation shorter.
This is the root mistake, and most of the others grow out of it.
Buyers often start with the line — "I want a 60,000 CPH placer, a ten-zone oven, and an AOI" — before they have written down what they actually need to build in the next three years.
The result is a machine list that does not match the product. Sometimes it is over-specified, which wastes capital and adds maintenance complexity. More often it is wrong in a specific dimension nobody checked.
Before you talk specifications, write down these five things for your top five boards:
Component count per board. This sets the placement throughput you need.
Smallest component and finest pitch. This sets accuracy, vision, and stencil requirements.
Largest component and heaviest board. This sets handling, nozzle, and conveyor requirements.
Board size and panel configuration. This sets machine footprint and conveyor width.
Annual and peak monthly volume. This sets the machine class and the number of lines.
Then add the constraints people forget: does the product need nitrogen reflow? Does it need underfill, conformal coating, or selective soldering downstream? Does it need traceability down to the individual board?
A classic example: a company buys a placement machine sized for consumer-electronics volumes, then wins a power electronics contract with heavy boards, tall components, and thick copper. The machine can technically place the parts, but the conveyor and support system struggle, and the reflow profile is wrong for the thermal mass. The line was not wrong in general. It was wrong for that product.
There is a mirror-image version of this mistake. Some buyers over-specify out of caution and buy capability for products they may never build. A machine rated for 01005 components when the smallest part on any roadmap board is 0402 adds cost, adds maintenance sensitivity, and adds calibration burden for no return. Over-specification feels safe. It is really just a different way of not defining the product.
The most expensive version of this mistake happens when the product definition is still moving while the machines are on the water.
A change from a 4-layer to a 6-layer board, or a shift from 0402 to 0201 components, looks small on an engineering drawing. On the floor it can mean a different stencil thickness, a different reflow profile, a different placement accuracy class, and in the worst case a different machine. If there is any chance your product roadmap shifts in the next eighteen months, build that uncertainty into the specification instead of discovering it during commissioning.
Write a one-page product definition before you read any quotation. Share it with every supplier and ask each one to confirm the line they propose can build those boards at the required volume.
If the answer is a generic "yes, absolutely," ask a follow-up: "Walk me through how the line handles board number three." A supplier who has actually done the analysis will answer in specifics. A supplier who has not will repeat the specification sheet.
Comparing machines by price is natural, and it is dangerous, because the machine is not what you are buying. You are buying output. Output depends on the machine, the feeders, the programming, the material flow, and the operator.
Two placements machines at the same price are almost never equivalent. One may include 40 feeders and a basic vision system. The other may include 80 feeders, an automatic nozzle changer, and a much better camera. On the invoice they look similar. On the floor they are different lines.
The feeder count is the clearest example. Two machines can look identical in a brochure photo. One arrives with 40 mechanical feeders, the other with 80 electric feeders. Changeover time, setup flexibility, and the number of products you can keep loaded all depend on that difference, and none of it appears in the machine's headline price.
The way to see past this is to stop comparing machines and start comparing complete lines against your product mix. That is how a high-speed SMT PCB assembly line built around your product mix is normally presented — as a configured answer to a defined requirement, not as a box with a price on it.
Here is how price comparison usually fails in practice.
Supplier A quotes 1,450,000 for a line. Supplier B quotes 1,780,000 for what looks like the same line. The committee picks Supplier A and saves 330,000.
Then the project starts. The feeders for the second product need to be purchased: 40,000. The offline programming software is not included: 25,000. Commissioning is five days instead of two weeks, so the ramp takes a month longer than planned. Training covers one operator, and a second needs to be flown to the factory: another 30,000 with travel. The stencil cleaner was never in the scope.
