Every PCB manufacturer has the same private statistic: the share of incoming jobs that cannot be quoted without an email asking for something. The missing item is rarely exotic. It is a drill file, a board outline, a layer whose function nobody can identify, or a set of coordinates that place the parts a few millimetres off the board. None of these are hard problems. They are simply invisible from inside the EDA tool, because the tool renders your design from its own database — not from the files you exported. This guide is deliberately not a menu-by-menu tutorial for KiCad, Altium or Eagle, because those menus move between releases and the screenshots rot within a year. Instead it describes what the output has to contain and how to prove it does, which is knowledge that survives every version bump.

What does a factory actually do with your Gerber files?

Understanding the downstream use explains most of the rules. A fabricator's CAM department imports your package, identifies each layer by its function, aligns everything to a common origin, and rebuilds a manufacturing dataset: photoplot data for imaging, drill programs for the mechanical and laser drills, routing paths for the profile, and a stencil aperture set if paste is included. Before any of that, they compare your geometry against process limits — trace and space, annular ring, drill-to-copper clearance, aspect ratio, mask sliver widths.

The critical point is that CAM works from your files, not your intent. If a layer is ambiguous, an engineer makes an interpretation and moves on, or stops and emails you. The first outcome costs you a wrong board; the second costs you two days. Both are avoidable at export time. At BELI Technologies, every PCB fabrication order includes a DFM review before production starts, and quotes are returned within 24 hours — but a DFM review can only comment on what is present in the package. It cannot infer a drill file that was never exported.

RS-274X versus the old format: why apertures must travel inside the file

Gerber began as a plotter control language. In the original RS-274D form, the file contained only coordinates and D-codes — instructions to select "tool 12" and flash it at a location. What tool 12 physically was lived in a separate aperture list, often a plain text table, sometimes never sent at all. When that list went missing, a fabricator had no way to know whether a pad was a 60 mil round or a 2 mm square, and a job effectively arrived as a drawing with no scale.

RS-274X, sometimes called extended Gerber, solved this by embedding aperture definitions inside each file, along with the coordinate format and unit declarations. This is the format any modern EDA tool produces by default and the only one you should send. Gerber X2 goes further, adding standardised attributes that declare what each file is — top copper, inner layer 3, solder mask bottom — plus drill span and net names, so the layer-identification guesswork disappears entirely. If your tool offers X2 output, enable it: it removes an entire category of misinterpretation at zero cost to you.

A practical test settles the question. Open a copper file in a text editor. If you see percent-delimited aperture definition commands near the top declaring shapes and sizes, you have RS-274X. If you see nothing but coordinates and bare D-codes, you are shipping the old format and the aperture list is somewhere else — or nowhere.

What files must a complete Gerber package contain?

Extensions vary by tool, and that variation is normal. What matters is that every function below is present and identifiable. The extensions in the middle column are the conventions you will most often see, not a requirement.

Why is the drill file the piece that most often goes wrong?

Drill data is separate from Gerber by history: it uses Excellon format, a different language with its own conventions, and it is exported through a different dialog in most tools. That separation is why it gets forgotten, and why its settings get missed even when it is present.

Three details deserve checking every time. First, plated and non-plated holes should be in distinct files or explicitly tagged, because plating a mounting hole changes its finished diameter and can create an unintended connection. Second, the tool table must be present, whether embedded in the file header or supplied as a drill report — this is the only place your hole diameters are stated in a machine-readable way. Third, confirm whether the diameters you exported are finished hole sizes or drilled sizes; plating reduces a hole by a measurable amount, and a tight-tolerance fit is where that difference stops being academic.

Process limits matter here too. BELI's fabrication runs to a minimum hole diameter of 4 mil with a maximum aspect ratio of 11:1, so on a thick board the smallest drillable via is set by thickness as much as by the drill itself. If your design pushes those numbers, say so in the fab note rather than letting CAM discover it — the full set of limits is published on the manufacturing capability page.

Where is the board outline, and how does the factory know?

Gerber has no concept of "this is the edge of the board". The outline is just geometry on a layer, and the fabricator identifies it by convention or by an attribute. Three rules make it unambiguous.

Draw the profile as a single closed contour with no gaps and no double-drawn segments. A 0.01 mm gap invisible on screen is a break in the contour, and a CAM engineer either has to close it manually or ask you which side is inside. Put nothing else on that layer — no dimension arrows, no title block, no courtyard outlines — because everything on the outline layer is a candidate for being routed. And state the line width convention: most fabricators treat the outline as the centreline of the routing path, so a thick outline drawn as a decorative border invites a board that is off-size by half a router bit.

Internal features follow the same logic. Cutouts and slots must be closed contours on the outline layer, or defined as slots in the drill file, and it is worth saying in the fab note which method you used. Rounded corners, castellations and edge plating all need to be called out in words as well as drawn, because they change process steps rather than just geometry.

Panelization: state your intent before the factory chooses for you

Small boards are not built one at a time. They run in arrays, and the array design affects both fabrication yield and assembly feasibility. You have three options and should pick one explicitly: supply a fully panelized dataset, supply a single board and specify the panel parameters you want, or supply a single board and ask the factory to propose the most economical panel.

If you choose the first, the package needs to show the panel outline, the individual board outlines, rails, tooling holes and global fiducials. If you choose the second, specify the separation method — V-score for straight-edged boards, tab-route with perforations for boards with irregular profiles or overhanging connectors — plus the array count and rail width. If you choose the third, say so in writing; a request to "panelize as you see fit" is a valid instruction, an absence of any instruction is not.

One constraint links panelization to assembly directly. BELI's SMT lines handle boards from 45 x 45 mm up to 680 x 500 mm, so a board below that minimum needs to reach assembly in an array rather than as a single piece. Fiducials matter for the same reason: SMT machines use them for optical alignment, and a panel needs global fiducials on the rails in addition to local fiducials near fine-pitch parts. With placement precision of ±0.03 mm and BGA ball pitch down to 0.4 mm on our PCB assembly lines, alignment accuracy is limited by the reference marks in your artwork, not by the machine.

Units, coordinate format and precision: the silent corruption

Gerber and Excellon both encode coordinates as integers, with a declared format telling the reader where the decimal point sits. A 2.4 format in inches means two integer digits and four decimals. Older files sometimes use leading or trailing zero suppression, dropping zeros to save bytes on 1970s media. Get that declaration wrong on import and every coordinate scales by a factor of ten — a mistake that produces a plausible-looking plot rather than an obvious error, which is precisely what makes it dangerous.

Modern practice removes the risk. Export in metric with a 4.6 format, use absolute coordinates, and disable zero suppression. Keep the same units across the Gerber set, the drill file and the pick-and-place file: a package where the copper is metric and the drill file is imperial is legal, importable, and an invitation to a units error. Precision also has a floor worth knowing — exporting at 2.3 inch resolution quantises everything to a thousandth of an inch, which is coarse relative to the 3.0/3.0 mil inner-layer line width and spacing BELI can fabricate.

Why do pick-and-place coordinates end up offset or rotated?

This is the classic assembly-side trap, and it has two independent causes.

The first is origin mismatch. Your centroid file states X and Y relative to some origin, and your EDA tool may use the drawing sheet origin, a user-defined origin, or the board origin depending on how the project was set up. If the Gerbers reference one origin and the centroid file another, every part is offset by a constant. The fix is to set the origin to a defined physical feature — the lower-left corner of the board outline is the safest convention — export everything from that same origin, and state it in the file header or the README. If the design is panelized, also state whether coordinates refer to the single board or the panel.

The second is rotation reference. Rotation in a centroid file is measured from the footprint's own zero-degree orientation as defined in your library, and library authors do not always follow the pin-1 conventions in IPC-7351. A tantalum capacitor footprint drawn with the anode to the right in one library and upward in another will export different rotation values for visually identical placements. The consequence lands on polarised parts — diodes, electrolytics, connectors, and any IC where pin 1 orientation is not obvious from the package outline.

Two habits eliminate almost all of it. Send an assembly drawing or a top-and-bottom 3D render showing pin-1 and polarity markers, so the machine program can be checked against a picture instead of a number. And state your bottom-side convention: some tools give bottom-side coordinates as viewed from the top through the board, others as viewed from below, and mirroring the wrong way puts every bottom part in the wrong place. The BOM, centroid file and Gerbers should all be exported from one design revision, with reference designators that match across all three — our guide to the files required for a PCBA quotation covers the BOM structure in detail.

What do ODB++ and IPC-2581 do better than Gerber?

Gerber describes each layer in isolation. Nothing in a plain Gerber set states how many layers there are, what order they sit in, which are planes, or what the board is supposed to connect. Every one of those facts is reconstructed at the factory from filenames, conventions and your notes — which is where interpretation errors are born.

ODB++ and IPC-2581 replace the loose collection with a single structured database. One file carries the layer stack in order with each layer's function declared, drill spans, the netlist, the profile, and optionally component data including placement and BOM references. IPC-2581 has the additional merit of being an open, vendor-neutral standard rather than a format controlled by one EDA vendor. In practice the benefit is measured in the questions that never get asked: no layer-order query, no plane-polarity query, no "is this file the outline or a mechanical drawing" query.

None of this makes Gerber obsolete. It remains the universal exchange format, every fabricator on earth reads it, and RS-274X with X2 attributes closes much of the gap. BELI quotes from either Gerber or ODB++. The practical recommendation is simple: if your tool exports ODB++ or IPC-2581 cleanly, send it alongside the Gerber set. The redundancy costs you nothing and gives the CAM engineer a cross-check.

How do you self-check a Gerber package in ten minutes?

The single most valuable habit in this entire article: open the exact zip file you are about to send in an independent viewer — a free standalone Gerber viewer, or a fabricator's online preview — and inspect it as if you had never seen the design. Your EDA tool renders its internal database, so it will show you a correct board even when the export is broken. A viewer shows only what the files contain, which is what the factory will see.

That last row deserves its own emphasis. Exporting into a folder that already contains yesterday's output is how a superseded inner layer or an obsolete drill file ends up in a package that otherwise looks perfect. Delete and recreate the output folder as part of the export routine.

What does a small dimensional error actually cost?

The abstract case for checking your files is easy to nod at and easy to skip. The concrete one is more persuasive. A US smart-sensor customer came to us after boards had already shipped, with a problem that only appeared at final assembly: technicians could not insert components without excessive force, risking pad lift, lead deformation and cracked joints. The DFM review traced it to the Gerber package itself — the drill diameters had been specified at 0.8 mm, without allowance for the insertion tolerance the connectors required. We enlarged the holes to 1.2 mm, changed the connector to pin headers for a reliable tool-free mating, and turned the fix around in 48 hours with zero rework. The full account is in our Gerber hole issue case study.

Note where the error lived. Nothing was wrong with the fabrication; the board matched the files exactly. The gap was between what the files said and what the designer assumed they said — 0.4 mm of it. That gap is what an independent viewer check and a hole-diameter cross-reference against the mating part's datasheet are designed to catch, and it is why generic design standards such as IPC-2221 publish lead-to-hole clearance guidance rather than leaving it to judgement.

What to send with the Gerbers, and what a DFM review catches afterwards

Gerber describes geometry and nothing else. Everything about materials and process lives outside it, so a short fab note — a PDF drawing or even a plain text file — should accompany every package. Cover the layer count and stack order, base material and Tg, finished board thickness, copper weight on outer and inner layers, surface finish, solder mask and silkscreen colours, any controlled-impedance requirements with target values and tolerance, and the quantity and delivery expectation. BELI fabricates up to 42 layers, in materials including FR-4, High-Tg FR-4, halogen-free laminates, Rogers, Arlon, Taconic and Nelco, with surface finishes including ENIG, HASL, lead-free HASL, immersion silver and immersion tin, and holds impedance control to ±10% — but only if the requirement is written down, since none of it is inferable from copper artwork.

For an assembly quote, add the BOM with manufacturer part numbers, the centroid file, and any assembly drawing that clarifies orientation. If you are early in a hardware programme, it is worth building this package discipline during your first prototype rather than at the pilot build, because the same dataset carries forward through EVT, DVT and PVT stages and gets re-exported at every revision. If the distinction between a bare board order and an assembled one is still fuzzy, our explainer on PCB versus PCBA sets out what each stage requires.

No export checklist catches everything, which is why the file check and the manufacturing review are complementary rather than redundant. Your viewer check catches what is missing or misdeclared; a fabricator's DFM review catches what is present but unmanufacturable — annular rings too thin after drill tolerance, clearances below process limits, mask slivers that will not survive development, aspect ratios beyond what plating chemistry can fill.

BELI Technologies includes a DFM review with every fabrication order, quotes within 24 hours, and takes Gerber or ODB++ for quoting, with no MOQ on prototypes. Quick-turn bare boards run as fast as 24 hours for 1-2 layer designs and typically 3-7 working days for multilayer, and prototype assembly takes 48 hours once boards and parts are on site. If you are not certain your export is complete, send us the package you have — the review will tell you exactly what is missing before it costs you a build.

Related reading: What Files Are Required for PCBA Quotation · How We Fixed a Gerber Hole Issue for a Sensor Client · Prototype PCBA in China: A Practical Guide

Engineers usually discover trade terms the hard way: a shipment of assembled boards sits at an airport for four days because nobody named an importer of record, or an invoice quoted "ex works Shenzhen" turns out to be well short of what the boards actually cost to get onto a bench in Boston, Munich or Tel Aviv. None of this is difficult, but it is unforgiving of assumptions. This guide covers the trade-terms vocabulary, the documents customs expects, the difference between courier and forwarder channels, and how sample shipments differ from production volume — written for hardware and procurement people who buy contract manufacturing, not for freight professionals.

One disclaimer up front, and it is not boilerplate: this article contains no duty rates, no tariff classifications, and no thresholds. Those change by country, by product category, and by the month — a number published today can be wrong by the time you read it. Every rate question in this article resolves the same way: check your own country's current customs rules and confirm with a licensed customs broker in the destination market. What we can give you is the structure of the decision, which does not change.

Why is landed cost the only number worth comparing?

A quote for assembled boards is not a price. It is one line in a cost stack, and the size of the remaining lines depends entirely on the trade term attached to it. Landed cost is the full amount it takes to put working boards in your building, and it is the only figure on which two suppliers — or two manufacturing regions — can be compared honestly.

The structure looks like this, and it is worth writing out for your own build because the composition surprises people:

Goods value + export handling and origin charges + main carriage (air or sea) + insurance + destination terminal or courier handling + import duty + import VAT / GST / sales tax as applicable + customs brokerage fee + inland delivery = landed cost.

Two things make this stack behave differently from what a spreadsheet assumes. First, several of those lines are charged per shipment rather than per unit, so they dominate a 20-board prototype run and nearly vanish on a 5,000-board production order — the same effect our PCB prototype versus mass production cost breakdown describes for NRE. Second, import taxes are usually calculated on a customs value that includes freight and insurance in many jurisdictions, so shipping fast does not only cost more freight, it can also enlarge the base on which tax is assessed. Our guide to what PCBA actually costs in China covers the goods-value side of the stack; this article is about everything that happens after the boards leave the factory door.

What do Incoterms actually decide?

Incoterms are a set of standard three-letter trade terms published by the International Chamber of Commerce. They are not law, they are not payment terms, and they say nothing about when title passes. They answer four operational questions, and only those four: who arranges and pays for carriage, who handles export clearance in China, who handles import clearance and pays duties and taxes in the destination country, and at what physical point risk of loss or damage moves from seller to buyer.

That last one deserves emphasis because it is the one buyers misread. Under several common terms, risk transfers long before the goods arrive. If a term hands risk to you when the boards are given to the first carrier in Shenzhen, then a pallet damaged in transit is your loss even though the seller paid the freight bill. Cost and risk are separate questions in the Incoterms framework, and the terms deliberately split them.

Always write the term with a named place, because the term alone is incomplete. "FCA Shenzhen" and "DAP Frankfurt" are usable instructions; "FCA" on its own is not. And always state which Incoterms revision you are using — the 2020 revision is current — so there is no argument about which version's rules apply.

Which Incoterm should you use for a PCBA shipment from China?

Read the table as a spectrum: obligations shift steadily from buyer to seller as you move down. For most overseas buyers of electronics, the practical working range is FCA at the light end and DAP at the heavy end. FCA gives you control of the freight and the visibility that comes with it, while leaving Chinese export formalities where they belong — with the party physically located in China. DAP hands the whole outbound journey to the seller and keeps import clearance, and therefore control of your own customs record, with you.

Why is EXW usually the wrong default for a cross-border buyer?

EXW appears on a great many Chinese quotations because it is the simplest thing for a seller to price: the goods are ready, come and get them. As a costing baseline it is genuinely useful, since an EXW figure strips out every logistics variable and lets you compare goods value against goods value.

As a shipping term, it creates a structural mismatch. Under EXW the buyer is responsible for export clearance from China — but a foreign buyer with no Chinese entity generally cannot file a Chinese export declaration in its own name. In practice the seller or its forwarder ends up doing it anyway, informally, which means the paperwork does not match the agreed term. That is fine right up until something goes wrong, at which point nobody's obligations are where the contract says they are. Risk is the second problem: under EXW it passes to you while the cartons are still sitting on the factory's floor, before loading.

The usual fix is one letter of difference. Ask for FCA at a named place instead. You keep control of carrier selection and freight cost, the seller handles export formalities as the party actually able to, and risk transfers at a defined handover rather than at a pallet on a warehouse floor.

What does DDP really mean — and when is it worth it?

DDP is the term buyers ask for when they want a single number and no customs involvement. It is legitimate and widely used, and for small shipments through courier networks it can be genuinely convenient. It also carries consequences worth understanding before you request it.

Under DDP the seller is responsible for import clearance and for paying duties and import taxes in your country. That normally requires the seller, or an agent acting on the seller's behalf, to be the declarant. Three things follow. First, the import tax may not be recoverable by you in the way it would be if your own company were the importer — in VAT and GST jurisdictions, the entity that pays import VAT is generally the entity that can reclaim it, and if that entity is a foreign supplier's agent, your reclaim path may be broken. Second, you lose the audit trail: you do not hold the entry documents, and if the classification used was wrong, the exposure can still find its way back to you. Third, the seller has to price uncertainty into the quote, because it is absorbing a duty and tax bill it cannot control.

The practical guidance is to use DDP for low-value samples where convenience genuinely outweighs control, and DAP for anything where you want to reclaim import tax, control classification, or build a clean import history. Whichever you choose, confirm with your accountant and your broker how import tax recovery works in your jurisdiction under each term — this is precisely the kind of question that has a different answer in the US, the EU, the UK and Israel.

What documents does customs need before your boards clear?

Clearance delays are rarely caused by duty. They are caused by paperwork that is incomplete, inconsistent, or too vague to classify. The list below is a generic checklist of what import entries commonly require for electronics; the exact set is determined by your destination country, your product, and your broker's requirements.

On the compliance row, be careful about what you are claiming. Marks such as CE and FCC attach to your finished product and are declared by the company selling it, not by the factory that assembles it — a distinction we spell out in who does what for CE, FCC and UL compliance on China-built PCBA. Bare boards and unfinished assemblies shipped to you for further integration usually sit in a different position from finished apparatus placed on the market, and your broker will want to know which one is in the box.

If you improve one thing about your import paperwork, improve the goods description. Customs classification is what determines duty treatment, and classification is done from the description. "PCBA" means nothing to a customs officer. "Populated printed circuit board assembly for industrial sensor module, not a finished apparatus, for further integration by the importer" tells the officer and your broker what they need.

Consistency matters just as much. The invoice, the packing list, the waybill and the purchase order should agree on quantity, value and description. Discrepancies are the single most common trigger for a document query, and a query costs days, not hours. Agree the invoice wording with your supplier once, then reuse it for every shipment of that item so the entries look consistent over time.

What happens if the declared value is too low?

Sooner or later, someone will suggest under-declaring — writing a lower value on the invoice, or marking a commercial shipment as a gift or a no-charge sample. Some suppliers offer it unprompted, as a favour.

Decline it, and get the reason on the record. The declaration is legally the importer's, which means the exposure lands on you, not on the factory that typed the number. Penalties for undervaluation typically far exceed the duty avoided, and repeat undervaluation is a route to a compliance audit of your entire import history. There are business consequences too. Cargo insurance pays out against declared value, so an under-declared shipment that disappears is only partly recoverable. Your customs record follows your company, and a bad one produces more physical inspections and slower clearance on every future shipment. And your internal numbers stop meaning anything, because the value you paid and the value on the entry no longer agree.

The correct approach is undramatic: declare the transaction value actually paid or payable, describe the goods accurately, and let your broker apply the right classification. If duty exposure is genuinely material to your business case, the legitimate levers are classification accuracy, origin rules, and any duty relief or drawback programme your country operates — all of which your broker can advise on, and none of which requires a false number.

Express courier or freight forwarder: which channel for which shipment?

BELI Technologies ships worldwide from Shenzhen by express couriers including DHL, FedEx and UPS, or by air and sea freight for larger volumes. Express delivery to North America, Europe and Israel typically takes 3 to 7 days depending on destination and customs — note that the customs half of that range is a function of your paperwork and your broker, not of the factory. Boards leaving our PCB assembly lines on the same day can clear in a day at one destination and sit for three at another purely because of how the entry was prepared.

One planning note that catches everyone at least once: assembled boards are light and bulky once packed in anti-static trays and foam. Courier pricing frequently uses volumetric weight rather than actual weight, so packaging efficiency has a direct and sometimes surprising effect on freight cost. Ask for carton dimensions and gross weight before you compare shipping quotes.

Do samples clear customs differently from production volume?

They are treated differently in practice, though not always in the way people hope. Many customs regimes have simplified procedures or de minimis handling for low-value consignments, and many recognise a category for samples — but the conditions, values and paperwork vary widely by country and change over time, so treat any rule of thumb you hear as something to verify with your broker rather than as a fact.

What is reliably true is that the failure modes differ. Sample shipments go wrong through carelessness: an invoice marked "no commercial value" that customs refuses to accept, a personal name as consignee, a missing description, a battery nobody declared. Production shipments go wrong through unpreparedness: an importer registration that was never obtained, a classification decided under time pressure on the day of arrival, a broker appointed after the goods landed.

The useful discipline is to use your prototype and pilot shipments as rehearsals for production. Register the importing entity, appoint the broker, agree the invoice wording, and settle the classification while the stakes are a handful of boards — so that when your DVT or PVT build ships, the customs path is already proven. It costs nothing extra to do the paperwork properly on a small shipment, and it removes an entire class of launch-week emergency.

Who is your importer of record, and what do the other terms on your paperwork mean?

The importer of record is the legal entity responsible for the declaration, for paying duties and taxes, and for the accuracy of everything on the entry. Naming it correctly is the single most consequential logistics decision you make, and it is frequently made by accident.

If you have a legal entity in the destination country, that entity is normally the importer of record — the cleanest arrangement, because it keeps the entry, the tax recovery and the audit trail in one place. If you do not, your options are a customer or distributor acting as importer, a third-party importer-of-record service, or a DDP arrangement in which the seller's agent handles it. All three work, and all three have trade-offs around tax recovery, liability and record ownership.

The mistake to avoid is discovering the question on arrival day. Decide before the goods ship, name the entity correctly on the waybill and invoice, and make sure whoever is named actually knows they are named and has the registrations and authorisations to act. A pallet whose consignee cannot be reached accrues storage charges from the day it lands.

The rest of the vocabulary on your entry is worth knowing well enough to read a broker's email without guessing. A customs broker is a licensed agent who files entries on the importer's behalf and advises on classification and valuation. Tariff classification is the code assigned to goods that determines duty treatment; it is derived from the goods description and, for anything non-obvious, is worth confirming with your broker rather than guessing. Customs value is the value on which duty is assessed, usually based on the transaction value, with freight and insurance included or excluded depending on the country's valuation basis. Rules of origin are the criteria determining where goods are considered to originate, and they govern any preferential duty claim. Consignee and notify party identify who receives the goods and who the carrier contacts on arrival; getting these wrong is a common cause of storage charges. Volumetric weight is a billing weight derived from package dimensions, used when goods are bulky relative to their mass, which assembled boards usually are.

What should you settle with your supplier before the first shipment?

Most import trouble is a communication failure that happened weeks earlier, at quotation time. The following should be agreed in writing before a purchase order is issued, and none of it is unreasonable to ask of any competent supplier.

• The Incoterm and named place, with the revision year. Written as a full term on the quotation, not implied.
• What is and is not in the quoted number. Export packing, pallets, origin handling — spell out which of these are included so you can compare quotes on the same basis.
• The exact invoice description and unit values to be used. Agree the wording once and freeze it.
• The consignee and importer of record details, including whatever registration your country requires of an importing entity.
• Carton dimensions, gross and net weight, and carton count, provided before shipment so you can validate freight quotes and prepare the entry in advance.
• Whether batteries, magnets or any restricted item is in the box. Ask explicitly; it is easy for a battery on a box-build BOM to become a transport problem nobody flagged.
• Who buys cargo insurance, and against what declared value. Only CIP and CIF oblige the seller to insure; under every other term, if you did not buy insurance, there is none.

Two of those items connect back to how you structured the build itself. If you are running consignment rather than turnkey, components you buy and ship into China are their own import event with their own paperwork — worth planning alongside the outbound shipment, as our comparison of turnkey versus consignment PCB assembly discusses. And supplier responsiveness on logistics questions is a reasonable proxy for responsiveness generally; the logistics questions in our list of questions to ask before outsourcing PCBA exist for that reason.

Ship from Shenzhen with the paperwork already agreed

BELI Technologies is a one-stop EMS manufacturer in Shenzhen, building PCB fabrication, assembly, component sourcing, wire harness and functional test work for customers across the United States, the European Union, the United Kingdom, Israel, the wider Middle East and Asia. Shipments leave Shenzhen by express courier — DHL, FedEx or UPS — or by air and sea freight for larger volumes. Import clearance at the far end stays where it belongs: with you and your customs broker, the only parties able to decide tariff classification, importer-of-record registration and duty treatment for your product in your market. Settle that before the first shipment moves, fix the goods description and the trade term with its named place at quotation time, and the customs half of the journey stops being the unpredictable half. If you are planning a build, send us your files and tell us where the boards are going.

Related reading: How Much Does PCBA Cost in China · Turnkey vs Consignment PCB Assembly · How to Choose an EMS Partner in China

Aluminum PCBs — also called metal-core PCBs, MCPCBs, or insulated metal substrate (IMS) boards — are the least understood product in most engineers' catalogue of options. They look like ordinary boards, they quote like ordinary boards, and then they behave nothing like ordinary boards the first time someone tries to route a two-layer design or drill a mounting hole through the middle of one. This guide covers when the material is the right answer, why the construction forces design choices that FR-4 never does, how to think about the thermal path without pretending you can calculate it from a datasheet number, and what a fabricator needs from you to quote it properly.

What is actually inside an aluminum PCB?

An aluminum PCB is a three-layer sandwich, and every design constraint that follows comes from that structure. At the bottom sits the base plate: a sheet of aluminum alloy, most often a 5052- or 6061-class alloy, typically somewhere in the range of 1.0 to 3.0 mm thick, which provides both the thermal path and the mechanical rigidity of the board. Bonded on top of it is a thermally conductive dielectric — a filled polymer prepreg, usually tens of microns thick — whose entire job is to be electrically insulating and thermally transparent at the same time. On top of that sits the copper foil that carries your circuit.

That is the whole stack. There is no glass-weave core, no inner layers, and no path for a plated hole to pass from the top copper down to anything else, because the only thing below the dielectric is a large conductive slab. Once you internalise that picture, most of the "why can't I do X on an aluminum board" questions answer themselves.

Aluminum PCB vs FR-4: which does your board actually need?

The comparison below reflects general industry behaviour of the two material families, not the specification of any one supplier. Treat it as a screening tool, then confirm the specifics of any construction with the fabricator quoting it.

The decision rule that survives contact with real projects: if your heat problem is a small number of high-dissipation parts spread across an area you can afford to make out of metal, aluminum is efficient. If your heat problem is one hot BGA in the middle of a dense digital board, aluminum solves nothing you could not solve with heavy copper, thermal vias, and a heatsink — and it costs you the layer count you need.

Why are almost all aluminum PCBs single-layer?

Because the metal base has no electrical role and cannot be perforated by plated holes without shorting to it. On an FR-4 board, a via is a plated barrel through insulating material; on a metal-core board, the equivalent barrel would land in aluminum. Any hole that must pass through the base plate needs a physical clearance around it, and once you accept that, you have lost the mechanism by which multilayer boards are normally built.

Double-sided and multilayer metal-core constructions do exist. They are built by bonding a conventional multilayer sub-stack onto the aluminum plate, or by using pre-formed insulation around isolated through-holes — approaches which add process steps, add cost, and constrain hole placement and stackup in ways that vary considerably between fabricators. Treat them as an engineered exception you discuss and confirm before layout, not as a checkbox you tick on a quote form.

The practical consequence for architecture is worth stating plainly: split the product. Put the power stage and the LEDs on the aluminum board, put the microcontroller, sensors, and communications on an FR-4 board, and connect them with a header or a cable assembly. That partition costs you one connector and buys you a routable control board plus a metal-core power board, each on its correct material.

How does heat actually move through a metal-core board?

Do not start with thermal conductivity numbers. Start with the chain of thermal resistances between the semiconductor junction and the air, because that chain is what determines the junction temperature you actually care about, and because the dielectric is only one link in it.

The path runs: junction to package case (fixed by the part you selected), case to copper pad (solder joint quality and pad area), copper pad laterally outward (copper weight and pad geometry), through the dielectric (this is the link the aluminum board improves), spread through the aluminum base, across the interface between the base and whatever it is bolted to, and finally into the air. Every link is in series. Improving one link only helps as much as its share of the total.

The useful intuition for the dielectric link is that its thermal resistance scales as thickness divided by the product of conductivity and area — thinner is better, more conductive is better, and wider is better. That last term is why pad and copper-pour geometry matters so much: a hot part sitting on a small pad is squeezing its entire heat flow through a small patch of dielectric, and no amount of base-plate aluminum fixes a bottleneck that happens upstream of the base plate. It is also why the biggest single-step improvement on many aluminum designs is simply enlarging the thermal pad and the copper around it.

The second intuition is that the interface at the bottom of the board frequently dominates. An aluminum board resting loosely against a chassis with an air gap has thrown away most of what you paid for. Specify the mounting: screw torque and pattern, thermal interface material, flatness of the mating surface. If you want numbers rather than intuition, run a thermal simulation with real boundary conditions, then verify on powered hardware — an afternoon of measurements on a real board tells you more than a week of arguing about W/m·K figures. Plan that verification into the same build stage as functional and environmental testing, and agree in writing who takes the measurements and against what pass criteria.

What has to be specified on an aluminum PCB drawing?

The parameters below are the ones that get missed and cause re-quotes or scrapped builds. Values are process-dependent and vary by fabricator, so the right column tells you what to decide and confirm rather than asserting a universal capability.

Two of those rows deserve emphasis because they are where real money is lost. Hole and edge clearances are fabricator-specific and are exactly the class of detail that a DFM review catches before it becomes scrap — we have written up how a hole-size problem in a customer's Gerbers was caught and fixed in 48 hours, and the same review discipline applies with more force on metal-core work because the failure mode is a short to a grounded plate rather than a loose fit. And isolation voltage is the parameter customers most often leave blank, then discover during safety testing that the laminate they were quoted does not support the clearance their standard demands.

Aluminum board or FR-4 with thermal vias?

This is the honest comparison, because heavy-copper FR-4 with a dense thermal via array under the hot part is a genuine competitor and is often the better answer.

FR-4 with thermal vias wins when you need multilayer routing, when the heat is concentrated under a few packages rather than distributed, when the board also carries signal integrity requirements, and when you already have a heatsink in the mechanical design that the vias can dump into. It loses when the number of vias needed becomes absurd, when the vias interfere with routing you cannot give up, or when solder wicking into open vias under a thermal pad starts costing you assembly yield.

Aluminum wins when heat is spread across many devices — an LED array is the canonical case, since each emitter is modest but there are dozens of them — when the board can serve as its own heat spreader and mechanical member, and when the circuit is simple enough to route on one layer. It loses on anything routing-dense. A third option worth remembering is a hybrid architecture: an FR-4 control board and an aluminum power board, which is how most commercial LED luminaires and many power supplies are actually built.

What changes when an aluminum board goes through SMT assembly?

Assembly is where teams get surprised, because the same property that makes the board good at cooling makes it awkward to solder. The metal base is a large thermal mass that pulls heat out of the joint during reflow, so the reflow profile has to be developed for that specific board — longer preheat and soak to bring the whole mass up, careful ramp control so the top-side components are not cooked while the base catches up. A profile borrowed from an FR-4 build of the same size will produce cold joints or overheated parts, or both on different areas of the same board.

Beyond profiling, expect these differences: the board is heavier and stiffer, so conveyor and support tooling settings change; large white-masked LED boards are cosmetically critical, so handling and glove discipline matter more than usual; V-cut depaneling of a metal-scored panel needs the right cutter and a controlled stress direction so the LEDs near the break line are not damaged; and since components sit on one side only, there is no double-sided reflow to worry about but also no room to hide anything on the bottom.

Ask your assembler how they profile metal-core boards and what inspection they run afterwards. BELI's SMT and through-hole assembly covers boards from 0.4 to 5 mm thick with AOI and X-ray inspection followed by ICT or functional test, and holds first-pass yield at or above 99.5% — but the point for you is to make profile development an explicit line item in the conversation rather than an assumption.

When should you step up to copper base or ceramic?

Aluminum is the volume default, not the ceiling. A copper base plate conducts substantially better and is chosen where the power density justifies both the material cost and the extra weight — typically in high-current power modules rather than lighting. Ceramic substrates such as alumina and aluminium nitride go further still, offering high thermal conductivity together with excellent high-voltage isolation and a coefficient of thermal expansion much closer to bare die, which is why they dominate in power semiconductor modules and high-reliability RF power stages. They are also the most expensive, the most brittle, and the most constrained in size.

The progression to keep in your head is: FR-4 with vias, then aluminum, then copper base, then ceramic — each step buying thermal performance and, on the ceramic step, isolation, at rising cost and falling design freedom. BELI's PCB fabrication material range spans FR-4 and high-Tg FR-4, halogen-free laminates, polyimide, aluminum, ceramic, and Rogers and other high-frequency substrates, so the material decision can be made on engineering grounds rather than on which substrate one supplier happens to stock. Copper-base constructions sit outside that standard list — treat them as a separate conversation with any supplier rather than assuming they are stocked. Full process specifications are listed on our manufacturing capability page; the layer counts and thicknesses published there describe conventional laminate constructions, and metal-core parameters should be confirmed per project.

What drives the price of an aluminum PCB quote?

No numbers here, but the structure of the cost is predictable, and knowing it tells you where to push.

The amortisation point catches teams comparing a five-board prototype quote against a production price they were shown a year earlier. Setup, tooling, and reflow profile development are one-time costs on a metal-core build just as they are on FR-4, so a prototype price per board says almost nothing about what the same design costs at a thousand units. Ask for a price break table across the quantities you actually plan to build, and ask what portion of the first-article price is non-recurring — that single question usually settles the argument about whether a quote is competitive.

What should you send to get an aluminum PCB quoted?

The file package is the same as for any board — Gerber or ODB++, plus BOM and pick-and-place data if you want it assembled, as detailed in the files required for a PCBA quotation. What must be added for metal-core work is a mechanical drawing or fabrication note carrying: base alloy and thickness; dielectric thickness and required thermal conductivity, or the thermal requirement it has to satisfy; isolation voltage and test regime; copper weight; solder mask colour and any shade requirement; surface finish; which edges are V-cut and which are routed; hole schedule with mounting holes identified; deburr and edge-treatment requirements; and the expected mounting arrangement, because a fabricator who knows the board bolts to a chassis will flag clearance problems you did not see.

If you are not sure of some of those, say so rather than guessing a value. A specification written as a requirement — "must withstand X volts, must keep junction temperature under Y at Z watts" — lets a fabricator propose a laminate and a stackup. A specification written as a guessed material grade only tells them what you assumed.

Getting a metal-core board quoted and built

BELI Technologies fabricates aluminum and ceramic substrates alongside FR-4, high-Tg FR-4, polyimide, and Rogers high-frequency laminates, with DFM analysis included on every order. Our SMT lines take boards from 45 × 45 mm up to 680 × 500 mm and 0.4 to 5 mm thick, so a lighting engine or power board can move from bare board through component sourcing to a tested assembly without changing suppliers. If you have an aluminum board in design, send us your Gerbers and the thermal requirement: we will come back with DFM feedback on the hole and edge clearances, questions about anything the fabrication drawing leaves open — dielectric, isolation voltage, edge treatment — and a quote. If the design is still open, tell us what the board has to dissipate and where it mounts — that is usually a more productive first conversation than a material choice already fixed.

Related reading: PCB vs PCBA: What's the Difference · What Files Are Required for PCBA Quotation · PCB Prototype vs Mass Production Cost

Both processes exist on the same value-add menu at most electronics manufacturing services suppliers, including ours, and they get chosen for the wrong reasons more often than almost any other decision in a PCBA build. Teams specify conformal coating because a customer questionnaire asked about "environmental protection", or specify potting because a competitor's board looked impressively solid when they cut one open. Neither is a reason. This article works through what each process actually protects against, what it costs you in mass, heat and repairability, and how to decide — with the design and documentation detail your assembler will ask for once you do.

What does conformal coating actually protect against?

Conformal coating is a thin polymer film — typically tens of microns thick — applied over a populated board so that it conforms to the topography of components, leads and solder joints. Its job is to raise the surface insulation resistance of the assembly and keep contaminants away from the conductors underneath.

What it genuinely defends against: condensing humidity and the leakage currents that follow, airborne dust and fibres that bridge fine-pitch pads, salt fog in coastal and marine settings, light chemical and solvent splash, and the classic slow killer — electrochemical migration between adjacent conductors under bias in a damp environment. On a board with 0.35 mm lead pitch or a 0.4 mm ball-pitch BGA, the distance between two nets carrying a potential difference is small enough that a film of condensed moisture with a little ionic residue in it is a resistor you did not design.

What it does not do, and what buyers routinely assume it does: conformal coating is not an ingress-protection rating. IP67 or IP68 is a property of an enclosure and its seals, not of a 50-micron film on a PCB. Coating will not survive prolonged immersion, will not stop a connector filling with water, and will not hold pressure. It also does not protect anything it cannot reach — the underside of a BGA, the inside of a connector housing, the mating faces of a press-fit terminal. If your risk is submersion or high-pressure washdown, the answer is enclosure design, not a coating line item.

What is potting, and how is it different from encapsulation?

Potting means placing the assembly in a housing, shell or mould and filling the void with a curable compound — epoxy, polyurethane or silicone — until the electronics are entombed in a solid mass. The housing usually stays as part of the finished product. Encapsulation is the closely related process where the compound itself forms the outer body and the mould is removed, and glob-top is the same idea applied to a single die or a small cluster of parts rather than a whole board.

The distinction that matters commercially is that potting is a bulk process and coating is a surface process. Coating adds grams; potting adds tens to hundreds of grams and changes the product's centre of gravity, shipping class and sometimes its thermal design. Coating can be selectively applied and locally removed; potting is, for practical purposes, permanent. And potting delivers something coating cannot: mechanical support. A potted assembly resists shock and vibration because every component body is mechanically coupled to the housing rather than cantilevered off its solder joints. For a board that will spend its life on an engine mount, a compressor skid, or a handheld tool, that is often the real requirement, and moisture protection is a side benefit.

Potting is also the standard answer to two non-environmental problems: tamper resistance, because you cannot probe or desolder what you cannot see, and high-voltage insulation, because a filled gap has a far higher dielectric strength than the same gap full of air. Neither of those is a reason to coat.

Conformal coating vs potting: the decision table

Read the last row twice. Potting is the more aggressive protection, but it is also the process more capable of destroying an assembly that would otherwise have been fine. A hard epoxy poured around a tall electrolytic capacitor and then thermally cycled to minus 40 degrees will try to tear that capacitor off the board, and it frequently succeeds. That is why the compound's glass transition temperature, cure shrinkage and modulus matter more than its datasheet moisture-resistance figure.

Which coating material types exist, and where does each belong?

The material families below are general industry practice, not a menu of what any one factory stocks — settle the specific product and its qualification data with your supplier before you release a drawing. Cured film thickness is not a universal number either — it varies by material family and by the acceptance criteria you and your supplier agree in writing, so put the required range on the drawing instead of assuming a house default.

Potting compounds map onto three of the same families. Epoxies give the hardest, most chemically resistant and most tamper-proof result and the highest internal stress. Polyurethanes sit in the middle: tougher than silicone, less brittle than epoxy, and generally the compromise choice for outdoor modules. Silicones are the softest and the kindest to solder joints under thermal cycling, which is why they dominate anything that must survive wide temperature swings, though they are also the easiest for a determined person to dig out.

When do you need neither?

This section exists because the most common mistake is specifying protection reflexively. If your product lives in a sealed enclosure in a climate-controlled room, coating buys you very little and costs you rework flexibility on every future field failure. Consider skipping both when all of the following hold: the operating environment is indoor and non-condensing; the enclosure already provides adequate dust protection; there is no salt, solvent or chemical exposure; the board is not subject to sustained vibration; and the assembly may realistically be repaired rather than replaced.

There is a cheaper intervention that resolves a surprising share of the moisture concerns coating is asked to solve: cleanliness. Ionic residue left on a board — from flux, from handling, from a poorly controlled process — is what turns condensation into leakage current. A properly cleaned, well-inspected assembly in a benign environment does not need a film. Conversely, coating over ionic contamination traps it against the copper and can make corrosion worse, not better, because the residue now sits in a humid microclimate it cannot escape. If you are considering coating primarily because you are worried about residues, fix the cleaning process first and then decide whether the film is still necessary.