Here is what the lower quotation was most likely missing:
Line item | Usually in the cheaper quote? | Why it matters later |
|---|---|---|
Spare feeder set | No, sold separately | Needed as soon as the product mix changes |
Offline programming seat | Often excluded | Without it, every program is written on the line |
Stencil cleaner | Priced as an option | Printing defects cost more than the cleaner does |
Buffer stations | First item cut | They absorb the small stoppages that stop everything |
Training days | One operator, one session | Knowledge leaves when the operator does |
Commissioning days | Shortened to fit budget | Process windows are never proven on real boards |
By the time the line is running, the savings have largely evaporated, and the buyer now manages two supplier relationships instead of one.
There is a second trap inside the same mistake: comparing prices across different currencies, delivery terms, and payment structures. A quote in a foreign currency with FOB terms and a 30% deposit looks cheaper than a quote in your local currency with delivered-and-installed terms and payments tied to milestones. They are not comparable. Normalize currency, Incoterm, and payment structure before you compare anything else.
Procurement teams are often measured on purchase price, not on total cost or line performance. That incentive quietly pushes projects toward under-specified lines, and nobody in the room is being unreasonable about it. They are optimizing the metric they were given.
The fix is procedural rather than personal. Ask procurement to compare three-year total cost instead of purchase price, including spares, training, utilities, software fees, and expected ramp losses. Change the metric and the decision changes with it.
Normalize every quotation against the same scope table before you compare a single number. Use the six cost blocks from the cost breakdown of an SMT production line as your structure: machines, inspection and software, peripherals, engineering, installation and training, commercial terms.
Then compare three-year totals, not purchase prices. And when you present the comparison internally, present capability first. If a line cannot build your third product, its price is irrelevant.
One more discipline that pays off: put the spare feeder set and the offline programming seat into the original order. They cost less there, the shipping is already paid, and you will not have to justify a second purchase order six months later.
Keep a written record of every question you ask and the answer you receive. When the quotations arrive three weeks later, that record is the fastest way to spot a supplier who agreed to something on a call and quietly dropped it from the document.
A line is only as fast as its slowest station. Buyers know this in theory. They still buy unbalanced lines, because machines are purchased one at a time and each one is optimized on its own.
Picture a typical case. The placement machine runs at 60,000 CPH. The stencil printer runs at 25 seconds per board. The reflow oven holds a fixed conveyor speed. The AOI inspects every board at 20 seconds.
If the product takes 45 seconds of placement time, the placer needs 45 seconds, the printer needs 25, the AOI needs 20, and the oven needs whatever the profile dictates. The line runs at 45 seconds per board, and the placer is the constraint. Buying a faster printer changes nothing at all.
Bottlenecks do not just cap output. They change how you operate.
The bottleneck station runs continuously, so it never gets maintenance windows. Unplanned stops become inevitable.
Everything upstream needs buffering, and everything downstream sits idle.
Product changeovers at the bottleneck take longer, because the queue has to clear first.
Any quality problem at the bottleneck multiplies through the shift.
Buyers also misread the constraint. A line that "cannot hit output" is often not short of speed. It is short of feeders, or short of material handling, or short of an operator to reload carts. Speed is just the easiest thing to blame and the easiest thing to buy.
Changeover time is the other constraint that hides behind a speed number. A line that runs 45 seconds per board in steady state can lose that entire advantage if every product change takes 90 minutes. Two changeovers a day at 90 minutes each is three hours of lost production — roughly 240 boards that never got made. When you compare lines, compare them across a realistic week of production, not across their best hour.
If you already run a line, measure before you buy. Log the stop reasons for two weeks: changeover, feeder reload, paste related, reflow related, inspection related, material shortages. Rank them by total minutes lost. The top item is your real constraint, and it is often not where you expected it to be.
If you are building a new line, ask every supplier to produce a balance sheet for your product mix. Then check whether any station exceeds the others by more than 20-25% of cycle time. If one does, ask how the supplier proposes to absorb it — buffering, a second station, or a change in the process. A supplier without an answer has not modeled your line.
Run a balance calculation before you buy, using your actual product mix. Take the cycle time at each station for each of your top five products, and find the slowest station in each case. Then design the line so that no single station is more than 20-25% slower than the others for your main products.
If the calculation is hard to do by hand, ask the supplier for it in writing. Any serious supplier will produce a line balancing sheet. If they cannot, that is a signal about how the line will be commissioned.
For high-mix operations, the layout decision matters as much as the balance. If your product mix is wide, inline or modular SMT line layout explains how modular cells absorb imbalance that a rigid inline line cannot.
Buyers spend months on machine selection and about twenty minutes on installation. That ratio is backwards.
Installation, training, and ramp-up are where projects actually succeed or fail. A perfectly specified line that is installed badly, commissioned quickly, and handed over to untrained operators will underperform a modest line that was properly set up.
The failures are predictable and boring.
Utilities are not ready. The machines arrive before the power, air, and exhaust are finished. The line sits in crates for three weeks.
Commissioning is compressed. The supplier gives five days because that is what the contract said. The line is technically running but the process windows are not proven across the product mix.
Only one operator is trained. Then that person goes on leave.
Nobody owns the process. The programs work, but nobody can explain why the profile is set the way it is. When a new product arrives, the team starts from scratch.
The ramp product is the hardest one. The first board run is the highest-margin, tightest-tolerance product the company makes. It does not go well, and confidence drops.
Treat installation and ramp-up as a project with its own schedule, not as a task that happens after delivery.
In practice, that means producing three documents before delivery. A utilities readiness checklist with dates and a named owner. A commissioning plan listing the products, the acceptance criteria, and the number of days. And a training plan that names the people, records the sessions, and includes a handover of the process documentation. Suppliers who produce those documents unprompted are usually the ones whose projects finish on schedule. Suppliers who produce them only after you ask are usually the ones who need chasing.
Write the utility readiness requirements into the contract with dates, and make them your responsibility with a named owner. Negotiate a commissioning scope that includes process window development on at least three of your real boards, not on a demo board. Train at least two operators and one process engineer, and record the sessions. Choose a forgiving product for the first two weeks of production and accept that you are learning on it.
Ask one specific question before signing: "What does day thirty look like?" A supplier with a commissioning plan will describe measurable milestones. A supplier without one will describe good intentions.
There is also a contractual detail that trips people up: what counts as acceptance. If the contract says the line is accepted on installation, you own it before you know whether it can hold a process window on your boards. If it says acceptance follows successful production of three of your products at the agreed cycle time and yield, the supplier shares your risk through the ramp. Ask for the second version. Suppliers who are confident about their equipment agree to it.
A rushed handover has a second cost that never appears in a project plan. It creates a knowledge gap.
The operators learn the buttons, not the process. When a new product arrives, nobody knows how to build the profile from first principles, so they copy the nearest existing program and adjust until the boards look acceptable. That habit is how a factory ends up with a process it cannot explain and a yield it cannot improve. Two extra weeks of commissioning is cheap compared with that outcome.
This mistake is invisible during the purchase and painfully visible afterwards.
A line that cannot be fixed quickly is not a line. It is a very expensive set of machines. The difference between a two-day repair and a twelve-day repair is usually decided by commercial terms that nobody read carefully.
How fast can a service engineer reach my factory? Is that response time contractual, or a sales claim?
Which spare parts will I hold locally, and what do they cost? Am I buying a starter kit or a real spares package?
Who pays for travel and accommodation on a warranty visit?
Is there a local partner or distributor who can service the equipment without waiting for a flight?
What is the expected lead time on a replacement control board or vision module?
Does the supplier still support this machine model, and for how long?
Pay attention to how the answers are given. "Usually within 48 hours" is a claim. "Within 48 hours for remote diagnosis, and within five working days on site, with travel included during the warranty period" is a commitment.
Run a simple calculation. Take the gross margin your line generates per day. Multiply by the number of days per year you expect it to be down for service. That number is the real value of a fast support agreement.
A line generating 8,000 of margin per day, down for twelve days instead of four, loses 64,000 in a year. Against that, a service agreement with a 5,000 annual premium that removes eight days of downtime is obviously worth it. Most buyers never run this math, and end up optimizing the machine price instead of the uptime.