The other honest answer is "not yet". Coating a prototype run rarely earns its keep, because you will be probing, cutting and reworking those boards. It is normal — and cheaper — to build engineering samples bare and introduce coating at the design-validation stage, once the layout is stable. Our walkthrough of EVT, DVT and PVT hardware build stages covers where each process change should land in that sequence.

Why is masking the biggest line item in a coating quote?

People assume conformal coating is priced on material. It is not. The film on a mid-size board costs very little; what you are paying for is the labour of deciding what must stay uncoated and then physically keeping the coating off it.

Every connector contact, test point, programming header, relay contact, potentiometer, switch actuator, LED lens, antenna, optical sensor window, MEMS microphone port, pressure sensor port, mounting hole with a grounding function and heatsink interface is a keep-out area. Each one has to be masked with tape, boots, dots or dispensed peelable latex before coating, then de-masked afterwards, then inspected to confirm the coating did not creep under the mask and that the masked area is genuinely clean. On a densely connectorised board that is minutes of skilled manual work per unit, twice, and it does not get much faster at volume unless the design cooperates.

Which means masking cost is a design decision, not a factory decision. Things that make it cheap: grouping all keep-out features into a defined region of the board rather than scattering them; choosing connectors that accept an off-the-shelf boot; leaving clearance around every keep-out so a mask has somewhere to seal; and specifying selective coating with a clear boundary on the drawing rather than "coat everything except the connectors". Things that make it expensive: a test point in the middle of a coated area, a keep-out feature within a millimetre of a component that must be coated, and a drawing that leaves the boundary to the operator's judgement. Ask your assembler for feedback on this before you release the design. A DFM review comes with every order we take, and the coating keep-out drawing is worth putting in front of the PCB assembly team at the same time, while moving a test point still costs nothing.

How much repairability are you giving up?

Reworkability is the trade nobody quantifies until the first field return arrives. Frame it as a straightforward expected-cost question: what fraction of units will need intervention after coating or potting, and what does each intervention cost under each option?

With coating, rework is possible but never free. The coating over the target area must be removed by solvent, abrasion or thermal parting; the repair is performed; then the area is cleaned and re-coated, and the re-coated area rarely matches the original film in appearance or in thickness uniformity. That whole loop needs a documented procedure if the product is in a regulated category — an undocumented touch-up on a medical or automotive assembly is a traceability problem, not just a cosmetic one.

With potting, there is no rework loop. Cutting a potted module open destroys it and usually damages the board underneath. The correct planning assumption is that a potted unit which fails in service is a scrapped unit, which pushes the cost of quality entirely upstream. That is not an argument against potting; it is an argument for stronger electrical and functional test coverage before the pour. In-circuit test coverage, functional test against the product specification, and a burn-in run all become more valuable the moment repair stops being possible, because the only remaining defence is not shipping a bad unit in the first place. Our functional testing and ICT service targets at least 95 percent node coverage with fixtures built in two to three working days, and that coverage figure is worth negotiating harder when the product is going to be potted.

What does coating or potting do to thermal design?

Coating has a modest thermal effect: a thin polymer film slightly impedes convective and radiative heat transfer from a hot component surface. On most designs this is negligible. On a design where a regulator or power stage is already running within ten degrees of its limit, it is not, and the correct response is to measure rather than assume — coat a sample, instrument the hot parts, and compare.

Potting is a genuine thermal redesign. Filling the enclosure removes internal air movement entirely, so every watt now leaves by conduction through the compound to the housing wall. Whether that is better or worse than the air it replaced depends on the compound: thermally conductive filled compounds can substantially improve heat transfer out of a sealed box, while an unfilled compound with poor conductivity can trap heat around a part that used to shed it into circulating air. The same fillers that improve conductivity usually raise viscosity, which makes complete void-free filling harder around tall or closely spaced components. Sort this out at the design stage with thermal measurements on a potted sample, not by reading the compound datasheet.

Component height is also a physical constraint worth checking early. Our SMT lines handle component heights up to 25 mm and boards up to 680 by 500 mm, which sets the practical envelope for what can be assembled before any fill operation; the full envelope is on our manufacturing capability page.

How do you prove the protection actually works?

A coating or potting specification without an acceptance test is a hope, not a requirement. Two kinds of evidence matter.

The first is process evidence: coverage and thickness. Industry acceptance documents publish criteria for what counts as acceptable coverage, thickness and freedom from defects such as bubbles, delamination and dewetting, and many coatings include a UV tracer specifically so coverage can be verified under black light. Agree with your assembler which criteria apply to your product, what gets inspected on every unit versus on a sample, and what evidence you receive — the answer should be in writing on the quotation, not assumed.

The second is performance evidence: does a coated or potted unit still work after the environment does its worst? This is where environmental stress screening pays for itself. Temperature and humidity cycling is the direct test of a moisture barrier, thermal cycling exposes the CTE-mismatch cracking that potting compounds cause, and a burn-in run under bias surfaces the leakage failures that a marginal film allows. At BELI we run temperature testing from minus 40 to plus 85 degrees Celsius at 20 to 98 percent relative humidity and burn-in for up to 24 hours in-house, and test reports go out with every shipment — which means a coated or potted sample set can be screened on our own equipment instead of being shipped out to a third party and back before you have an answer. Where a formal qualification standard applies to your market, the accredited-lab work still belongs at a lab; the split of responsibilities is the same one described in our guide to CE, FCC and UL compliance for China-built PCBA.

What does your assembler need in order to quote it?

Coating and potting are the two value-add processes most likely to be quoted vaguely and then argued about later. A quotable package includes: the material family or a specific approved product; the required thickness range or fill depth; a marked-up drawing or a dedicated fabrication layer showing every keep-out area; the application method you expect, if you have a preference; the acceptance criteria and inspection method; whether the housing is supplied by you or sourced by the factory; and the cure and handling constraints that follow, since some systems need days of full cure before the unit can be boxed.

Two of those items drive most of the price. The keep-out drawing drives masking labour, as above. The fill volume drives potting material cost, and it is worth calculating rather than estimating, because compound is bought by mass and a housing with more void than you think will quietly double that line. If the housing and any thermal interface parts are to come through the factory rather than from you, treat that as a sourcing question as much as a process one, and put every one of those items on the BOM with a specific part number. On the electronic side, component sourcing through authorized distributors and original manufacturers keeps traceability intact; mechanical and enclosure items deserve the same level of specification so nothing is left to the buyer's interpretation. The rest of the quoting package is the same set of files described in what files are required for a PCBA quotation, and the cost structure it feeds into is broken down in our guide to PCBA cost in China.

A short decision path

If you want a single sequence to run through: start by writing down the actual environment — temperature range, condensing or not, chemical exposure, vibration profile, expected service life. Then ask whether the enclosure can solve it, because sealing is usually cheaper than filling. If the enclosure cannot, ask whether the threat is chemical and electrical only, or also mechanical. Chemical and electrical alone points to conformal coating. Add sustained shock, vibration, high voltage or a tamper concern and it points to potting. Finally, ask what happens when a unit fails in year two: if the answer is "we replace it", potting is on the table; if the answer is "we repair it", coat it and keep the repair path open.

The one thing not to do is specify either process late. Both change the drawing package, both change the enclosure, and potting can change the thermal design outright. Decide during design validation, prove it on a sample set, and freeze it before the production validation build.

If you are weighing conformal coating against potting for a board we could build for you, send us your design files and a description of the operating environment. We quote within 24 hours, include a DFM review with every order, and can tell you where the masking cost is hiding in your layout before you commit to it. Conformal coating, potting and box-build are all available as value-add steps on the same lines that run SMT and through-hole assembly, with no MOQ for prototypes.

Related reading: EVT, DVT, PVT: Hardware Build Stages and Your EMS · Key Questions to Ask Before Outsourcing PCBA · How Much Does PCBA Cost in China

Buyers who have run several PCBA orders often assume box build is the same relationship with a bigger deliverable. It is not. A PCBA order is defined almost entirely by three files that the industry has standardised — Gerber, BOM, pick-and-place — and a factory can quote from them without a phone call. A box build order is defined by mechanical drawings, a wire list, a label spec, a test spec and a packing spec, and none of those have a universal format. That difference is why box build quotes come back slowly, why they come back with assumptions attached, and why two quotes for the same product can differ by a factor that has nothing to do with labour rates. This article maps the boundary between the two, then works through what you actually have to hand over to get a box build priced properly.

What does box build assembly actually include?

Box build — also called system integration, final assembly, or simply "the box" — is the step that turns tested circuit boards into a shippable product. There is no single industry definition of its scope, which is exactly the problem, but a typical box build covers some or all of the following:

• Enclosure assembly: fitting boards, standoffs, gaskets, thermal pads, fans, displays and mechanical hardware into a plastic or sheet-metal housing, to a torque and sequence defined by your drawings.
• Harness and internal cabling: building and routing the wire harnesses that connect boards to each other, to connectors on the panel, and to power inputs, plus strain relief and cable management.
• Panel work and labelling: front and rear panel fitment, membrane keypads, light pipes, product labels, rating plates, serial numbers, regulatory marks and barcodes.
• Software or firmware loading: flashing the product image before or after enclosure closure, then verifying the version that shipped.
• Final functional test: testing the assembled product against its specification rather than testing the board against a netlist.
• Packing: foam or moulded pulp inserts, accessories, documentation, retail or bulk cartons, pallet configuration and shipping marks.

BELI Technologies lists complete box-build assembly as a value-added option alongside conformal coating and potting on its turnkey PCB assembly line. Which of the six activities above are in scope for a specific programme is a conversation, not a default — and having that conversation before the quote is the single highest-leverage thing a buyer can do.

Where does PCBA stop and box build begin?

The boundary matters commercially because it decides who owns a failure. If a unit fails final test, the question "was that a board defect or an assembly defect?" only has a clean answer when both sides agreed in advance where one process ended and the other started. The table below is the boundary as most EMS providers draw it.

Two rows deserve a flag. Enclosure fabrication is the one most buyers assume is included and most often is not — an EMS may coordinate a mechanical supplier, but machining, moulding and tooling are a distinct sourcing exercise with its own lead time and its own tooling cost. And firmware loading sits on the boundary deliberately: whether it happens at all, and whether it happens before or after the box closes, is a per-programme decision you must confirm with your manufacturer rather than assume. If the difference between a bare board and a populated assembly is still fuzzy, our explainer on PCB versus PCBA sets out the earlier boundary in the same terms.

Why is box build more document-dependent than PCBA?

A PCBA is described by a data set that machines consume directly. The pick-and-place file tells a placement head where to go; the Gerber tells a stencil cutter where to open apertures. Ambiguity is limited because the file formats themselves force precision, and a DFM review — included with every BELI order — catches most of what remains.

Box build is human work described by drawings. An operator has to interpret "route the display cable behind the standoff, not over it", and there is no file format that enforces that instruction. The consequences are asymmetric in a way that surprises people: a board assembly defect is usually caught by AOI or ICT within minutes of being created, while a box build defect — a cable pinched under a lid, a gasket seated one turn too tight, a label 3 mm off centre — is often only caught at final test or, worse, by your customer. That is why box build is priced on documented labour minutes and why a factory that cannot see your documentation cannot honestly quote it.

The practical implication is that the time you spend producing a clear assembly drawing and a clear test spec is not overhead. It is the mechanism by which you convert your engineer's tacit knowledge into something a line of operators can repeat ten thousand times. Programmes that skip that step do not save the time; they spend it later on rework, on debugging failures over video calls, and on arguing about whose defect it was.

What do you need to provide before a box build quote?

Below is the package a manufacturer needs before it can quote a box build without wrapping the number in assumptions. The right-hand column is the useful one: it tells you what actually goes wrong when a given item is missing, which is usually more persuasive to a busy engineering team than "please send everything".

If you have already assembled the file package for board-level work, this list is the mechanical counterpart to it — our guide to the files required for a PCBA quotation covers the electrical half in the same detail.

How does the wire harness fit into a box build?

Harnesses are where box build most often stalls, because they are the one sub-assembly that is genuinely custom to your mechanical design and cannot be bought from a catalogue. Length depends on routing, routing depends on the enclosure, and the enclosure is frequently still changing when the first harness is ordered.

Sourcing harnesses from the same partner that does the box build removes a hand-off at exactly the point where the hand-off hurts. BELI builds custom wire harnesses and cable assemblies to IPC/WHMA-A-620 Class 2 and Class 3 workmanship standards, in wire gauges from AWG 28 to AWG 10, using JST, Molex, TE and other connector systems, with 100% continuity testing on every assembly. There is no minimum order quantity, samples turn around in 1 to 2 days, and batch production typically runs about two weeks — which means a harness revision driven by a mechanical change does not have to be a three-week schedule event.

Two specification points are worth settling early. First, name the workmanship class explicitly: Class 2 and Class 3 differ in crimp inspection rigour and defect tolerance, and quoting the wrong one either overpays or under-specifies. Second, put the wire colour and gauge for every circuit in the wire list rather than only on the schematic — operators build from the wire list, and a mismatch between the two documents is one of the most common causes of a mis-wired first article.

Who owns the enclosure, and how does that change the quote?

There are three workable arrangements, and the difference between them is worth more money than any labour-rate negotiation.

You supply the enclosure. You buy housings from your own mechanical supplier and consign them to the assembly factory. This gives you complete control over tooling ownership and cosmetic quality, and it puts the entire inbound logistics burden on you. It works well when you already have a tooling relationship, and the consignment mechanics are the same ones described in our comparison of turnkey versus consignment assembly: agree receiving checks, attrition allowance and excess disposition before anything ships.

The factory coordinates a mechanical supplier. You keep the design and the factory manages the buy — an arrangement to ask about rather than assume, since enclosure sourcing sits outside standard PCBA and box-build scope. It reduces your coordination load but requires you to be explicit about who approves first articles and who owns the tool.

Off-the-shelf housings. For instrumentation, industrial control and many low-volume products, a standard extruded or die-cast housing with custom panel machining avoids tooling cost entirely. It is almost always the right answer below a few thousand units a year, and it is the arrangement most often overlooked by teams who default to custom moulding because that is what consumer products do.

Whichever route you take, settle tooling ownership in writing. A tool paid for by you and held by a supplier is a commercial hostage; a tool paid for by the supplier and amortised into unit price is a different deal with different exit costs. Neither is wrong, but the two are frequently confused.

What testing happens at box-build level versus board level?

Board-level and product-level tests answer different questions and neither substitutes for the other. Board test asks whether the assembly matches the design: at BELI that means AOI and X-ray on every order, with in-circuit test reaching at least 95% node coverage, or flying probe where a dedicated fixture is not economical for the quantity. Custom test fixtures are built in 2 to 3 working days, and first-pass yield on the SMT lines runs at or above 99.5%.

Product-level test asks whether the finished unit does what the specification says. It exercises the interfaces a user touches — buttons, displays, connectors on the panel, power input — and it catches the failure classes that only exist once the box is closed: a connector seated but not latched, a fan blocked by a cable, a display ribbon that works on the bench and fails when the lid is torqued down. BELI's functional testing service also covers burn-in for up to 24 hours and environmental stress screening from -40 to +85 degrees C at 20 to 98% RH, with test reports delivered alongside every shipment.

What you have to supply is the definition of pass. Write the test specification as a list of measurable checks with numeric limits, in the order they should be performed, with a stated action for each failure mode — retest, rework, or scrap. A test spec written as prose descriptions of desired behaviour cannot be turned into a test station. Mechanical qualification such as drop, vibration and transit testing sits with an accredited test laboratory rather than with the assembly factory, and it is a separate scope item you appoint and budget for yourself.

Firmware loading, labels and packaging: what to confirm with your EMS

These three items are in scope for some box build programmes and out of scope for others, and each one has a specific question you should ask before assuming.

Firmware. Ask whether programming happens at the factory at all, at what point in the build, how the loaded version is recorded against each serial number, and what happens when you issue an update mid-run. Ask also how the image is protected — whether your binary is held in escrow, how many copies exist, and whether programming is done through a fixture that limits read-back. If your product carries intellectual property in software, this deserves a clause, not an email.

Labels and regulatory marks. Ask who supplies the artwork, who owns placement accuracy, and how a change to your regulatory marks propagates to the line. Certification marks are placed at your instruction and on your legal authority — the factory prints what you specify, and no manufacturer can certify your product for you. Our guide to CE, FCC and UL compliance for China-built PCBA sets out that split in full.

Packaging. Ask whether the factory expects you to supply a finished packing design or whether it can propose one, whether inserts are sourced locally, and how transit protection will be validated. Packaging design and drop testing are specialist disciplines; treat them as a scope item to be agreed rather than something that arrives free with the assembly quote.

How do box build lead times and volumes actually work?

Box build lead time is rarely set by the assembly itself. It is set by whichever input arrives last, and for most programmes that is either the enclosure or a long-lead component. Working backwards from the ship date, the chain typically runs: component procurement, then bare board fabrication, then PCBA, then harness build, then box build and final test, then packing.

The board-level portions of that chain are fast and well characterised. BELI returns board-level quotes within 24 hours, sources standard components from authorized distributors and original manufacturers in 3 to 7 days, runs quick-turn prototype assembly in 48 hours once boards and parts are on site, and completes turnkey orders including sourcing in typically 2 to 3 weeks. Harness samples take 1 to 2 days with batches around two weeks. Enclosures, by contrast, can take anywhere from days for machined prototypes to many weeks for a new injection mould — which is why the mechanical path, not the electronic one, usually determines your launch date.

On volume, the useful principle is that box build labour does not benefit from automation the way SMT does. A 52-million-placement-per-month SMT capacity means board cost falls sharply with quantity; a hand-assembled enclosure costs roughly the same per unit at 100 as at 1,000 until fixtures and jigs are justified. That is not a reason to avoid low-volume box build — it is a reason to expect the cost curve to look different from the one you know from PCBA, and to ask specifically at what quantity fixtures become worthwhile for your product.

How should box build fit into EVT, DVT and PVT?

Box build is where hardware programmes most often discover that their stage gates were optimistic. The pattern to avoid is validating electronics and mechanics on separate schedules and integrating them for the first time at the pilot build.

A more reliable sequence puts a small number of fully integrated units into the DVT build, however rough the enclosure is, because that is the earliest point at which cable routing, thermal behaviour inside a closed box, connector accessibility and assembly ergonomics can be observed at all. By PVT the build sequence, work instructions, fixtures and final test flow should be the ones that will be used in mass production, and the number worth demanding from the factory is not just yield but assembly time per unit — that figure is what your unit cost rests on, and it only becomes real when operators build to written instructions rather than to an engineer's supervision. Our breakdown of EVT, DVT and PVT hardware build stages covers the gate criteria in detail.

One more scheduling point: build the first article review into the plan explicitly. A first article on a box build should be inspected against the drawing dimension by dimension, photographed from every side, and signed off in writing before the balance of the run starts. Skipping it to save three days regularly costs three weeks.

How does shipping change for a finished product?

Finished goods are bulkier and more fragile than boards, and the freight arithmetic changes accordingly. Bare PCBAs ship in trays and travel well by express courier; boxed products are volumetric, which means air freight is priced on dimensional weight and sea freight becomes attractive far earlier than it does for boards.

BELI ships worldwide from Shenzhen via express couriers including DHL, FedEx and UPS, or by air and sea freight for larger volumes, with express delivery to North America, Europe and Israel typically taking 3 to 7 days depending on destination and customs. Beyond that, plan your own import arrangements with your freight forwarder and customs broker — duty classification, import documentation and delivery terms for a finished consumer or industrial product are a different exercise from importing components, and they belong to your side of the transaction.

Getting a box build quote that holds

The fastest route to a box build number you can budget against is to send the mechanical package at the same time as the electrical one, rather than after the PCBA quote is settled. BELI Technologies runs PCB fabrication, component sourcing, SMT and through-hole assembly, wire harness build, functional testing and box-build assembly under one roof in Shenzhen, under ISO 9001:2015, ISO 14001 and ISO 13485:2016 quality systems, with full process specifications on the manufacturing capability page. Send drawings, BOMs, the wire list and whatever test and packing specification exists — even in draft — and ask us to quote the whole product rather than just the boards. Where something is missing, we will tell you what it is and why it changes the number, which is more useful than a quote that quietly assumes.

Related reading: What Files Are Required for PCBA Quotation · EVT, DVT, PVT: Hardware Build Stages and Your EMS · Key Questions to Ask Before Outsourcing PCBA

Most buyers pick a model by reflex: startups go turnkey because they have no purchasing team, established OEMs go consignment because they already have distributor contracts. Both reflexes are defensible and both are frequently wrong for a specific build. This article goes past the headline definitions into the operational detail that actually decides which model works: how a BOM gets split under partial turnkey, what happens to your parts when they land at the factory, who eats the cost when a reel comes up short, and where the leftovers go when the run finishes. Our PCBA cost breakdown for China manufacturing compares the two models from the pricing side; this one is written from the shop floor.