Spare parts strategy deserves its own conversation. There are two schools, and both are reasonable. Some factories hold a deep local stock of common wear items — nozzles, belts, filters, sensors, feeder parts — and rely on the supplier for major assemblies. Others hold almost nothing and depend on air freight when something fails. The first approach ties up working capital and saves days of downtime. The second saves cash and costs you hours whenever something small breaks. Whichever you choose, choose it deliberately, and put the list in writing.
Response time is the number suppliers advertise. The support model is what decides whether that number is real.
Does the supplier have engineers in your region, or do they fly in from another continent? Do they have a local partner with trained technicians? Can they diagnose remotely with secure access to the machine? A 48-hour response promise backed by local engineers is a completely different commitment from the same words backed by an international flight schedule.
For lines that must run at high utilization, spares and support also shape how you plan maintenance. If you intend to add MES integration on an SMT production line later, ask early whether the equipment exposes the data you need. Retrofitting data collection onto older machines is possible, and much cheaper when it is planned from the start.
Make support a scored criterion in supplier selection, not a tiebreaker. Get spares pricing in writing before you sign. Confirm the spare parts list by name and part number. And confirm what happens to the warranty if you source consumables from a third party, because that clause surprises people.
Reference checks are the cheapest risk reduction available to a buyer, and they are the step most often skipped under schedule pressure.
A supplier's own brochure cannot tell you what it is like to work with them. Their existing customers can.
If you can visit the supplier's factory, go. Look at the production process, not the showroom.
Is the assembly area active and organized, or is it a display space?
Do they build their own machines or integrate third-party equipment? Both models are legitimate, but the answer changes your support expectations.
Are there machines in test running real processes, or only static units?
Can they show you the quality control process for incoming components and outgoing machines?
How big is the engineering team, and how many of them work on process rather than sales?
If travel is not possible, ask for a live video walkthrough. A supplier who cannot arrange that is telling you something.
Pay attention to what the factory says about its own process. Ask how they validate a machine before shipping, how long the test run lasts, and whether it runs a real product or a demo board. Ask what happens when a machine fails final inspection. Those answers tell you how likely you are to receive a machine that has genuinely been run, rather than one that was assembled and boxed.
Ask for three references, and ask for a mix: a customer with your product type, a customer in your region, and a customer who bought more than two years ago. The third one is the most valuable, because they have lived with the equipment long enough to be honest.
Then ask these questions, and listen to the pauses as much as the words.
Would you buy from them again? Why or why not?
How long did commissioning actually take compared with what was promised?
When something broke, how fast did they respond, and who paid?
Did the line reach the output you expected, and how long did that take?
What do you wish you had specified differently?
Question five usually produces the most useful answer of the whole conversation.
One caution on references. Suppliers usually offer their happiest customers, and those customers may be genuinely happy. That is fine. Make sure you also ask about problems, and give the reference an opening to talk about them. A question like "what went wrong during installation?" is easier to answer honestly than "were you satisfied?" — and it produces far more useful information.
Ask to see the quality documentation you will receive with the machines: inspection records, calibration certificates, and compliance documentation. For workmanship expectations on the assemblies you will produce, the widely used reference is IPC-A-610, Acceptability of Electronic Assemblies. Knowing which class you are building to should shape your inspection and training plan, and it is worth stating in the purchase specification.
Print this. Bring it to every supplier meeting. Fill it in for each supplier separately.
[ ] Top five boards documented with component count, smallest component, finest pitch, largest component, board size, and volume.
[ ] Peak monthly demand confirmed, with a 20-30% growth allowance.
[ ] Real placement requirement calculated using utilization and balance factors, not brochure speed.
[ ] Line balance sheet produced by the supplier and reviewed by your engineering team.
[ ] Changeover target agreed in minutes, for a real product change.
[ ] Every quotation mapped against the same six-block scope table.
[ ] Spare feeder set and offline programming seat included in the original order.