What is turnkey PCB assembly?

Turnkey means the manufacturer takes the whole job: bare board fabrication, component procurement, assembly, inspection and test, and often box-build. You supply design data — Gerber or ODB++, a structured BOM, pick-and-place data — and receive tested assemblies. One purchase order, one supplier, one line of accountability.

The commercial consequence people miss is that turnkey converts a procurement problem into a schedule commitment. When the EMS owns the BOM, it also owns every lead-time surprise in it. At BELI Technologies, component sourcing runs through authorized distributors and original manufacturers, with standard parts typically available in 3 to 7 days and spot-market reach for hard-to-find or end-of-life items. A turnkey order including sourcing typically takes 2 to 3 weeks in total — that figure exists because procurement, not placement, is the long pole.

What is consignment PCB assembly?

Consignment — sometimes called "customer-supplied material" or CSM — means you buy the components yourself and ship them to the factory, which quotes and invoices assembly labour only. The factory receives, verifies, kits and places your parts.

Consignment is not simply "turnkey minus the parts cost". It changes what the factory can promise. Quick-turn assembly becomes possible — quick-turn prototype assembly at BELI takes 48 hours once boards and components are on site, with no MOQ — but the clock only starts when the last reel arrives. Any date you are given is conditional on your own kit being complete and correct.

Turnkey vs consignment across five dimensions

The last row is where most schedule disputes originate. A consignment quote showing "48 hours" and a turnkey quote showing "2–3 weeks" describe different clocks, not different factory speeds. Compare door-to-door dates, measured from the day you issue the PO, or you will conclude that consignment is five times faster than it is.

Partial turnkey: how a BOM actually gets split

In practice, most mature programmes run neither pure model. They run partial turnkey, where the BOM is divided line by line. BELI supports turnkey and consignment with no fixed ratio between them, which means the split can follow engineering logic rather than a contractual minimum.

The split that works is usually by risk class, not by cost:

• Customer-supplied: parts you already hold stock of; parts under a negotiated corporate contract you cannot beat locally; qualified single-source ICs where the vendor relationship carries firmware, NDA or allocation privileges; anything safety- or certification-critical where you refuse to delegate the sourcing decision.
• Factory-supplied: passives, connectors, discretes and commodity logic where the local supply chain is faster and cheaper than air-freighting reels from your side of the world; parts with volatile availability where you want the EMS chasing alternates; anything where the reel minimum buy vastly exceeds your build quantity.

Two administrative points make or break a split BOM. First, the split must be a column in the BOM file itself — a "supplied by" field on every line, agreed before quoting, not an email thread. Our guide to the files required for a PCBA quotation covers the rest of the BOM structure that a factory needs. Second, the shortage rule must be written down for both halves: what happens if a customer-supplied line is late, and what happens if a factory-supplied line goes on allocation. A split BOM without an agreed shortage rule is the single most common source of blame arguments in contract manufacturing.

What happens to your parts when they arrive at the factory

Consigned material should not bypass receiving control, and the receiving process is worth settling in writing before your kit ships rather than after a discrepancy appears. For the material BELI buys under turnkey, incoming components go through 100% visual inspection plus X-ray sampling, with traceability records maintained through the supply chain and original packaging preserved. If you are consigning, ask which of those checks will be applied to your kit, who signs off the count, and what happens when something fails — and get the answer attached to the quote, not to a later email.

The checklist below is what to agree with any assembler before a consignment kit leaves your building. Treat the right-hand column as what to ask for in writing rather than as a default that comes with every order:

Packaging deserves emphasis because it silently destroys consignment savings. Reels and full tape feed a pick-and-place head. Loose parts in a bag do not. If you save on unit price by buying cut tape and tubes, some of that saving comes back as manual placement labour and slower throughput — and hand-placed fine-pitch parts do not carry the same process confidence as machine placement.

Attrition: why your kit needs more parts than your build quantity

Every SMT process loses a small number of parts to feeder setup, first-article verification, misplacements caught by inspection, and rework. That loss is called attrition, and under turnkey the factory absorbs it inside its own material planning. Under consignment it is your problem, and it is the most common reason a consignment run stops one board short of the order quantity.

Agree an attrition allowance per line before you ship, and expect it to be higher in percentage terms on tiny quantities and on small passives — a 0201 chip capacitor has a far higher setup loss rate than a 64-pin QFP, and a 20-board prototype run cannot amortise feeder setup over anything. High-value single-source ICs are usually treated separately, with a small fixed spare count rather than a percentage. Whatever numbers you land on, put them in the kit list so the incoming count either passes or generates a shortage notice on day one.

Who is responsible when a part is missing or defective?

This is the clause that matters and the one most often left vague. The workable principle is that responsibility follows the sourcing decision.

Note the asymmetry in the counterfeit row. Consignment does not transfer the inspection duty away from the factory, but it does transfer the sourcing judgement to you. If you buy from a broker to beat a 40-week lead time, factory inspection is a safety net, not an indemnity.

Where does the leftover material go?

Excess is guaranteed under consignment, because reels and MPQs rarely divide neatly into build quantities. Decide the disposition before the run, and put it in the purchase order:

• Return with the shipment. Cleanest for prototype and one-off builds. Ask for excess to be returned in original packaging with a reconciliation sheet showing issued, consumed, scrapped and returned quantities per line.
• Hold as bonded stock for the next run. Sensible for repeat programmes. Agree in writing that the material remains your property, how long it will be held, and under what storage conditions — MSD parts in particular have a shelf life once the dry pack is opened.
• Scrap on your authority. For cheap passives where return freight exceeds the value. It should still appear on the reconciliation, never disappear quietly.

The document to insist on either way is that material reconciliation. It is the only thing that lets you close the loop between what you paid for, what went onto boards, and what you still own. Without it, consignment inventory quietly leaks and the loss surfaces two builds later as an unexplained shortage.

Which model should you choose?

One more consideration for scaling teams: whichever model you start with, staying with a factory that supports both means you can change your mind without changing supplier. Moving a programme between factories to switch commercial models means requalifying the process — a bad trade, as our guide to EVT, DVT and PVT build stages explains.

Does turnkey really cost more?

Turnkey quotes look higher because they contain the BOM. Whether they are more expensive is a different question, and one you can only answer by comparing like with like: your landed component cost plus your procurement labour plus inbound freight plus inventory carrying cost, against the factory's material line.

The honest generalisation is that a well-connected Shenzhen EMS often buys commodity passives, connectors and standard logic more cheaply than an overseas customer buying in prototype quantities, while a large OEM with corporate contracts usually buys its key ICs more cheaply than any factory can. That is precisely why partial turnkey exists. A BOM review that flags availability, lifecycle status and qualified alternates before the quote is the tool for this decision — our guide to cutting BOM costs without cutting corners shows what that kind of review can surface.

What to send for a quote under each model

The design package is the same either way — Gerber or ODB++, structured BOM, pick-and-place data. What differs is the commercial detail attached to it:

• Turnkey: BOM with manufacturer part numbers on every line, approved alternates flagged, no-substitute lines marked explicitly, and target build quantity plus any forecast that would change price tiers.
• Consignment: the same BOM plus a kit list stating packaging format per line, the attrition allowance you are supplying, expected arrival date, and the disposition rule for excess.
• Partial turnkey: both, with a "supplied by" column filled in on every single line — no blanks, because a blank is where an assumption grows.

Under either model, expect AOI and X-ray inspection with ICT or functional test as the quality baseline; BELI keeps first-pass yield at or above 99.5%, and its manufacturing is certified to ISO 9001:2015, ISO 14001 and ISO 13485:2016. Quality process does not change with who bought the parts, and it should not.

Work with a factory that supports both

BELI Technologies runs turnkey and consignment PCB assembly with no fixed ratio between them, so you can hand over the full BOM, ship us a complete kit, or split it line by line and change the split between builds. There is no MOQ for prototypes, quick-turn prototype assembly takes 48 hours once boards and parts are on site, and turnkey orders including sourcing typically run 2 to 3 weeks. If you are not sure which model suits your build, send us your BOM and design files — we quote from them and review the BOM for availability, lifecycle and alternates, which is usually enough to show which lines are worth consigning and which are quicker to buy here. Small-batch work is covered in more detail in our low-MOQ and small-batch PCB assembly guide.

Related reading: How Much Does PCBA Cost in China · What Files Are Required for PCBA Quotation · Key Questions to Ask Before Outsourcing PCBA

Almost every contract manufacturer in Shenzhen describes itself as one-stop. The phrase costs nothing to put on a website, and there is no certification body that audits it. What varies enormously is what sits behind it: some suppliers genuinely run fabrication, procurement, assembly and test as one operation with one production plan; others are sales offices that forward your Gerbers to a board house, your BOM to a broker, and your kit to a nearby SMT shop, then re-badge the invoice. Both quote the same job. Only one of them can actually control your schedule. This article is a verification method, not a pitch — a set of stage-by-stage checks, questions and a single-RFQ test you can run on any supplier, including us.

Why does "one-stop" mean so little on a supplier website?

The term has no definition and no threshold. A supplier that outsources 80% of the process chain and a supplier that outsources none can both write it truthfully in their own minds, because "one-stop" is describing the customer's experience — one contact, one PO, one invoice — rather than the production reality. Single-invoice convenience is real value, and for a small prototype order it may be all you need. But it is a different product from integrated manufacturing, and it fails in a different way when something goes wrong.

The failure mode matters more than the label. When an integrated supplier has a problem, it reschedules internally and tells you a new date. When a coordinator has a problem, it has to negotiate with the vendor that caused it, and you find out late because the bad news has to travel two hops before it reaches you. Buyers rarely discover which kind of supplier they hired during the quote. They discover it the first time a schedule slips or a batch fails inspection.

What does a genuine one-stop EMS actually own?

Working backwards from a finished, boxed, tested product, the chain has seven stages. A supplier is one-stop to the extent that it controls each one — and "controls" means it can change the schedule, hold a lot, and answer a root-cause question without making a phone call to a third party.

• Bare board fabrication — the physical PCB, its stack-up, drilling, plating and surface finish.
• Component procurement — buying parts, choosing channels, and screening what arrives.
• SMT assembly — paste printing, placement, reflow and automated inspection.
• Through-hole and selective soldering — connectors, transformers, anything not reflowable.
• Wire harness and cable assembly — the parts of the product that are not on a board.
• Test — electrical test, functional test, fixtures and test programmes.
• Box build — enclosure, final assembly, labelling, packaging.

Very few suppliers of any size own all seven under a single roof in the literal sense, and demanding that is not the point. Some stages are legitimately performed at affiliated plants or long-standing partner sites. The distinction that matters is whether the supplier is accountable for a stage — plans it, controls its quality records, and absorbs its problems — or merely passes it along. A supplier that says "our second plant does that, here is how we schedule between them" is being straight with you. A supplier that says "yes we do that" and then cannot describe the process is not.

How do you tell integrated manufacturing from relabelled subcontracting?

Each stage leaves different fingerprints depending on whether it is run in-house or bought in. The right-hand column is the question that forces a real answer, because it asks for a specific, checkable detail rather than a yes.

None of these questions are hostile, and none of them require you to reveal that you are testing. They are all questions an engineer would legitimately ask. What you are watching is not the content of the answer so much as its specificity and latency. In-house capability produces fast, concrete, slightly opinionated answers. Brokered capability produces answers that arrive the next day and repeat your own words back to you.

Why is the cost of a multi-vendor chain invisible on the quote?

Compare three quotes from a board house, a broker and an assembler against one integrated quote and the split chain often looks cheaper. It usually is cheaper on the line items, because each vendor is quoting only its own step and none of them is pricing the seams. The seams are where the money goes, and they never appear on any of the three documents.

The first cost is coordination labour. Somebody on your side becomes the programme manager: chasing the board house for a ship date, telling the assembler when to expect boards, re-explaining the same ECO three times. That person is usually your hardware engineer, and the hours come out of design work. The second is duplicated incoming inspection — every hand-off re-verifies what the previous vendor already verified, or worse, nobody does because each assumes the other did. The third is queue time. Three vendors mean three production queues, and your job waits at the back of each one; a chain of three five-day queues is fifteen days of calendar time even if every vendor hits its quoted lead time perfectly.

The fourth and largest is diagnostic time on failure. When a board fails functional test in a split chain, the question "is this a bare-board defect, a counterfeit part, or a soldering defect?" has no owner. The assembler blames the laminate, the board house blames the reflow profile, the broker blames neither because it has already been paid. Getting to root cause requires cross-sections, X-ray and dated process records that live in three different companies — and in practice, many teams give up and rebuild the batch instead. Our guide to how PCBA cost is built up in China covers the direct cost side; these four are the ones you only see afterwards.

What should a one-stop coverage checklist actually contain?

Take the seven stages and turn them into a document. Send it as part of your RFQ and ask for written answers. A supplier that can fill this in without hedging is worth a factory audit; one that cannot is worth a smaller first order.

Two columns are deliberately absent: price and promises. The checklist is designed to detect capability, not to negotiate. If you mix commercial pressure into the same email, you will get optimistic answers on both.

How can one RFQ tell you whether a supplier is really integrated?

You do not need an audit trip to run a first-order test. Send one enquiry that deliberately spans several stages at once — a small board with at least one fine-pitch part, a few through-hole connectors, a short harness, and a request for functional test — and then read the response as evidence rather than as a price.

An integrated supplier tends to answer with one document, one lead time built from internal stages, and cross-stage observations: a note that a connector choice affects selective soldering, or that the harness sample can be ready before the boards. A coordinator tends to answer in instalments, because each number has to come back from a different vendor, and the harness or test line will arrive a day or two after the PCBA price. Watch for a single consolidated schedule versus a stack of separate quotes stapled together, and for whether anyone comments on the interaction between your stages at all. Our list of questions to ask before outsourcing PCBA and the framework for choosing an EMS partner in China go deeper on scoring the answers; the file-level requirements are in what files a PCBA quotation needs.

When is one-stop the wrong answer?

One-stop is not universally superior, and pretending otherwise is how buyers end up disappointed. It is the wrong shape when you already own the capabilities it bundles.

A large OEM with a mature supply chain has negotiated component pricing that no EMS purchasing department will beat, an approved vendor list built over years, and a supplier quality team that can run a split chain competently. For that company, handing procurement to the assembler destroys pricing leverage and hides the supply base behind a middleman. Consignment or partial turnkey usually wins — see turnkey versus consignment PCB assembly for how the BOM gets split line by line.

One-stop is also the wrong answer when a stage genuinely requires specialisation the integrator does not have: exotic RF substrates, safety-critical qualification with a specific accredited scope, or a certification body's own test regime. And it can be the wrong answer when second-source resilience matters more than convenience — a single integrated supplier is a single point of failure, and some programmes are required to dual-source. The honest test is whether integration removes work you cannot do well, or removes control you would rather keep.

Where does one-stop pay off most?

The pattern is consistent: integration is worth most to teams that are small relative to the complexity of what they are building, and to programmes still changing shape. A five-engineer startup running its first build has no purchasing function, no supplier quality function, and no slack to spend chasing three vendors — the coordination cost lands directly on the person who should be writing firmware.

Integration also pays disproportionately during iteration. Between prototype and production a design changes repeatedly, and each change has to propagate through fabrication data, the BOM, the pick-and-place file and the test programme. Inside one company that propagation is a work-order revision. Across three vendors it is three email chains with three chances of a stale file reaching the floor. That is exactly why the EVT, DVT and PVT build stages are easier to run with one manufacturing partner: the same team carries what it learned at EVT into the PVT line rather than re-teaching a new supplier at every stage.

What should you audit once the paperwork checks out?

Written answers only get you to a shortlist. If the order matters, the next step is verification of a different kind: certificates with scope and expiry dates you can read, not logos on a page; a factory visit or a live video walk-through of the specific lines that will run your job; and a first article with real inspection records attached. Ask for the quality-system certificates by number and check what scope each one actually covers, because a certificate held by an affiliated plant does not automatically cover the line building your product. Then place a small paid order before a large one. A supplier's behaviour under a real PO tells you more than any audit checklist.

How does BELI's coverage check out against this list?

Run the checklist above against us rather than taking the label. PCB fabrication covers rigid, flexible, HDI and impedance-controlled boards up to 42 layers, with inner-layer line and space to 3.0/3.0 mil, impedance control of ±10%, DFM analysis included on every order, and quick-turn bare boards as fast as 24 hours for 1–2 layer designs or 3–7 working days for multilayer. Component sourcing runs through authorized distributors and original manufacturers with full traceability, original packaging preserved, and 100% visual inspection plus X-ray sampling on incoming parts; standard parts are typically available in 3–7 days, with spot-market reach for shortages, and both turnkey and consignment are supported with no fixed ratio.

PCB assembly covers SMT down to 0201/01005 and 0.4 mm BGA pitch at ±0.03 mm placement accuracy on boards from 45 × 45 mm to 680 × 500 mm, plus through-hole with wave and selective soldering, at a first-pass yield of ≥99.5% with AOI and X-ray on every order; quick-turn prototype assembly takes 48 hours once boards and parts are on site, and turnkey orders including sourcing typically run 2–3 weeks. Wire harness assembly is built to IPC/WHMA-A-620 Class 2 and Class 3 across AWG 28–10 with JST, Molex and TE connector systems, 100% continuity tested, samples in 1–2 days. Functional testing covers ICT at ≥95% node coverage, flying probe, functional test, burn-in up to 24 hours and environmental testing from −40 to +85 °C at 20–98% RH, with custom fixtures built in 2–3 working days and test reports delivered with every shipment. Value-added options after assembly include conformal coating, potting and complete box-build assembly. Full numbers are on the manufacturing capability page, and the manufacturing facilities behind them have over 15 years of operation under ISO 9001:2015, ISO 14001 and ISO 13485:2016.

If you are evaluating suppliers right now, the most useful thing you can do is send the same multi-stage enquiry to three of them, put the coverage checklist above in the same email, and compare how the answers come back. Send us your Gerber, BOM and pick-and-place files to start that comparison: quotes are returned within 24 hours, and DFM analysis is included with every order.

Related reading: How to Choose the Right EMS Partner in China · Key Questions to Ask Before Outsourcing PCBA · Turnkey vs Consignment PCB Assembly

Most hardware startups do not fail because the firmware was wrong. They lose weeks because a reel of one MCU sat in customs, a fab house and an assembly house disagreed about who owned a footprint error, or nobody noticed a part had gone end-of-life until the boards were already panelised. Turnkey exists to collapse those hand-offs into a single accountable supplier. This guide covers what turnkey actually includes, exactly who owns which deliverable, how the model maps onto EVT, DVT and PVT builds, and the three assumptions that most often cost early-stage teams a month.

What does turnkey PCB assembly actually cover?

In a turnkey engagement, the manufacturer owns the entire path from design files to a tested assembly. That means procuring the bare board, buying every line on the BOM, printing and placing, reflowing, soldering through-hole parts, inspecting, testing, and shipping. You supply data and decisions; the factory supplies materials, process and labour.

The important word is accountable. If a joint fails on a turnkey board, there is no argument about whether the pad was wrong, the part was counterfeit or the profile was off — one supplier owns all three variables. That single-throat-to-choke property is worth more to a small team than any line-item saving, because debugging a cross-supplier defect burns engineering time you cannot buy back.

BELI Technologies runs turnkey PCB assembly alongside fabrication, sourcing, wire harness and test, so the whole chain sits under one roof rather than being brokered out.

Turnkey deliverables checklist: who is responsible for what

The single most useful thing you can do before a first build is agree, in writing, who owns each item below. Ambiguity here is where schedules die.

Two rows deserve emphasis. First, alternates: turnkey does not mean the factory silently swaps parts. A responsible EMS proposes an alternate and waits for your written approval, because a 10% tolerance change on a feedback divider is your problem, not theirs. Second, compliance: no manufacturer can CE-mark or FCC-certify your product. They supply the evidence trail; you own the declaration.

Turnkey or consignment — where does the split actually change?

Consignment means you buy the parts and ship them in; the factory only assembles. The practical difference is narrower than it sounds: it comes down to who carries procurement risk, who eats the cost of a shortage, and who is accountable when a reel arrives short or a part turns out to be counterfeit. Under turnkey the factory carries all three. Under consignment you do, along with kitting accuracy and attrition allowances. Both are legitimate, and BELI supports turnkey and consignment with no fixed ratio, so a hybrid — you supply the two long-lead parts you already stock, the factory buys the other 180 lines — is entirely normal. The cost and risk mechanics of each model are worked through in our breakdown of what drives PCBA cost in China.

Misjudgement 1: "Turnkey must be more expensive"

This is the most common and the most expensive assumption. It compares the wrong two numbers: the factory's turnkey material line against the price you would pay on a distributor website, ignoring everything else the buy actually costs you.

The lines startups routinely forget when they self-source: minimum packaging quantities and reel breaks on parts you only need 40 of; multiple inbound freight charges from several distributors; import duty and brokerage on parts entering China; attrition stock so the line does not stop at the 47th board; the cost of holding leftover inventory for a design that will change; and the engineering hours your own team spends chasing purchase orders instead of debugging firmware. On low volumes the last item usually dominates everything else.

Turnkey also changes the shape of the risk. If a part is quoted, ordered and then arrives wrong, that is a supplier problem to solve rather than a founder problem. Genuine material savings come later, from consolidation, approved alternates and packaging choices — not from paying distributor spot prices on 40-piece buys.

Misjudgement 2: "We should wait until the design is frozen"

Teams often delay contacting a manufacturer until layout is final, on the theory that there is nothing to discuss before then. The opposite is true: the cheapest input a factory gives you arrives before the design is frozen.

Send a draft BOM while you are still choosing parts and a competent sourcing team will flag the lines that are single-source, on allocation, or already heading for end-of-life — while changing them is a symbol edit rather than a respin. Send preliminary Gerbers and DFM feedback catches annular-ring, clearance and footprint problems at the point where fixing them is free.

What that is worth is easiest to see in a case where it did not happen. A US smart-sensor client only discovered a problem after boards had been built and shipped: their team could not insert components without excessive force. A DFM review of the incoming Gerbers found drill holes specified at 0.8 mm with no allowance for lead insertion tolerance, and the 1.2 mm hole correction plus a pin-header swap was turned around in 48 hours with zero rework. Run before the first build, that review is free; run after, it is paid for on the assembly floor.

DFM review is included with every BELI order, prototype or production, so there is no commercial reason to hold files back until the last minute. If you are unsure what is worth sending at an early stage, our guide to the files required for a PCBA quotation lists what each file must contain and what is optional.

Misjudgement 3: "Nobody will take a 10-board order"

Plenty of volume-oriented factories will indeed decline, because changeovers cost them line time. But prototype-friendly EMS providers structure around exactly this. BELI has no MOQ for prototypes — a single board is a valid order — and quick-turn prototype assembly takes 48 hours once boards and components are in hand. A full turnkey order in which we also source the parts typically runs 2 to 3 weeks end to end, with the component lead time, not the assembly, being the variable.

What matters more than the minimum is whether the same supplier can still build you at volume. Lines that run 0201/01005 chips and 0.4 mm-pitch BGAs at ±0.03 mm placement, on boards from 45 × 45 mm to 680 × 500 mm, with roughly 52 million placements a month of capacity, do not need to change when your order goes from 10 to 10,000. The small-batch economics are covered in more depth in our low-MOQ and small-batch PCB assembly guide.

From first board to mass production: what turnkey looks like at each stage

Turnkey is not one service; what you are actually buying changes as the product matures. The stage names below follow the standard EVT/DVT/PVT gates, explained in full in our guide to EVT, DVT and PVT hardware build stages.

Quantity ranges here are the common industry pattern rather than a rule — a simple sensor node may need 15 EVT boards, a complex consumer device several hundred PVT units. The structural point is that the model does not have to change between rows. If your first bring-up board and your first production run happen on the same lines, with the same stencils, fixtures and inspection criteria, nothing has to be requalified at the point where you can least afford a delay.

What "tested board" should mean before you accept it

Turnkey ends at a tested assembly, so define what tested means. At minimum, expect automated optical inspection on every board and X-ray on hidden joints — the connections under a BGA or QFN are invisible to optical inspection, so without X-ray those joints ship unverified. Beyond that, expect either in-circuit test or functional test against your specification.

Every BELI order runs through AOI, X-ray, and ICT or functional test, at a first-pass yield of ≥99.5%. Where a dedicated fixture is not yet economical — typically at bring-up and EVT — flying-probe testing covers prototypes and small batches instead. For products that need reliability evidence before certification, our functional testing service also covers burn-in up to 24 hours and environmental stress screening from −40 to +85 °C at 20–98% RH. Ask for the test report to ship with the boards, not on request.

What to send for a turnkey quote

Three files unlock a real number: Gerber or ODB++, a BOM with manufacturer part numbers and approved alternates, and a pick-and-place file with rotations. Add assembly notes, target quantity, target IPC class, and your delivery destination and the quote comes back without a round of clarification emails.

What slows quotes down is predictable: BOMs listing "100 nF capacitor" with no MPN, alternates written as "or equivalent" with no tolerance or voltage constraint, missing rotations in the centroid file, and quantities described as "a few hundred, maybe". Each of those turns a 24-hour turnaround into a two-day conversation. BELI returns quotes within 24 hours when the file package is complete, and signs an NDA before you share anything.

How to sanity-check a turnkey supplier before you commit

Ask five questions and the answers will separate a real manufacturer from a broker: Do you fabricate, source, assemble and test in-house, or subcontract? Where do components come from, and can you show traceability and original packaging? What inspection runs on every order, and what first-pass yield do you actually hold? Is there an MOQ, and does the prototype line become the production line? Will you sign an NDA and give me a named engineer? A supplier who answers all five concretely — with numbers, not adjectives — is one you can scale with. BELI publishes its PCB and SMT manufacturing capability specifications in full, and every project is handled by a dedicated English-speaking project engineer from quotation through delivery.

Start a turnkey build

If you are a small hardware team weighing up whether to hand off sourcing and assembly, the cheapest next step is not a decision — it is a file review. Send your Gerber, BOM and pick-and-place files for a turnkey PCBA quote and you will get a DFM review, a BOM checked for availability, lifecycle and alternates, and an itemised quotation back within 24 hours. There is no MOQ for prototypes, so the first build can be a single board, and the same lines carry you through to volume without a supplier change.

Related reading: Prototype PCBA in China: A Practical Guide · EVT, DVT, PVT: Hardware Build Stages and Your EMS · Key Questions to Ask Before Outsourcing PCBA

This guide breaks down where the money actually goes, how quantity changes the per-unit math, who pays for what under turnkey versus consignment, and what an itemised quote must contain before you can fairly compare two suppliers.

The Five Cost Blocks of a PCBA Quote

Components Usually Dominate

For most assembled boards the BOM is the largest single share of total cost — often well over half once ICs, connectors and power parts are counted. That is why two boards of identical size can differ in price by 10x, and why a serious quote requires your BOM rather than a board photo.

It is also where the biggest savings live. Qualified alternates sourced through authorized distributors and original manufacturers can cut BOM cost substantially without touching your design — our case study on cutting BOM costs without cutting corners shows how that works in practice, always with written client approval before any substitution.

Bare PCB Cost

Board cost scales with layer count, area and material. A standard 2-4 layer FR-4 board is inexpensive at volume; I stack-ups, Rogers high-frequency laminates, heavy copper and tight impedance control all add cost. Efficient panelization — fitting more boards onto each production panel — meaningfully lowers per-unit fabrication cost as quantity rises.

Assembly and Test

Assembly is priced largely per placement, with premiums for fine-pitch BGAs (which require X-ray verification), double-sided reflow and through-hole work. Testing adds cost but protects you from shipping failures: AOI and X-ray are baseline at BELI, with ICT or functional testing quoted to your required coverage.

How Quantity Changes the Per-Unit Price

One-time costs — stencils, test fixtures, machine programming, line setup — are identical whether you build 5 boards or 5,000. Divided across a small run they dominate the invoice; across a production run they nearly vanish. This single mechanic explains most of the confusion about "why was my prototype so expensive per board".

The arithmetic below uses illustrative figures, not BELI pricing, and deliberately holds the per-unit build cost constant so the NRE effect is isolated. Assume one-time costs of 400 currency units and a per-unit build cost of 12:

In reality the build cost falls as well, because component pricing steps down at distributor volume tiers and setup time is spread over a longer run — so the real curve is steeper than the table shows. The practical takeaway: judge prototype pricing against prototype economics, never against your projected mass-production unit cost. Our prototype vs mass production cost breakdown works through that comparison in detail.

Turnkey or Consignment: Who Pays for What

The engagement model changes both the quoted number and where your risk sits. BELI supports both, with no fixed ratio between them — you can hand off the entire BOM, supply everything yourself, or split the list.

A frequent and expensive mistake is comparing a turnkey quote from one supplier against a consignment quote from another. They are not the same number.

What an Itemised Quote Should Contain

Two quotes are only comparable if they break out the same lines. Ask every supplier for:

• Bare PCB cost per unit, with layer count, material, finish and panel utilisation stated
• BOM cost per unit, line by line, with manufacturer part numbers and the channel each part comes from
• Assembly cost per unit, with placement count and any fine-pitch or through-hole premiums shown separately
• Test and inspection scope — what is baseline, what is extra, and what coverage the extra buys
• NRE broken out as a one-time line, never folded into the unit price
• Tooling ownership — who keeps the stencils and fixtures you paid for
• Lead time, and what it is measured from (order date, or the date the last component lands)

If a quote arrives as a single number with no breakdown, you cannot tell whether it is cheap because the supplier is efficient or because something has been left out.

Costs Buyers Forget to Budget

Why the Cheapest Quote Often Costs More

Price differences between competent suppliers are usually modest. A quote that is dramatically below the others is normally explained by something that has been removed rather than something that has been optimised. Common causes:

• Test scope reduced to visual inspection only, so failures reach you instead of the factory
• Parts sourced from open-market brokers rather than authorized channels, trading counterfeit risk for margin
• NRE hidden inside the unit price, so the number rises on your second order
• The quote covers assembly only while you assumed it covered components
• No DFM review, so manufacturability problems surface as scrap and rework after the build

The questions worth asking before you commit are collected in key questions to ask before outsourcing PCBA.

Why Shenzhen Remains Cost-Competitive

Shenzhen's advantage is not mainly labour rates — it is supply-chain density. Components, laminates, stencils and tooling are all sourced locally within days, and a one-stop EMS keeps fabrication, sourcing, assembly and test under one roof, removing both the coordination overhead and the margin stacking of a multi-vendor chain. For small teams that also means one point of contact instead of four vendors in a foreign supply chain.

How to Get an Accurate Quote

No honest manufacturer publishes a flat PCBA price list, because the BOM makes every board different. What we can promise is a fast, itemised number: send your Gerber (or ODB++), BOM and pick-and-place file — here is exactly what each file needs to contain — and BELI returns an itemised quotation within 24 hours. We sign an NDA before you share anything, DFM review is included, and there is no MOQ for prototypes.

Request your free PCBA quote and get a real number for your board, not an average.

Common Questions

Can you give me a ballpark before I send files? Only a meaningless one. Without the BOM, the two largest cost blocks — components and placement count — are unknown. Sending three files takes minutes and gets you a real number within 24 hours.

Is there a minimum order quantity? No MOQ for prototypes. A single board can be built on the same lines that run volume production.

Why is my 10-piece prototype so expensive per board? One-time costs are spread across 10 units instead of thousands. See the quantity table above — the per-unit NRE, not the build itself, is what you are reacting to.

Does a cheaper board mean lower quality? Not necessarily, but it should prompt a question about what changed. Compare the itemised lines above before concluding one supplier is simply cheaper than another.

Related reading: PCB Prototype vs Mass Production Cost Breakdown · How We Cut Client BOM Costs Without Cutting Corners · What Files Are Required for PCBA Quotation

This guide draws the exact line between what your EMS partner does and what you, the brand owner, must own for CE, UKCA, FCC, UL, and Israeli import approval. We build electronics for these markets, so we will be blunt about what a factory legally cannot do — the supplier pages that promise "we handle all certifications for you" are the ones to walk away from.

The #1 misconception: a factory certificate does not certify your product

Manufacturer quality systems and product certifications are two different legal objects. ISO 9001 certifies a factory's quality management process; CE marking declares that a specific product meets EU product legislation. A supplier's ISO 9001 certificate — or ISO 13485, or any other management-system standard — does not make your product CE compliant, FCC compliant, or UL Listed. If a sales page implies otherwise, the vendor either misunderstands compliance or hopes you do.

Both layers matter, for different reasons. Product certification proves your design passes safety and EMC requirements once, on the samples the lab tested. A quality system is what keeps unit number 10,000 identical to that tested sample. When you evaluate an EMS, verify both: the process certificates the factory holds and the concrete compliance support it provides. Our certificates are ISO 9001:2015, ISO 14001, and ISO 13485:2016, and we tell customers plainly that they support — never replace — product-level certification.

Who is legally responsible for CE, FCC, and UL compliance?

The company that places the product on the market under its own brand carries the legal responsibility for compliance — not the contract manufacturer. This holds across every major regime and answers the most common question new customers ask.

CE (EU/EEA). Under EU law, the "manufacturer" is the entity that markets the product under its own name — you, not your factory. Your Chinese EMS can print the CE mark on your product at your instruction; that part is trivial. But the act that gives the mark legal force — signing the EU Declaration of Conformity and maintaining the technical file — is yours, as the official EU CE marking guidance spells out. Most PCBA-based products under the EMC and Low Voltage Directives can be self-declared — in practice on the strength of accredited-lab test reports, which is what market-surveillance authorities and customers expect; a Notified Body is only mandatory in specific cases, most commonly some radio equipment routes under the RED.

FCC (US). Products with only unintentional radiators (a digital device with no radio) use the Supplier's Declaration of Conformity — and FCC equipment authorization rules require the SDoC responsible party to be located in the United States. A Chinese factory cannot fill that role. If your company has no US presence, your US importer — or a designated US agent — serves as the responsible party; budget for that arrangement before launch. Products with intentional radiators (Wi-Fi, Bluetooth, LoRa, cellular) need FCC Certification through a Telecommunication Certification Body — typically a few thousand dollars and 4-8 weeks — and the grantee is your company, unless you integrate a module with full modular approval, in which case the module maker holds the grant and your host device is normally handled under SDoC, subject to the module's integration conditions. That is why pre-certified radio modules are the standard cost-saver.

UKCA (UK). Post-Brexit Great Britain has its own UKCA mark, but the UK government now recognizes CE marking indefinitely for most electronics categories. In practice, most teams test once to the harmonized standards and declare both — confirm your specific product category before relying on CE alone.

Israel (SII / MoC). Missing from almost every guide on this topic: Israeli imports of electrical products typically require approval involving the Standards Institution of Israel (SII), and radio equipment needs type approval from the Ministry of Communications. The Israeli importer of record runs this process, and existing CE and FCC test reports are normally accepted as core evidence — one more reason to test against harmonized standards early. We cover the logistics side in our guide to PCB assembly for Israeli hardware startups.

Which mark applies where? A one-table reference

The UL trap: a "UL certified" factory is not a UL Listed product

This is the single most abused claim in China electronics sourcing. Many PCB fabricators advertise "UL certified" — which almost always means UL 796 recognition of their bare-board fabrication process and UL 94 flammability ratings on laminates. Those are component-level recognitions of the bare board. UL Listing applies to a complete end product and requires a separate UL (or other NRTL) investigation of your finished device, followed by ongoing factory surveillance — a fab's UL 796 file does not make your product UL Listed. UL Listing commonly starts around $5,000-15,000 plus quarterly follow-up service fees, which is why it is pursued only where retailers or local inspection authorities demand it.

To be equally transparent: BELI does not hold UL certification and does not claim to. What we do is source the components on your BOM through authorized distributors and original manufacturers, preserving certificates of conformance, datasheets, and manufacturer documentation — including UL file numbers where the component maker publishes them — so your NRTL evidence pack is ready when the investigation starts. When a supplier says "UL certified," always ask: which UL file number, and what exactly does it cover?

What a good EMS actually contributes to your compliance

The factory is the technical executor, and a competent one removes most of the friction. Your EMS should provide:

• Compliant materials: RoHS and REACH compliant processes, solder, and laminates, with material declarations traceable to authorized distributors and original manufacturers.
• Component traceability and build records: lot-level records tying each production run to the exact BOM revision that was tested — the backbone of your technical file. This is what disciplined PCB assembly under ISO 9001:2015 and ISO 13485:2016 buys you.
• Critical-component evidence: datasheets, certificates of conformance, and — where the manufacturer publishes them — UL file numbers for the safety-relevant parts on your BOM, preserved during component sourcing rather than reconstructed in a panic before the audit.
• DFM and EMC-aware layout review: catching broken return paths, poor stackup choices, and unfiltered connector lines before they become a failed radiated-emissions scan. We include DFM review on every order.
• Production test data: AOI, X-ray, ICT, and functional test records showing that shipped units match the qualified design.
• Local lab proximity: a Shenzhen EMS sits near accredited EMC labs — ask whether your partner can help you book pre-compliance scans locally. Informal pre-compliance scans often cost a few hundred dollars with results in days; formal EMC testing at a CNAS/ISO 17025 Shenzhen lab typically runs $600-1,200 and two to four weeks — a fraction of US or EU lab pricing.

The step-by-step path from design to Declaration of Conformity

Compliance goes wrong when it is treated as a final checkbox. Sequence it with your NPI milestones instead:

1. During design: identify the directives and standards that apply (EMC, LVD, RED, FCC Part 15) and design to them — pick pre-certified radio modules, plan filtering and grounding.
2. At prototype: run a DFM and layout review, then build a small lot. Our prototype PCBA process exists so you have test-ready hardware early.
3. Pre-scan: take prototypes to a local EMC lab for pre-compliance scans. Fixing a failure at this stage costs a respin; fixing it after tooling costs a launch date.
4. Formal testing: full testing that generates the reports your DoC and SDoC rest on. ISO/IEC 17025 accredited labs are mandatory for FCC certification and strongly recommended for CE and SDoC so your reports are defensible.
5. Technical file: compile schematics, BOM, risk assessment, test reports, and manufacturing records — your EMS supplies the manufacturing half.
6. Declare and mark: sign the DoC (CE/UKCA), file the SDoC or certification grant (FCC), and instruct your factory on exact label artwork.

Do not let production drift void your certification

The dirtiest secret in low-cost assembly is the certified golden sample: a factory passes testing with one BOM, then quietly substitutes cheaper components in production. Any change to a safety- or EMC-relevant component after testing can invalidate your CE, FCC, or UL compliance — a different switching regulator or connector is legally a different product. The safeguards are contractual and procedural: a locked approved-vendor list, written ECO approval for every substitution, component-level traceability, first-article inspection, and retesting after significant changes. This is why quality-system discipline matters even though it certifies no product: it keeps the product you ship identical to the product you certified. Our checklist of questions to ask before outsourcing PCBA includes the change-control questions that expose weak suppliers.

If you are planning a build for the US, EU, UK, or Israeli market and want a manufacturing partner that treats your technical file as seriously as your solder joints, send us your files for a quote — you will get a DFM review with every order, an English-speaking project engineer, and honest answers about what we can and cannot certify.

Related reading: How to Choose an EMS Partner in China · How to Choose a Reliable PCBA Manufacturer in China

What does NPI mean in hardware manufacturing?

New Product Introduction (NPI) is the structured process of taking a hardware design from a working prototype to stable, repeatable mass production. In practice, NPI is organized as a sequence of builds — EVT, DVT, PVT — each with a defined purpose, a defined quantity, and defined exit criteria. You do not move to the next stage until the current one passes its gate. The point of the structure is simple: it is far cheaper to find a problem on 30 boards than on 3,000 units sitting in a warehouse.

On the EMS side, NPI is also where most of the engineering work happens: design-for-manufacturing (DFM) review, stencil design, test fixture development, first-article inspection, and yield analysis all live inside these three stages.

EVT, DVT, PVT: one-sentence definitions

Engineering Validation Test (EVT) is the build stage that proves your design works — the electronics perform every core function on boards assembled with production-representative processes.

Design Validation Test (DVT) is the build stage that proves the complete product — electronics, enclosure, and firmware together — meets its full specification, including environmental, reliability, and regulatory requirements.

Production Validation Test (PVT) is the build stage that proves the factory can manufacture the product at target yield, rate, and cost using final tooling, fixtures, and test coverage.

The difference between DVT and PVT trips people up, so put it this way: DVT validates the product; PVT validates the production process. A unit that passes DVT proves the design is right. A run that passes PVT proves the factory can make that design correctly, quickly, and repeatedly.

Stage by stage: quantities, testing, and exit criteria

Quantities scale with product complexity and channel commitments — a simple sensor node might do EVT on 15 boards, while a complex consumer device might need 500 PVT units. On timelines, plan roughly 4–6 weeks per EVT loop (including your redesign time), around 8 weeks for DVT because environmental testing and certification sample preparation gate it — formal certification testing usually runs in parallel with PVT — and 4 or more weeks for PVT. The build itself is rarely the bottleneck; the analysis and rework between builds is.

Where prototype, pilot, and mass production map onto an EMS engagement

From the factory's side, the three stages translate into three familiar engagement modes:

EVT = quick-turn prototype assembly. What you need here is iteration speed and no volume commitment. At BELI Technologies we run prototype PCB assembly in 48 hours once boards and parts are ready, with no MOQ — so an EVT loop of 20 boards, a week of bench testing, and a respin does not cost you a purchasing negotiation each time. If you want us to handle parts, turnkey including sourcing from authorized distributors and original manufacturers typically runs 2–3 weeks.