[ ] Three-year total cost compared, including spares, utilities, and software fees.
[ ] Contingency of 8-12% approved before project start.
[ ] Payment schedule linked to milestones, not just to shipping dates.
[ ] Utility readiness requirements documented with dates and a named owner.
[ ] Commissioning scope includes process window development on at least three real boards.
[ ] Minimum of two operators and one process engineer trained, with sessions recorded.
[ ] Ramp product selected and agreed as a low-risk board.
[ ] Acceptance criteria written down and measurable, so both sides know when the project is done.
[ ] Service response time contractual, with travel costs defined.
[ ] Local spare parts list confirmed by part number, with prices.
[ ] Three references checked, including one customer with over two years of ownership.
[ ] Factory visit or live video walkthrough completed.
[ ] Compliance documentation and compliance marking requirements confirmed for your market.
Two more checks are worth adding, because they catch different problems. Confirm the line's upgrade path in writing: if you might add inspection, handling automation, or a second line later, ask which components would be reused and which would be replaced. Then confirm which parts of the line your own team can service, because a line that needs the supplier for every adjustment stops whenever the supplier is busy.
If the numbers are the main open question, it helps to see how the industry splits the cost. The full cost breakdown of an SMT production line explains where the money goes and which items are safe to cut.
If the line will run thick copper, heavy boards, or high-power assemblies, the specification errors are different from consumer electronics. Start with SMT lines for power electronics and PCBA work before you finalize the machine list. And if your product mix changes often, review the trade-offs in semi-automatic vs fully automatic SMT lines before you decide how much automation to buy.
One final piece of advice. The single most common trigger for every mistake in this article is schedule pressure. Committees rush because a customer order is waiting. That pressure is real, but the line will outlast the order. Spending two extra weeks on the product definition and the scope table is almost always cheaper than spending two extra years working around a line that was specified too quickly.
When you move from the checklist to the shortlist, the same discipline applies to the vendor conversation: how to evaluate a turnkey SMT line supplier covers the scoring approach. And at some point you will face a question that comes before all of this — whether to change the line at all, which is the subject of upgrade or replace: choosing the right path.
Specifying machines before defining the product. Buyers choose a placement rate and an oven zone count from a brochure, then discover that their real boards need different handling, different accuracy, or different thermal profiles. Every other mistake in this article tends to follow from that one. If you fix only one thing, write a one-page product definition with your five main boards and their component data before you read a single quotation.
No. Compare scope first, then price within the same scope. A lower quote usually reflects a narrower scope, not a better deal. Ask every supplier to quote against the same line-by-line table, then compare three-year totals including spares, training, utilities, and software. If a line cannot build your third product at the volume you need, its price does not matter at all.
For a fully automatic line with automatic handling and inspection, plan for one to two operators per shift. For a full line with manual loading, plan two to three. For a semi-automatic line, three to five is realistic, and that number rises with product changeover frequency. These are planning ranges, not guarantees. Your product mix, board size, and inspection criteria will all move the number.
It is the cheapest risk reduction available. If you cannot travel, ask for a live video walkthrough of the production floor and the test area. You are looking for an active manufacturing environment, real machines running real processes, and a visible quality control step. A supplier who cannot arrange a walkthrough is giving you useful information about how they will behave after the order.
Ask whether they would buy again, how long commissioning actually took, how fast support responded when something broke, whether the line reached expected output, and what they wish they had specified differently. The last question is usually the most valuable. Ask for references with different profiles: a customer in your region, one building your type of product, and one who has owned the equipment for more than two years.
Ask for the upgrade path in writing before you sign. Confirm which components would be reused if you later add inspection, automatic handling, or a second line, and which would be replaced. Pay attention to conveyor heights, machine footprint, utility capacity, and floor space. A line with no documented upgrade path usually means replacing equipment that should have been reusable.
Want a second opinion on your line specification before you sign?
Send us your five main boards and the specification you are working from. We will review the balance, the scope, and the upgrade path, and tell you where the specification is likely to cost you later.