DVT and PVT = small-batch pilot runs. Builds of 50–1,000 units on production SMT lines, with real fixtures, real test coverage, and yield data reported back after each run. This is also where first-article inspection (FAI) reports and process documentation get formalized.

MP = the same lines, scaled. This matters more than it sounds: if your pilot builds run on the same lines and processes that will build mass production, nothing needs requalification at the MP gate. The stencils, fixtures, programs, and inspection criteria proven at PVT carry straight over. If your prototype shop and your volume factory are different companies, you effectively repeat part of PVT — at the worst possible moment.

Before any EVT build starts, your EMS needs a complete file package: Gerbers (or ODB++), a structured BOM, pick-and-place data, assembly drawings, and whatever test spec exists. We covered the details in what files are required for a PCBA quotation. Cost-wise, expect NRE items — stencils, programming, fixtures — to dominate early builds, while per-unit cost only becomes meaningful at PVT; the breakdown in prototype vs. mass production cost explains why quoting EVT boards against MP pricing misleads everyone.

How DFM and test coverage advance with each stage

The EMS's job changes character across the stages, and test coverage is the clearest way to see it.

At EVT, the highest-value input is DFM feedback before the build: footprint errors, acid traps, insufficient annular rings, unbuildable component orientations. BELI includes a DFM review on every order, prototype or production. Test coverage at EVT is deliberately light — AOI plus flying probe — because building dedicated fixtures for a design that will change is wasted money.

At DVT, the board is stable enough to invest in fixtures. This is when ICT (in-circuit test) fixtures and functional test jigs get built — ours are ready in 2–3 working days — and test coverage jumps: our ICT targets at least 95% node coverage, paired with functional testing against your spec. DVT is also where environmental screening earns its keep: we run burn-in up to 24 hours and temperature/humidity testing from -40 to +85°C at 20–98% RH in-house, catching marginal designs before certification; drop, vibration, and EMC pre-compliance typically run at a test lab, with the EMS building and serializing the sample sets.

At PVT, nothing new should be invented — the goal is to run the full production test flow (SPI, AOI, X-ray for BGAs, ICT, functional test) at line rate and measure the result. The deliverable you should demand from your EMS after a PVT run is data: first-pass yield, a defect Pareto, and the corrective actions for anything below target. Sustained first-pass yield of 99.5% or better is the standard we hold our own lines to; full process capabilities are on our capability page.

When should FCC, CE, and UKCA testing happen?

Certification testing belongs at DVT. Your golden samples must represent the final design — same PCB revision, same BOM, same enclosure — so submitting before DVT wastes lab fees, and submitting after PVT delays launch. Book the test lab as DVT planning starts, because FCC equipment authorization and CE marking lab queues routinely run several weeks. UKCA follows the same golden-sample timing as CE; since the UK continues to recognize CE marking for most product categories, many teams cover both markets with a single DVT test campaign.

Be clear about roles here: product certification is the brand's responsibility, not the factory's. An EMS supports it with build records, material declarations, and serialized traceability. BELI operates under ISO 9001:2015, ISO 14001, and ISO 13485:2016 with RoHS- and REACH-compliant processes and materials, and provides the documentation your compliance file needs — but no factory can "CE certify" your product for you, and anyone who claims otherwise is a red flag.

Common failure modes (learn from other people's scars)

Skipping DVT. The design worked at EVT, money is tight, so the team jumps to a big build. Then units fail EMC testing, or the enclosure warps at 60°C, and the fix requires a board respin — now multiplied across 1,000 assembled units instead of 100. DVT exists precisely to make failures cheap.

BOM churn during PVT. Every part swap after DVT — a cheaper connector, an alternate MCU because of lead time — invalidates some of your validation data and can trigger re-quoting, new fixtures, or even re-certification. Lock the BOM at the DVT exit gate and route any change through a formal ECO with your EMS, so both sides know exactly which revision is on the line.

Requalifying a new supplier at MP. Switching factories after PVT to chase a lower unit price means new FAI, new fixtures, new operator learning curves, and a fresh yield ramp — usually costing more than the savings. Qualify the factory that will scale with you before DVT, not after PVT. Our guide on choosing an EMS partner in China covers what to check.

Can startups compress the stages? A low-volume playbook

Yes — carefully. Combining EVT and DVT into one build is reasonable when the board is simple, derived from a proven reference design, and the regulatory path is short. It is a bad idea for anything with RF, high power density, safety implications, or a novel mechanical design. What you should never compress is PVT's core question — does the production test flow catch real defects at real line speed? — because that is the only thing standing between you and shipping bad units.

And yes, PVT units can usually be sold, provided they pass the full production test flow, certification is complete, and each unit carries serialized traceability so a later ECO can be mapped to affected serial numbers. Many startups sell PVT units to early customers or use them as certification and channel samples.

If you are heading into an EVT build — or trying to get from a working prototype to a pilot run without changing factories midstream — send us your files. We sign an NDA before you share anything, quote within 24 hours, and assign a dedicated English-speaking project engineer who will run your builds on the same lines from first prototype to mass production.

Related reading: Prototype PCBA in China: The Complete Guide · Low-MOQ Small-Batch PCB Assembly

Why Israeli Hardware Teams Manufacture in Shenzhen

The pull is the same as for US and European companies, but sharper for startups on tight seed budgets: Shenzhen's supply-chain density means components, bare boards, stencils, and tooling are all local, so iteration is fast and pricing works even at prototype volumes. A one-stop EMS — fabrication, component sourcing, assembly, and test under one roof — also means a two-person hardware team doesn't need to manage four vendors in a foreign supply chain.

Working Days and Communication

Israel runs Sunday–Thursday; China runs Monday–Friday, five to six hours ahead (UTC+8 vs UTC+2/+3). In practice that gives four full shared working days, Monday to Thursday — and because an Israeli morning is a Chinese afternoon, questions sent at the start of your day get answered the same day. At BELI Technologies, every project is handled by a dedicated English-speaking project engineer, and quotes and technical questions are answered within 24 hours.

Shipping from Shenzhen to Israel

Prototype and small-batch orders ship by express courier (DHL, FedEx, UPS), typically arriving in Israel in 3–7 days depending on customs. Larger production runs can move by air or sea freight. We prepare the commercial invoice and shipping documentation your customs broker needs, and mark shipments per your instructions.

Prototypes Without MOQ — Built for Iteration Speed

Startup hardware development is iterative: EVT boards, a respin, a pilot batch, then production. A volume-oriented factory with high minimum order quantities fights that rhythm. BELI runs no MOQ for prototypes — you can build one board — and quick-turn prototype assembly takes 48 hours once boards and components are ready. A full turnkey order where we also source the components typically takes 2–3 weeks. The same lines then scale to volume, so nothing gets re-qualified when you ramp: see our guide to turnkey PCB assembly for startups.

Medical Devices and Quality Systems

Medical technology is one of Israel's strongest hardware verticals, and it raises the certification bar for any manufacturing partner. BELI's processes are certified to ISO 13485:2016 (medical devices quality management) alongside ISO 9001:2015 and ISO 14001, with RoHS- and REACH-compliant materials — and every order is verified with AOI, X-ray, and ICT or functional testing at a ≥99.5% first-pass yield. For CE-marked products bound for European or Israeli markets, we provide the build records and component traceability your regulatory file needs.

Component Sourcing You Can Audit

Counterfeit components are the risk buyers worry about most when moving production to China — reasonably. BELI sources exclusively through authorized distributors and original manufacturers, preserves original packaging and full traceability records, and runs 100% visual inspection plus X-ray sampling on incoming parts. During shortages, substitution proposals come to you with full documentation and nothing changes without your written approval.

IP Protection: NDA Before Files

For IP-driven Israeli startups, design security is non-negotiable. Every BELI client engagement begins with a signed NDA, and design files are stored on access-restricted servers with strict internal controls. You retain ownership of custom tooling and test fixtures you pay to develop — worth confirming with any manufacturer you evaluate, as covered in our list of questions to ask before outsourcing PCBA.

How to Start

Send three files — Gerber (or ODB++), BOM, and pick-and-place — and you will have an itemized quotation within 24 hours (here is exactly what each file needs to contain). From there, a DFM review catches manufacturability issues before anything is built.

Building hardware in Tel Aviv, Haifa, or anywhere in between? Get a free quote from BELI — prototype to production, with no MOQ to start.

Related reading: Turnkey PCB Assembly for Startups: A Practical Guide · Prototype PCBA in China: A Practical Guide

The short version: a PCB is the bare board; a PCBA is that board with all components assembled onto it — a functional circuit ready for testing and integration into your product.

What is a PCB

A Printed Circuit Board (PCB) is the structural foundation of an electronic device. It is made of insulating material — most commonly fiberglass (FR-4), but also Polyimide for flex circuits, aluminum for thermal management, or high-frequency laminates like Rogers — with etched copper traces that will connect the electronic components.

A PCB by itself contains no components. It only provides the conductive pathways and mounting structure for the circuit. Typical PCB elements include:

• Copper layers — from a single layer up to 42 layers in advanced multilayer designs
• Solder mask — the (usually green) protective coating that prevents accidental shorts
• Silkscreen — printed reference designators and markings
• Vias and holes — plated connections between layers and mounting points

Producing the bare board is called PCB fabrication — a chemical and mechanical process of imaging, etching, laminating, drilling, and plating. To quote it, a fabricator needs only your Gerber (or ODB++) files.

What is PCBA

PCBA (Printed Circuit Board Assembly) refers to the process — and the resulting product — of assembling electronic components onto the bare PCB. Resistors, capacitors, integrated circuits, and connectors are soldered onto the board to create a functional electronic circuit.

Modern PCB assembly combines two technologies, often on the same board:

• SMT (Surface Mount Technology) — components are placed onto solder paste by high-speed pick-and-place machines and soldered in a reflow oven. This handles the vast majority of modern parts, down to 01005 passives and fine-pitch BGAs.
• THT (Through-Hole Technology) — component leads pass through drilled holes and are wave- or selective-soldered, typically used for connectors and parts needing mechanical strength.

Assembly also involves component procurement, solder paste inspection (SPI), AOI and X-ray inspection, and electrical testing — which is why quoting a PCBA requires more from you than quoting a bare PCB: a BOM and a pick-and-place file in addition to Gerbers.

PCB vs PCBA Comparison

What the Difference Means When You Buy

The distinction has three practical consequences for anyone requesting quotes:

• Files. A PCB quote needs Gerber/ODB++ files only. A PCBA quote also needs your Bill of Materials (BOM) and pick-and-place (CPL) file — our guide to the files required for a PCBA quotation covers the details.
• Cost structure. For most assembled products, the bare board is a minor share of total cost — components usually dominate, followed by assembly and test. If you compare quotes, make sure you are comparing the same scope (bare PCB vs. full PCBA).
• Supplier scope. Some vendors only fabricate; some only assemble boards you supply. A one-stop EMS partner does both — plus component sourcing and functional testing — so responsibility for the finished, working board sits with a single team.

A Common Misconception

"PCB manufacturer" is often used loosely to mean either a board fabricator or a full assembly house, which causes real confusion in sourcing. When you contact a supplier, be explicit: are you buying bare boards (PCB), assembly of parts you will supply (consignment PCBA), or finished boards including parts (turnkey PCBA)? Each is quoted differently.

Applications

PCBs and PCBAs are at the core of products across consumer electronics, industrial automation, medical devices, robotics, IoT, and telecommunications — anywhere an electronic circuit needs a reliable physical platform.

Conclusion

PCB and PCBA represent two essential stages in electronics manufacturing: the PCB provides the physical structure, and assembly transforms it into a functional electronic product. BELI Technologies provides both PCB fabrication (up to 42 layers) and turnkey PCBA services under one roof, from a single prototype — there is no MOQ — to mass production.

Not sure which you need? Send us your files and our engineers will scope it with you — we respond within 24 hours.

Related reading: What Files Are Required for PCBA Quotation · How Much Does PCBA Cost in China

The Three Essential Files

Gerber or ODB++ Files

Gerbers describe the physical board — every copper layer, solder mask, paste layer, silkscreen, and the drill file. Export the complete set from your EDA tool (KiCad, Altium, Eagle, etc.) and zip it. For bare-board fabrication quotes, Gerbers alone are enough; also state layer count, material, thickness, copper weight, and surface finish if they are not embedded in a fab notes layer.

BOM (Bill of Materials)

The BOM drives most of the cost of a turnkey order, so its quality directly affects quote accuracy. A good BOM includes one row per part with:

• Reference designators (R1, C3, U2…) matching the silkscreen and CPL
• Manufacturer and manufacturer part number (MPN) — not just a distributor SKU
• Quantity per board and package/footprint
• Approved alternates, or a note that substitution proposals are welcome — this is where a good sourcing team can save real money on long-lead parts
• DNP markings for parts that should not be populated

Pick-and-Place / Centroid File (CPL)

The CPL gives each component's X/Y coordinates, rotation, and board side, and is used to program the SMT machines. Any modern EDA tool exports it in one click. Mismatched reference designators between BOM and CPL are a leading cause of clarification rounds — export both from the same design revision.

Helpful but Optional

• Assembly drawings or a 3D render — resolves orientation questions for polarized parts and connectors
• Test requirements — if you need ICT or functional test, describe pass/fail criteria so fixtures can be quoted
• Special processes — conformal coating, potting, or box-build requirements
• Panelization preferences — or let the manufacturer propose the most economical panel

Common Mistakes That Delay Quotes

• BOM rows without an MPN ("10k resistor" is not orderable)
• Missing drill file in the Gerber set
• BOM and CPL exported from different design revisions
• No indication of which parts allow alternates during shortages
• Quantities that don't state whether they are per-board or per-order

What Happens After You Send Files

At BELI, incoming files go through a DFM (Design for Manufacturability) and BOM review before quoting — footprint issues, sourcing risks, and long-lead parts get flagged up front rather than mid-build. Quotes are returned within 24 hours. For turnkey orders including component sourcing, plan for 2–3 weeks end to end; quick-turn prototype assembly takes 48 hours once boards and parts are ready.

Have your files ready — or not sure they're complete? Send what you have and our engineers will tell you exactly what's missing.

Related reading: PCB vs PCBA: What's the Difference? · Key Questions to Ask Before Outsourcing PCBA

Prototyping in Shenzhen is fast when the data package is complete and slow when it is not. The delays that hurt a first build are almost never machine time — they are a missing drill file, an ambiguous part number, or a question that sat unanswered overnight because nobody on your side had authority to answer it. This guide covers what a prototype PCBA supplier actually needs, where the calendar days genuinely go, why first articles get rejected, and what to test at this stage instead of over-testing a design that is still moving.

What do you need to prepare before ordering a prototype PCBA?

Split your preparation into three categories: files (machine-readable data), decisions (choices only you can make), and information (context the factory cannot infer). Missing files stop the quote. Missing decisions stop the build after the quote — which is worse, because you have already burned a week feeling productive.

If you want the file-level detail — what a usable BOM column set looks like, which Gerber layers to export — our guide to the files required for a PCBA quotation goes line by line.

What does a prototype PCBA lead time actually consist of?

"Two weeks" is not one process; it is four, and only one of them is assembly. Understanding the split tells you where to push and where pushing achieves nothing. For turnkey PCB assembly at BELI Technologies, quick-turn prototype assembly takes 48 hours once boards and components are on hand, and turnkey orders including component sourcing typically run two to three weeks end to end. Here is where those weeks go.

Read that table once and the strategy is obvious: the schedule is set by your BOM and your response speed, not by the SMT line. A team that answers a DFM question in two hours and pre-approves alternates will beat a team with an identical design by a week.

Why do first articles get sent back? The five recurring causes

Prototype rejections tend to cluster into a short list, and almost all of them are data problems rather than workmanship problems. Each one is worth checking against your own package before you release it.

• Footprint and land pattern errors. A part that does not match its pad geometry cannot be rescued on the line. This is what a proper design-for-manufacturability review is for, and it is why we run DFM on every order rather than as a paid extra.
• Hole and clearance geometry that only fails at the next assembly step. The board passes electrical test, then will not accept a connector or a standoff. We documented exactly this in a Gerber hole issue we fixed for a US sensor client, where holes specified at 0.8 mm were opened to 1.2 mm and the corrected boards shipped in 48 hours with zero rework.
• Polarity and orientation. Silkscreen says one thing, the centroid file says another. When the two disagree, the machine follows the file.
• Unapproved substitutions. A part was out of stock, someone fitted an equivalent, and nobody told the designer it had a different tolerance or thermal rating. Fix this by writing your alternates policy into the order rather than leaving it to goodwill.
• Mismatched expectations on test and cosmetics. A prototype built to functional-sample expectations arrives with minor cosmetic marks; the customer expected production finish. Define the acceptance criteria in advance and this disappears.

Notice that four of the five are settled before any machine starts. The prototype stage rewards paperwork discipline more than it rewards factory selection.

What should you test at the prototype stage — and what should you not?

Over-testing a moving design wastes money; under-testing it wastes a build cycle. The rule of thumb: at prototype, test what tells you whether the design is right. Defer anything that tests whether the process is repeatable, because there is no stable process yet.

If you want the full staging logic, the EVT, DVT and PVT build stages lay out which validation belongs to which gate. And when the design does freeze, functional testing and ICT pick up where flying probe leaves off, with node coverage of 95% or better on a purpose-built fixture.

Why prototype and production unit prices cannot be compared directly

Nearly every hardware team does this once: receives a prototype quote, divides by the quantity, multiplies by the production forecast, and panics. The arithmetic is meaningless, because a prototype price is dominated by costs that do not scale with quantity.

Setup work — stencil, machine programming, feeder loading, first-article check, engineering time — is charged once whether you build 5 boards or 5,000. Bare-board fabrication at prototype quantities cannot fill a production panel, so you pay for panel area you do not use. Components bought in reel-break quantities cost far more per piece than components bought on full reels, and some suppliers will not break a reel at all. None of these disappear at volume; they are simply divided across far more units.

The practical consequence: use prototype quotes to compare suppliers on responsiveness and engineering quality, and ask separately for an indicative volume structure to compare on unit economics. The mechanics of the split are set out in prototype versus mass production cost, and the wider cost drivers in what PCBA actually costs in China. What you should never do is choose a long-term manufacturing partner because their prototype was slightly cheaper — the sample is not the product, and the quote is not the price.

Turnkey or consignment for a first prototype run?

Consignment — you buy and ship the parts — makes sense when you already hold inventory, when a part is on allocation and you have a distributor relationship the factory does not, or when a specific lot must be used. It also puts every purchasing error on your side of the line.

Turnkey means the factory quotes the BOM, buys the parts, and owns the shortage risk. For a first build this usually wins, because one missing low-value passive holds the whole kit hostage just as effectively as a missing microcontroller. We handle electronic component sourcing through authorized distributors and original manufacturers with full traceability and original packaging preserved, with 100% visual inspection plus X-ray sampling on incoming parts — which matters more at prototype than people assume, because a counterfeit or out-of-spec part in a ten-board build is indistinguishable from a design fault until you have wasted a week chasing it. Both models are supported with no fixed ratio, so a hybrid — you consign the two parts you already have, we buy the rest — is entirely normal.

How do you work across the time-zone gap without losing a day per question?

Shenzhen runs at UTC+8. That is roughly 15–16 hours ahead of US Pacific, 12–13 ahead of US Eastern, 6–7 ahead of Central Europe, and 5–6 ahead of Israel, depending on daylight saving. The consequence is not that communication is hard; it is that each unanswered question costs a full calendar day, and the cost is invisible until you add it up at the end of the project.

Three habits remove most of that loss. First, batch: when a DFM report arrives, resolve every point in one reply rather than sending three separate answers over three days. Second, pre-authorize: decide in advance who may approve an alternate part or a minor geometry change, and say so in writing, so the factory does not stop and wait for a decision that was never contentious. Third, give conditional answers — "if the 0.8 mm hole is a problem, open it to 1.2 mm and proceed" is worth a day compared with "please advise". Europe and Israel additionally have a genuine overlap window with the Chinese afternoon; teams in those regions should schedule the one weekly call inside it rather than at their own convenience. US teams do not have overlap, so the discipline matters more: an answer written before you finish your working day is waiting when the factory opens.

How many prototype boards should you build?

More than you think, but not many more. Building only the units you plan to bring up leaves no margin for a board destroyed during debug, none for destructive cross-sectioning if a joint looks suspicious, and none to ship to a colleague in another location. Because there is no MOQ for prototypes, the marginal cost of a few extra assembled boards is small next to the setup cost you have already paid — and small next to the two weeks it takes to get more. Ordering a handful of extra bare boards is cheaper still, and gives you spares for a partial rebuild if a component substitution needs evaluating.

Does the prototype supplier need to be the production supplier?

Not strictly — but changing suppliers between prototype and volume means repeating work you have already paid for: a new DFM cycle, new stencils and programs, new fixtures, and a new learning curve on your board's quirks. It also removes the one thing that makes a prototype genuinely predictive, which is that the same processes and inspection criteria will be used at volume.

The practical test when choosing a prototype partner is whether they can still build your board at 100 times the quantity. Ask about board-size limits, the smallest components and BGA pitches the lines handle, layer count, and whether inspection at volume includes AOI, X-ray and ICT as standard. Our full process window is on the manufacturing capability page, and boards up to 42 layers from PCB fabrication run into the same assembly lines that hold first-pass yield at 99.5% or better in production.

A realistic first-build sequence

Put together, a well-run first prototype looks like this: send a complete package and get a quote within a day, with DFM review included; resolve every DFM point in one exchange; approve the BOM including any alternates; bare boards fabricate while parts are purchased; assembly runs in 48 hours once both are in-house; AOI and X-ray inspection before packing, plus flying-probe test if you asked for it at quotation; express shipment lands in 3–7 days. Two to three weeks, most of it purchasing, almost none of it machine time — and the parts you control are the parts that move the date.

If you have a board ready to build, send us your Gerber, BOM and pick-and-place files and you will get a quote within 24 hours, with DFM review included on every order — plus a straight view of which BOM lines are going to set your schedule. There is no minimum order for prototypes, and the same lines that build your first ten boards will build your first ten thousand.

Related reading: Low-MOQ and Small-Batch PCB Assembly · PCB Prototype vs Mass Production Cost · What Files Are Required for a PCBA Quotation

"No MOQ" is one of the most repeated phrases in contract electronics, and one of the least examined. Engineers who take it at face value are surprised when a 20-board build still requires buying 5,000 resistors, or when a 30 mm square board gets quoted as a panel of twelve. None of that contradicts a no-MOQ policy — the assembler genuinely has no quantity floor, but several upstream and downstream steps do. This guide separates the constraints that are real from the ones that are negotiable, and shows how to structure a small run so it does not become an expensive dead end.

What Does "No MOQ" Actually Mean at a PCBA Factory?

Minimum order quantity is a commercial policy, not a physical limit. Volume-focused factories set high MOQs because every changeover costs them line time: feeders are unloaded and reloaded, a new stencil is mounted, the placement program is called up, and the first boards through are consumed proving the setup. If a line is booked with six-figure runs, a 25-board job is a net loss of capacity.

A prototype-friendly EMS makes a different trade. At BELI Technologies there is no MOQ for prototypes, and the same SMT lines carry a job from a handful of boards through to mass production — capacity across those lines runs to roughly 52 million placements per month, so a small job runs on the same equipment as volume work rather than on a separate prototype cell. The practical value of that is not the low quantity itself. It is that the stencil, the placement program, the inspection criteria and the test coverage you prove on 20 boards are the same ones that will build 20,000, so nothing needs requalifying when you scale. Our turnkey PCB assembly service is built around that continuity.

What "no MOQ" does not mean: that a small build costs the same per unit as a large one, that every component can be bought in the exact quantity you need, or that a board of any outline can be run as-is. Those are separate constraints, covered below.

Low Volume vs Low MOQ: Are They the Same Thing?

These two terms get used interchangeably and they are not the same. The distinction matters when you are comparing suppliers.

Low MOQ describes the supplier's policy — the smallest order they will accept. It is a threshold question with a yes or no answer.

Low volume describes your production regime — you build in the tens or low hundreds, repeatedly, as an ongoing business, not as a one-off validation exercise. Industrial instrumentation, medical devices, broadcast equipment, laboratory hardware and capital-equipment sub-assemblies live here permanently.

A supplier can have a low MOQ and still be a poor low-volume partner: happy to take your 50-board order once, but treating every repeat as a new setup, with no retained programs, no held fixtures, and no stable pricing. Conversely, a supplier with a nominal MOQ of 100 units may be an excellent low-volume partner if they retain your stencils and test fixtures, keep the placement program on file, and re-run the job in days.

Small batch is the loosest of the three and generally means a build somewhere between prototype and production — a pilot run, a DVT or PVT build, a first commercial batch. If you build in this range, ask about repeat-order handling and tooling retention, not just about the minimum. See EVT, DVT and PVT build stages for how these builds are normally sequenced.

How Cost Structure Shifts Across Quantity Tiers

The most useful thing to understand about small-batch pricing is that the composition of the invoice changes with quantity, not just the total. The table below describes that structural shift. The relative index in the last column is illustrative and structural, not BELI pricing — it exists to show the shape of the curve, and real numbers depend entirely on your BOM, board and test scope.

Two readings follow from this. First, the steepest part of the curve is between single-digit and low-hundred quantities — which is exactly where most validation builds sit, and exactly why a prototype quote looks alarming next to a projected production cost. Second, above roughly a thousand units, further savings come from the BOM rather than from the assembly line, which is a different negotiation entirely. The five cost blocks behind these numbers are broken out in how a China PCBA quote is built.

Why NRE Weighs So Heavily on a Small Batch

Non-recurring engineering is the set of costs incurred once per design, independent of quantity: stencil fabrication (one per soldered side), placement program creation, feeder setup and line changeover, DFM engineering, first-article inspection, and any dedicated test fixture. Build five boards and you carry the whole of that on five units. Build five thousand and the same total lands on five thousand.

Three things follow that are worth acting on. NRE is charged per design revision, not per order, so three respins in a quarter means paying it three times — batching design changes into fewer revisions is real money. It is also charged per side for stencils, so a layout that keeps everything on one side of the board is cheaper to set up as well as faster to build. And it should always appear as a separate line on the quotation; if a supplier folds NRE into the unit price, a small-quantity quote becomes impossible to compare against anyone else's. The full arithmetic of NRE amortisation, worked through with clearly labelled illustrative figures, is in prototype vs mass production cost.

Which Process Steps Carry a Minimum, and Which Do Not

This is the table most engineers actually need. "MOQ" is not one number; it is a set of independent constraints, each with its own workaround.

Note the pattern: almost nothing on the assembly side imposes a quantity floor. Two entries do constrain you — the 45 x 45 mm minimum board size, which is a machine-handling reality, and component packaging, which belongs to the distributor rather than the factory. On the test side, the choice is economic: BELI builds custom ICT and functional fixtures in 2 to 3 working days with ICT node coverage of 95% or better, but for a genuinely small batch, flying-probe testing avoids the fixture cost entirely. Details of both routes are on the functional testing and ICT page.

MPQ: The Component Minimum That Really Sets Your Floor

Minimum packaging quantity (sometimes standard packing quantity) is the smallest unit a distributor will sell a part in. It is set by how the manufacturer packages the component, and it is the single most common reason a small build costs more than expected. Typical figures, which vary by manufacturer and distributor:

Two practical consequences. First, a 0402 capacitor whose per-piece cost is negligible still carries a full-reel minimum, so a 20-board build can end up holding thousands of spares on dozens of line items — normal, and rarely worth fighting. Second, packaging format affects assembly, not just purchasing: tape-and-reel feeds fastest, tubes and trays need different feeders, and bulk parts usually mean hand placement. When a BOM is quoted, ask which lines are only available in bulk or in awkward pack sizes, because those are the lines that drive setup time up on a small run. BELI's component sourcing from authorized distributors and original manufacturers covers standard parts in 3 to 7 days, with 100% visual inspection plus X-ray sampling on incoming material, and consignment or turnkey with no fixed ratio — so you can hand over the awkward lines and buy the rest yourself if that suits your cash flow.

End-of-Life Risk Bites Harder in Small Batches

Low-volume products often stay in production for years, which means their BOMs outlive several component life cycles. The risk profile is the opposite of a high-volume consumer product: you are not exposed to allocation during a demand spike, you are exposed to quiet obsolescence between orders. A part that was fine on last year's build is not stocked this year, and because you buy in small quantities you have no leverage and no contractual supply.

Three habits keep this manageable. Run a lifecycle check on the BOM at every build, not only at design time, so you catch a not-recommended-for-new-designs status while alternatives still exist. Qualify a second source for anything critical during the design phase, when a footprint change is free. And when a part does go end-of-life, decide deliberately between a last-time buy and a redesign rather than drifting into a shortage. BELI's purchasing team flags risky lines during BOM review and can reach the spot market for hard-to-find or end-of-life components; the approach to substitutions, always with written client approval, is described in cutting BOM costs without cutting corners.

Panel Utilisation: How Your Board Outline Changes the Bill

Bare boards are not manufactured individually — they are imaged, drilled and plated on a production panel and separated afterwards. BELI fabricates on panels up to 18 by 24 inches, and the cost that matters is how much of that panel your board actually uses. An outline that tiles efficiently might yield forty boards per panel; the same area with an awkward shape or an unnecessarily generous outline might yield twenty-eight. At volume that difference is a line item in a cost-reduction review. On a small batch it can decide whether you pay for one panel or two.

Panelisation also solves the minimum board size problem. Boards below 45 x 45 mm cannot be handled individually on the SMT line, so they are assembled in an array with breakaway rails carrying tooling holes and fiducials, then depanelised by V-score or routed tabs after assembly. Design the array early: the rails need enough width for conveyor edge clearance, and components too close to a V-score line risk cracking at depanelisation. A DFM review — included with every order at BELI, prototype or production — catches these before boards are cut, alongside the usual footprint and annular-ring checks. Materials, finishes and stack-up options for the bare boards themselves are on the PCB fabrication service page.

Common Small-Batch Mistakes Worth Avoiding

• Ordering exactly the quantity you need. Build in a small overage. Boards get destroyed in bring-up, cut for cross-sections, sent to a test lab, or sacrificed to a firmware bug. Adding 10% to a 30-board run costs little; a second run costs the whole NRE again.
• Comparing a turnkey quote to a consignment quote. One includes components and one does not. This is the single most common apples-to-oranges error in supplier comparison.
• Sending an incomplete file package. Missing manufacturer part numbers, no pick-and-place file, or an unstated board revision costs days of clarification. Delays hurt small runs disproportionately, because the schedule, not the unit price, is usually why you chose a small run.
• Changing the design between quote and build. A revision after quoting can invalidate the stencil, the program and part of the BOM. Freeze, then order.
• Choosing a prototype-only shop. If your assembler cannot scale, you pay for requalification later — new supplier, new fixtures, new yield ramp — at the moment you can least afford the delay.

How to Structure a Small Batch So It Scales

Treat the first small build as the beginning of a production process rather than a one-off purchase. Concretely, that means agreeing four things with your assembler up front: that the pilot runs on the same lines and processes as future volume; that stencils, fixtures and placement programs are retained and that you own the tooling you paid for; that first-pass yield and a defect breakdown are reported after each run, not just a pass or fail; and that any component substitution requires your written approval.

Timing is the other half. Quick-turn prototype assembly at BELI takes 48 hours once boards and components are in hand, while a turnkey order that includes sourcing typically runs 2 to 3 weeks — the difference is procurement, not the line. Bare boards move faster still: as quick as 24 hours for 1 to 2 layer boards and 3 to 7 working days for multilayer. Plan your build calendar around the sourcing window, because that is what actually sets the date. Small batches run under the same controls as volume work, with AOI and X-ray on every order and first-pass yield held at 99.5% or better.

If you have a low-MOQ or small-batch build coming up, send us your Gerber, BOM and pick-and-place files. We sign an NDA before you share anything, return an itemised quotation within 24 hours, and include a DFM review with every order — and the same lines that build your 20-board pilot will build your production run.

Related reading: Prototype PCBA in China: A Practical Guide · Turnkey PCB Assembly for Startups

But not all EMS partners are created equal. Choosing the wrong one can lead to quality failures, communication breakdowns, IP concerns, and costly delays. Choosing the right one can accelerate your time to market, reduce unit costs, and give you a reliable production backbone for years to come.

This guide walks you through the key factors to evaluate when selecting an EMS partner in China — whether you are a startup shipping your first product or an established OEM diversifying your supply chain.

Understand What EMS Actually Covers

Before evaluating partners, it helps to clarify what EMS means in practice. An EMS provider is not just a factory that assembles boards. A full-service EMS partner typically offers:

• PCB fabrication — manufacturing bare printed circuit boards to your specifications
• Component sourcing — procuring electronic parts from authorized distributors or qualified alternates
• SMT and through-hole assembly — populating and soldering components onto the PCB
• Testing and inspection — AOI, X-ray, ICT, functional testing, and burn-in
• Box build and final assembly — enclosure assembly, cable harness integration, and packaging
• Engineering support — DFM review, BOM optimization, and test fixture development

The more of these services your partner can handle under one roof, the fewer coordination gaps you will encounter. A true one-stop EMS partner eliminates the need to manage multiple vendors across different stages of production.

Evaluate Technical Capabilities and Equipment

Not every EMS provider can handle every type of product. Before you engage, assess whether the facility has the right equipment and experience for your specific requirements.

Key questions to ask include:

• What SMT line equipment do they use? High-speed pick-and-place machines from brands like Yamaha, Fuji, or Panasonic indicate serious investment in precision.
• Can they handle fine-pitch components, BGAs, QFNs, and 01005 passives?
• Do they have X-ray inspection capability for hidden solder joints?
• What PCB layer counts and materials can they fabricate? Can they support HDI, flex, or rigid-flex boards?
• Do they offer conformal coating, potting, or other protective processes for harsh-environment products?

At BELI Technologies, our SMT lines place components at 0.15 seconds per chip with ±0.03 mm precision — down to 01005 passives and 0.4 mm-pitch BGAs — backed by SPI, AOI, and X-ray inspection, supporting everything from simple 2-layer boards to complex 20+ layer HDI assemblies. See our full manufacturing capability for the detailed specifications.

Assess Quality Systems and Certifications

Quality is non-negotiable in electronics manufacturing. A credible EMS partner should hold relevant certifications and demonstrate a mature quality management system.

Look for the following:

• ISO 9001 — the baseline for quality management across all manufacturing
• ISO 13485 — essential if you are building medical devices
• IATF 16949 — required for automotive electronics
• IPC-A-610 — the industry standard for acceptability of electronic assemblies
• RoHS and REACH compliance — mandatory for products sold in the EU and increasingly expected worldwide
• UL certification — important for safety-critical products

For reference, BELI's own manufacturing is certified to ISO 9001:2015, ISO 14001, and ISO 13485:2016; we are an IPC member, and our processes are RoHS and REACH compliant.

Beyond certifications, ask to see their defect tracking process, corrective action procedures, and first-pass yield data. A factory that openly shares quality metrics is one that takes quality seriously.

Prioritize Communication and Responsiveness

This is where many EMS relationships fail — not because of technical capability, but because of poor communication. When your partner is 8,000 miles away and 12 time zones ahead, communication discipline becomes critical.

Here is what to look for:

• Dedicated project manager — you should have a single point of contact who understands your project inside and out, not a rotating support desk.
• English proficiency — your primary contact should be fluent in English, capable of discussing technical details without ambiguity.
• Responsiveness — a good EMS partner should respond to emails within 24 hours, even if just to acknowledge receipt.
• Proactive updates — the best partners do not wait for you to ask. They flag potential issues, share production photos, and keep you informed at every stage.

At BELI Technologies, every client is assigned a dedicated English-speaking project engineer who manages the project from quotation through delivery. We believe clear, proactive communication is the foundation of every successful manufacturing partnership.

Consider Scalability and Flexibility

Your production needs will change over time. A good EMS partner should be able to grow with you — from initial prototypes of 5-10 units, through pilot runs of a few hundred, to volume production of tens of thousands per month.

Evaluate whether the partner can:

• Handle both low-volume, high-mix orders and high-volume production runs
• Ramp production up or down based on your demand forecast
• Store safety stock of critical components to buffer against supply chain disruptions
• Support multiple product lines simultaneously without quality trade-offs

Flexibility also means the willingness to accommodate engineering changes mid-production, manage revision control, and adapt processes as your product evolves. A partner that is rigid on process may be efficient at scale but painful during the iterative early stages of a product lifecycle.

Choosing an EMS partner is not just a procurement decision — it is a strategic one. The right partner becomes an extension of your engineering team, helping you navigate supply chain challenges, optimize your design for manufacturing, and deliver a quality product on time and on budget.

If you are evaluating EMS options in China, BELI Technologies offers a free project review and quotation. We will assess your design files, recommend optimizations, and provide a transparent cost breakdown — so you can make an informed decision with zero obligation.

Contact BELI Technologies

Related reading: How to Choose a Reliable PCBA Manufacturer in China · Key Questions to Ask Before Outsourcing PCBA

Whether you are a startup building your first proof of concept or an established company launching a new product line, understanding the cost breakdown between prototyping and mass production is essential for making smart financial decisions.

Why Prototypes Cost More Per Unit

PCB manufacturing involves significant fixed costs that are amortized over the total number of boards produced. When you order 5 prototype boards instead of 5,000 production boards, those fixed costs are spread across far fewer units — dramatically increasing the per-unit price.

The major fixed costs include:

• Tooling and setup — every new PCB design requires creating photo tools, stencils, and configuring the production line. Whether you make 1 board or 10,000, this setup cost is roughly the same.
• Engineering review — DFM checks, Gerber file verification, and impedance calculations happen once per design, regardless of quantity.
• SMT programming — the pick-and-place machine must be programmed with component coordinates, rotation, and feeder assignments for each unique design.
• Test fixture development — functional test jigs and ICT fixtures are designed per product, not per unit.

For a prototype run of 10 units, you might pay $50-100 per board. The same board at 10,000 units might cost $3-8 per board. The board itself has not changed — but the economics of spreading fixed costs have.

Component Costs: Volume Pricing Makes a Big Difference

Electronic components are priced on steep volume curves. A microcontroller that costs $4.50 in single quantities might drop to $1.80 at 1,000 units and $0.95 at 10,000 units. This pattern applies across virtually every component category — resistors, capacitors, connectors, ICs, and sensors.

During prototyping, you typically buy components in small quantities from distributors like Digi-Key or Mouser at the highest price tier. At production volumes, your EMS partner can leverage bulk purchasing, long-term supply agreements, and alternative sourcing strategies to secure significantly lower pricing.

At BELI Technologies, component cost optimization is a core part of our component sourcing service. During the transition from prototype to production, we conduct a detailed BOM review to identify cost reduction opportunities — including qualified alternative components, packaging optimizations, and consolidated purchasing across multiple client projects.

PCB Fabrication: Panelization and Yield

PCB fabrication costs also benefit significantly from scale. Here is why:

• Panelization — at production volumes, multiple individual boards are arranged on a single manufacturing panel. This maximizes material utilization and reduces per-board fabrication cost. Prototype orders often cannot be panelized efficiently, resulting in more material waste.
• Material optimization — production runs allow the fabricator to select optimal panel sizes and layer stackups, reducing raw material cost per board.
• Yield improvement — production processes are tuned and optimized over the run. First-article inspection catches issues early, and process adjustments improve yield as the run progresses.

A 4-layer prototype board might cost $25-40 per unit in quantities of 5-10. The same board in production quantities of 5,000+ might cost $2-5 per unit — a reduction of 80-90%.

Assembly Costs: Setup Time vs Run Time

SMT assembly costs follow a similar pattern. The majority of the cost for a small prototype run is setup time — loading feeders, calibrating the placement machine, running the first board, and performing initial inspection. Once the line is running, the incremental cost per board drops significantly.

Consider this simplified comparison:

• Prototype (10 units) — 2 hours of setup time, 30 minutes of actual run time. Setup cost dominates.
• Production (5,000 units) — 2 hours of setup time, 20 hours of run time. Setup cost becomes negligible per unit.

Testing costs follow the same logic. A manual functional test on 10 prototype boards is expensive per unit because the test procedure is developed and executed individually. At production scale, automated test systems amortize development cost across thousands of units and execute each test in seconds.

How to Budget Effectively Across Both Phases

Understanding these cost dynamics helps you plan your budget realistically:

• Expect prototype costs to be 5-20x higher per unit than production pricing. This is normal and not a sign that you are being overcharged.
• Do not over-order prototypes. Order only what you need for validation and testing. The goal of prototyping is to verify your design, not to produce inventory.
• Plan for a pilot run. Before committing to full production, consider a pilot run of 50-200 units. This intermediate step validates your production process, catches remaining issues, and provides units for field testing — at a per-unit cost between prototype and production pricing.
• Optimize your BOM before production. Work with your EMS partner to review component selections for cost, availability, and alternative options. Small changes to passive component values, connector brands, or IC packages can yield significant savings at volume.
• Design for manufacturing from the start. Following DFM guidelines during the design phase reduces production costs by minimizing rework, improving yield, and enabling efficient panelization.

At BELI Technologies, we guide clients through every stage — from rapid prototyping with fast turnaround, through pilot production, to full-scale manufacturing. We provide transparent pricing at each stage so there are no surprises, and we actively help optimize costs as you scale.

Ready to understand the true cost of your next PCB project? Send us your Gerber files and BOM for a free, detailed quotation covering both prototype and production scenarios.

Get a Free Quote from BELI Technologies

Related reading: How Much Does PCBA Cost in China · Prototype PCBA in China: A Practical Guide

The difference between a successful outsourcing experience and a painful one often comes down to the questions you ask before signing a contract. Too many companies choose a PCBA supplier based on price alone, only to discover hidden costs in rework, delays, and quality failures.

Here are the critical questions you should ask any potential PCBA partner — and why each one matters.

Questions About Manufacturing Capabilities

Understanding what a manufacturer can and cannot do is the foundation of a successful partnership. Not every PCBA provider is equipped to handle every type of product.

Ask these questions:

• What SMT component sizes can you place? If your design uses 01005 passives or micro-BGAs with 0.3mm pitch, you need a manufacturer with the equipment and experience to handle them reliably.
• What is your maximum PCB layer count? If you are designing complex multilayer boards (8+ layers, HDI, blind/buried vias), confirm the manufacturer can fabricate them in-house or has a trusted fabrication partner.
• Do you support mixed-technology assembly? Many products require both SMT and through-hole components on the same board. Confirm the manufacturer can handle both in a single production flow.
• What testing capabilities do you have? AOI and visual inspection are baseline. For production quality, you should expect X-ray inspection for BGAs, in-circuit testing (ICT), and functional testing capabilities.
• Can you handle box build and final assembly? If your product requires enclosure assembly, cable integration, labeling, or packaging, a manufacturer that offers these services eliminates the need for a separate assembly house.

Questions About Quality and Process Control

Quality problems in electronics manufacturing are expensive to fix and can damage your brand reputation. Understanding a supplier's quality systems before you commit is essential.

Key questions include:

• What certifications do you hold? ISO 9001 is the minimum. Depending on your industry, you may also need ISO 13485 (medical), IATF 16949 (automotive), or AS9100 (aerospace).
• What is your first-pass yield rate? A manufacturer with a consistently high first-pass yield (above 99%) demonstrates strong process control. Ask for actual data, not just claims.
• How do you handle defects? Ask about their corrective action process. Do they perform root cause analysis? Do they track defect trends? A mature quality system prevents recurring issues rather than just fixing them one at a time.
• Do you perform DFM review before production? A proactive DFM review catches design issues that could cause manufacturing problems — before they result in defective boards. This service should be standard, not an upsell.
• Can you provide inspection reports and traceability data? For regulated industries, traceability from component lot numbers to finished assemblies is not optional. Even for commercial products, traceability data helps diagnose field failures.

Questions About Intellectual Property Protection

IP protection is a legitimate concern when outsourcing manufacturing, particularly to overseas partners. While the vast majority of professional manufacturers operate ethically, it is reasonable to seek assurances.

• Will you sign a Non-Disclosure Agreement? Any reputable manufacturer should be willing to sign an NDA before you share design files. If they hesitate, consider it a red flag.
• How do you control access to client design files? Ask about their internal data security practices. Are Gerber files and BOMs stored on access-controlled systems? Who has access?
• Do you manufacture for my competitors? This is a reasonable question, especially for niche markets. While a manufacturer may not be able to disclose client names, they should be willing to confirm they have policies to prevent conflicts of interest.
• Can I retain ownership of custom tooling and test fixtures? If your project requires custom stencils, test jigs, or molds, clarify ownership upfront. You should own anything you paid to develop.

At BELI Technologies, we take IP protection seriously. Every client engagement begins with a signed NDA, and all design files are stored on access-restricted servers with strict internal controls.

Questions About Lead Time and Logistics

Delivery reliability can make or break your product launch. Understanding realistic lead times and logistics capabilities is critical for planning.

• What is your typical lead time for prototype orders? For prototypes, 5-10 business days after component procurement is standard. If a manufacturer quotes significantly longer, their production scheduling may not prioritize small orders.
• What is your production lead time at volume? Typical PCBA production lead times range from 2-4 weeks depending on complexity and component availability. Get a realistic estimate based on your specific BOM.
• How do you handle component shortages? Supply chain disruptions are a reality. Ask how the manufacturer handles situations where a component on your BOM becomes unavailable. Do they have alternative sourcing strategies? Will they notify you before making any substitutions?
• What shipping methods do you support? Understand the options for air freight, sea freight, and express courier — along with the cost and transit time for each. Also ask about customs documentation and export compliance support.

Questions About Communication and Project Management

Poor communication is the number one complaint companies have about overseas manufacturing partners. Establishing clear communication expectations from the start prevents most problems.

• Will I have a dedicated project contact? You should have a single point of contact who knows your project, not a generic support email. This person should be reachable during your business hours at least part of the day.
• What language does your team communicate in? English fluency — particularly technical English — is essential for discussing engineering details without miscommunication.
• How often will you provide production updates? Expect updates at key milestones: component procurement complete, SMT assembly started, testing complete, shipment dispatched. The best partners provide photo documentation at each stage.
• What happens if there is a problem during production? Ask how they escalate issues. Will they stop the line and notify you, or proceed and inform you after the fact? The answer reveals a lot about their values.

Asking these questions before you commit to a PCBA partner is not about being difficult — it is about building a manufacturing relationship on a foundation of transparency, capability, and mutual respect. The best suppliers welcome tough questions because they are confident in their answers.

At BELI Technologies, we encourage prospective clients to ask every question on this list — and more. We believe the right manufacturing partnership starts with complete transparency. Contact us today for a no-obligation project discussion, and see how we answer these questions for your specific product.

Start a Conversation with BELI Technologies

Related reading: How to Choose the Right EMS Partner in China · What Files Are Required for PCBA Quotation

An Industry-Wide Storm

Between 2020 and 2023, the global electronics industry faced a supply chain disruption unlike anything seen in modern manufacturing history. A perfect storm of pandemic-driven factory shutdowns, surging consumer demand, geopolitical tensions, and logistics bottlenecks converged simultaneously — sending component lead times soaring and prices through the roof.

Connectors, microcontrollers, passives, and discrete semiconductors that had been reliably available for cents were suddenly backordered for 52+ weeks or priced at 10× their pre-crisis cost. For our clients — OEMs building industrial controllers, medical devices, and smart consumer hardware — this was a direct threat to their ability to ship product.

We Saw It Coming — And We Acted

As a PCBA manufacturing partner deeply embedded in global supply chains, our team at BELI Technologies began seeing warning signs early. Lead times on key components started stretching. Allocations were tightening. Spot-market prices were climbing.

Rather than wait for clients to come to us in crisis, we proactively initiated 1-on-1 project reviews across our active client base. For each project, we analyzed the BOM line by line, flagged components at risk of severe price increases or supply disruption, and began researching qualified alternatives — before clients even asked.

Our Approach: 1-on-1 BOM Optimization

Our methodology was deliberate and client-first. We didn't apply a blanket substitution policy across all projects. Instead, every review was tailored to the specific project, application, and risk tolerance of that client.

Quality Chinese Alternatives: A Better-Kept Secret

One of the most impactful areas of substitution was connectors — a component category where Western brands like Molex, TE Connectivity, and JST had seen dramatic price spikes and lead time extensions during the shortage.

What many product companies outside China don't fully appreciate is that the Chinese connector manufacturing ecosystem has matured enormously over the past decade. Several domestic manufacturers now produce connectors functionally equivalent to their Western counterparts — same pin pitch, contact plating, current ratings — while meeting RoHS, REACH, and UL certifications.

Our Commitment: Transparency First

We want to be clear: we never made component substitutions without telling our clients. This is a principle we hold firmly, and it distinguishes a true manufacturing partner from a contract house that optimizes its own margin at the client's expense.

Every proposed change was documented, shared with the client's engineering team, and approved in writing before any production action was taken.

The Results

Across clients who participated in the BOM optimization program, we consistently achieved 30–50% reductions in per-unit component cost compared to what the same components would have cost at peak spot-market pricing.

Just as importantly, clients were able to keep production moving. During a period when many companies were halting production lines due to component unavailability, our clients with optimized BOMs continued shipping — in some cases, the difference between maintaining customer commitments and losing key accounts.

What This Means Going Forward

The acute component shortage has eased, but the lessons remain. Global supply chains are not as stable as they once appeared. A single geopolitical event, a factory fire, or a demand spike in an adjacent industry can rapidly change the availability and pricing of components your product depends on.

The best time to build a supply chain resilience strategy is before you need one. At BELI Technologies, we now include a proactive BOM risk review as part of our onboarding for all new clients. We maintain a continuously updated database of qualified alternatives for the most common components in our production — so when market conditions shift, we can respond within days, not months.

If you are designing a new product or reviewing an existing BOM, we'd be glad to run a complimentary cost-optimization review and show you where there may be opportunities to reduce risk and save cost — without compromising quality.

Related reading: How Much Does PCBA Cost in China · PCB Prototype vs Mass Production Cost Breakdown

Almost every PCBA supplier in Shenzhen publishes the same page: SMT lines, AOI, X-ray, ISO 9001, "advanced equipment", "strict quality control". Those pages are not useful for selection because they are indistinguishable from each other. What separates a reliable manufacturer from a risky one is not what they claim — it is what they can produce as evidence when you ask, and how they behave when something goes wrong.

This guide is written for the procurement or engineering lead who has to run that evaluation. It is deliberately about verification. If you are still at the stage of drafting your enquiry, our companion articles cover the questions to ask before outsourcing PCBA and how EMS engagement models differ in China. This one assumes you already have three quotes in hand and need to work out which of them is real.

What "Reliable" Actually Means in a PCBA Supplier

Reliability in contract manufacturing is not the absence of problems. Every programme hits a shortage, a marginal solder joint, or a design file that does not match the BOM. Reliability is the supplier's ability to detect the problem before you do, tell you about it unprompted, and correct it without silently changing your product.

That reduces to four measurable dimensions, which map to the four sections below: real line capability, a quality system that is actually operated rather than framed, a component channel with traceability, and engineering support that engages with your files rather than just processing them.

The Supplier Evaluation Checklist: Evidence and Red Flags

Send this table to each shortlisted supplier as a written request. The speed and specificity of the reply is itself a data point — a factory that routinely handles export customers should be able to answer most of it without escalating internally.

How to Verify Manufacturing Capability Instead of Reading the Spec Sheet

Published specifications describe the best case a factory has ever achieved, not what it does routinely. Three checks convert a spec sheet into something you can plan against.

First, ask for the numbers that bound your design specifically. If your board carries a 0.4 mm pitch BGA, the relevant question is not "can you do BGA" but "what is your minimum BGA ball pitch, your placement accuracy, and do you X-ray every BGA or a sample". For reference, our own PCB assembly lines place components down to 0201/01005 and BGAs down to 0.4 mm ball pitch at ±0.03 mm placement precision, on boards from 45 × 45 mm up to 680 × 500 mm, with capacity of roughly 52 million placements per month — the point is that those are the four figures a supplier should be able to state without checking.

Second, separate fabrication capability from assembly capability. Many "PCBA manufacturers" broker the bare board to a third-party fab. That is not automatically bad, but it changes who owns a stack-up or impedance problem. Ask directly whether the bare board is made in-house, and if not, who makes it. Our PCB fabrication side runs to 42 layers with impedance control at ±10%, and our full manufacturing capability page lists the fabrication and SMT parameters side by side so an engineer can check a design against them in one pass.

Third, check whether the capability is reachable at your volume. A factory geared to 100k-unit runs will quote your 30-piece pilot, then schedule it behind everything else. Ask what the MOQ is, what a prototype turnaround looks like, and whether prototype and production run on the same lines. Quick-turn prototype assembly at BELI takes 48 hours once boards and components are ready, with no MOQ for prototypes; a turnkey order where we also source the parts typically takes 2 to 3 weeks end to end.

What Each Certification Means — and What It Does Not

Certifications are widely misread. They are evidence that a management system exists and was audited; they are not a warranty on your specific board. The distinction matters when you are comparing two suppliers who both list the same logos.

For transparency: BELI's manufacturing is certified to ISO 9001:2015, ISO 14001 and ISO 13485:2016, we are an IPC member, and our processes are RoHS- and REACH-compliant. Apply the same "what it does not mean" column to us as to anyone else — ask for the certificate, check the scope line, check the expiry.

How to Read IPC-A-610 Class 2 vs Class 3

This is the single most commonly fudged item in Chinese PCBA quotations, and it is worth understanding properly because it directly changes cost, yield and rework rules.

Three practical rules follow. One: a class only becomes binding when it appears on the purchase order or the agreed drawing — a sentence on a supplier's website is not a contract term. Two: do not order Class 3 reflexively. It costs more and will not compensate for a design that is not qualified for it; the correct time to decide is at the DVT stage, which we cover in the EVT, DVT and PVT build-stage guide. Three: ask who does the inspecting and to what training level, because the standard is only as good as the operator applying it. Where a documented class already applies to a whole product family it should be stated plainly — our wire harness assemblies, for example, are built and inspected to IPC/WHMA-A-620 Class 2 or Class 3 with 100% continuity testing on every unit.

How to Verify the Component Sourcing Channel

Counterfeit and mis-graded parts remain the highest-consequence failure mode in China PCBA, because they usually pass functional test and fail in the field. The verification here is concrete.

• Ask for the source per line item, not a general statement. A quotation that names the channel for each part is a different document from one that says "authorized distributors" at the bottom.
• Compare against public distributor pricing. If a quoted unit price for a branded MCU is well below the authorized distributor price at your quantity, the part is coming from somewhere else. Ask where.
• Require original packaging and certificates of conformance for the parts that carry the most risk — programmable logic, MCUs, memory, RF front ends, power semiconductors.
• Ask what incoming inspection actually happens. "We check the parts" is not an answer. A specific answer sounds like: 100% visual inspection plus X-ray sampling, which is the standard we apply on component sourcing, with traceability and original packaging maintained through to kitting.
• Get the substitution rule in writing. The only acceptable version is that no alternate part reaches the line without your prior written approval.

Also ask what happens when a part is genuinely unavailable. A supplier with real purchasing depth will describe both routes rather than simply pushing the delay back to you. At BELI, standard parts come from authorized distributors and original manufacturers and typically arrive in 3 to 7 days, while hard-to-find or end-of-life items are chased on the spot market. If you would rather carry the channel risk yourself, check that consignment is supported; we run consignment and turnkey with no fixed ratio, and the PCBA cost guide explains which risks move with which model.

How to Test Engineering Support Before You Place an Order

Engineering support is the cheapest thing to claim and the easiest thing to test, because you can test it for free during quotation. Send your real Gerber and BOM and read what comes back.

A superficial DFM reply restates your own parameters. A substantive one raises specifics: annular ring or drill-size margins, solder mask web widths between fine-pitch pads, thermal relief on large copper pours, panelisation and tooling-strip proposals, silkscreen clashing with pads, parts on the BOM that are NRND or have a long lead time. If nothing comes back but a price, you have learned that DFM is not part of that supplier's process — at BELI, DFM analysis is included with every fabrication order rather than sold as an extra.

Two further probes are worth running. Ask what happens if their DFM finds an issue after you have already approved for production — a real answer describes stopping and asking, not proceeding. And ask for a worked example of a design problem they caught and fixed; our own Gerber hole-size case study describes a US sensor client whose through-holes were specified at 0.8 mm, too tight for assembly, corrected to 1.2 mm and returned within 48 hours with zero rework. Whatever example a supplier gives you, it should contain a root cause, a specific change and a time.

If a supplier struggles to tell you which files they need, that is also diagnostic. Our note on the files required for a PCBA quotation sets out what a complete package looks like, and a capable manufacturer will tell you exactly what is missing rather than quoting around the gap.

Quality Data: What to Ask For and How to Read It

Any supplier can say "high quality". Ask for numbers with definitions attached, because the definitions are where the meaning lives.

• First-pass yield, and how it is calculated. Yield measured after touch-up is not first-pass yield. Ask at which test station it is measured. BELI reports first-pass yield of ≥99.5% on assembly.
• Test coverage, not just test names. "We do ICT" is incomplete without node coverage; our functional test and ICT service targets ≥95% node coverage, with custom fixtures built in 2 to 3 working days.
• Whether reports ship with the goods. Test records that exist only on request tend not to exist. Ours are delivered with every shipment.
• What reliability testing is available in-house if your product needs it. At BELI that means burn-in for up to 24 hours and environmental stress screening from −40 to +85 °C at 20–98% RH.
• The corrective-action format. Ask to see a blank or redacted 8D report. A factory that cannot produce one does not run a closed-loop quality process, whatever the certificate says.

Red Flags That Should End the Evaluation

• Refusal to sign an NDA before receiving design files.
• A quotation with no line-item breakdown, so you cannot see whether the saving is in board, parts, labour or test.
• Certificates supplied as website screenshots rather than PDFs with certificate numbers you can check with the issuing body.
• A price materially below every other quote with no explanation of where the difference comes from.
• Reluctance to name the factory or allow a video walkthrough of the line.
• Answers that change between emails — particularly on lead time, MOQ and acceptance class.
• No named engineer on the account, only a sales alias.

A Three-Step Qualification Sequence That Works

Rather than trying to decide everything at once, stage the risk.

Step 1 — paper audit (about one week). Issue the checklist table above to three suppliers, request certificates with scope, and send your Gerber and BOM for DFM. Score the replies on specificity and turnaround, not on price. Most shortlists reduce themselves at this stage.

Step 2 — sample build (one to three weeks). Order a small prototype batch from the two survivors, with the acceptance class written on the PO. Judge the boards, but judge the paperwork too: did the test report arrive, is traceability there, were questions raised before production rather than after?

Step 3 — pilot run, then release. Only after a sample build behaves as promised should you commit a pilot volume, and only after the pilot should you release recurring production. Suppliers that pass steps 1 and 2 but degrade at volume usually do so because the pilot was scheduled on a different line — which is exactly why you ask, at step 1, whether prototypes and production run on the same lines.

Where BELI Technologies Fits

BELI Technologies is a one-stop EMS provider in Shenzhen, Guangdong, China, backed by manufacturing facilities with over 15 years of operation, serving hardware teams and OEMs primarily in the United States, the European Union, the United Kingdom, Israel, the wider Middle East and Asia. We cover fabrication, component sourcing, SMT and through-hole assembly, wire harness and functional test under one roof, which removes the hand-offs where responsibility usually gets lost.

We would rather be evaluated against the checklist above than against a brochure. Send us your Gerber, BOM and pick-and-place files — the same package any supplier needs in order to quote — and our engineers will run a DFM and BOM review before quoting, flagging footprint issues, sourcing risks and long-lead parts up front. Run the same request past two other suppliers at the same time and compare what comes back. Contact BELI Technologies to start that comparison, or send your questions to our engineering team first if you would like the checklist adapted to your specific product class.

Request a DFM Review and Quotation

Related reading: Key Questions to Ask Before Outsourcing PCBA · How to Choose the Right EMS Partner in China · EVT, DVT, PVT: Hardware Build Stages and Your EMS

The Challenge

When one of our U.S.-based smart sensor clients reached out, they had a frustrating problem — one that only became apparent after production had already shipped. Their engineering team was struggling to mount boards during final assembly, and the root cause traced back to the original Gerber files: the through-holes were simply too small.

The undersized holes made it physically difficult to insert and secure components. Technicians were forced to apply excessive force, increasing the risk of board damage, component misalignment, and production delays. What should have been a smooth assembly process had turned into a time-consuming bottleneck — costing real money on the production floor.

Root Cause Analysis

Upon receiving the client's Gerber files, our engineering team performed a thorough Design for Manufacturability (DFM) review — the same review that is included with every BELI fabrication order. The issue was clear: the original drill hole diameters had been specified without accounting for the standard insertion tolerances required for the connectors used during assembly in the U.S.

When hole diameters are undersized relative to the component lead diameter — even by fractions of a millimeter — assembly workers face significant resistance during insertion. This not only slows down the production line but also creates the very real risk of PCB pad lifting, lead deformation, or cracked solder joints.

Our Engineering Solution

Once the root cause was confirmed, our team acted swiftly. The fix involved two complementary changes to the Gerber files and BOM that, together, completely eliminated the assembly issue.

The Result

The outcome was immediate and measurable. With the revised Gerber files and updated component specification, the client's assembly team in the United States was able to mount the boards smoothly and without any specialized tooling or excess force.

Beyond resolving the immediate issue, our DFM review also surfaced several additional areas for layout improvement, which we documented in a detailed engineering report. The client now submits all new designs through our DFM review process before tape-out — helping them catch issues upstream, before they become production-floor problems.

Key Takeaways

This case is a great example of why DFM (Design for Manufacturability) review is a critical part of any PCB production workflow — especially when designs cross borders and are assembled in different facilities with different tooling setups.

A few millimeters in your Gerber file can translate to hours of production delays and thousands of dollars in rework costs downstream. At BELI Technologies, we don't just manufacture boards — we act as an extension of our clients' engineering teams, proactively identifying and solving issues before they ship.

Have a design you'd like checked before production? Send us your Gerber and BOM for a free DFM review — we respond within 24 hours.

Related reading: What Files Are Required for PCBA Quotation · How We Cut Client BOM Costs Without Cutting Corners