DEV Community: Xingliao Smart Home

The Power Cord Is a Certified Part, Not an Accessory

Xingliao Smart Home — Mon, 02 Nov 2026 06:50:36 +0000

A buyer orders one product for five countries and expects one carton. Then the first shipment lands and four markets come back with the same complaint: the plug does not fit.

The failure is not that anyone forgot. It is that the power cord was treated as a packing item when it is really a certified component, and it was specified at the end of the project when it has to be decided at the start.

Plug shape and voltage are two separate problems

The first thing to separate is the two things everyone merges.

Plug shape is about the socket in the wall. It varies by country, and a few countries have their own regardless of their neighbours: the UK, Brazil, South Africa, India, Switzerland, Italy.

Voltage and frequency are about what the device can accept. Most modern switch-mode supplies are labelled for a wide input range, and those genuinely do run on both. Heater elements and motors usually do not, and they are single-voltage parts.

So the same device can be electrically fine in five countries and still be unsellable in four of them because nobody ordered the right cord.

The two are independent, and "100-240V" on the label answers only the second one.

The cord carries its own approval

Here is the part that catches people.

A detachable power cord is not covered by the approval of the device it plugs into. The cord is a component with its own approval, and the combination is what gets tested. When a product ships with a cord, the cord's approval and the moulded plug's approval are part of the file.

That is why a product approved with a European cord is not automatically approved with a British one. It is not a paperwork technicality. The plug differs, the cord standard differs, and the test file has to name what was actually tested.

It is also why the fix that seems obvious is the one that creates the problem. A distributor who cannot sell the fitted cord and swaps it for a local one has changed a certified part, and the approval no longer describes the product in the box.

Fixed or detachable changes the answer

Which cord arrangement you choose is a design decision with regulatory consequences.

A fixed, captive cord is part of the appliance. It is tested as part of the product, and it cannot be swapped by anyone downstream, which is exactly the protection you want. It also means the product SKU has to be built for its market, because you cannot change the cord at the destination.

A detachable cord with a standard inlet, such as an IEC C13 or a figure-8, separates the two approvals. The device is approved for the inlet, and the cord is approved as a cord. This is the arrangement that lets one product serve many markets, because the local cord can be fitted in the destination country or packed at the factory.

The cost of the second approach is the inlet itself and the volume behind it. The benefit is that you stop shipping five variants of the same housing.

The markets that need their own approvals

Then there are the countries where the cord is the smallest of the problems.

Brazil wants INMETRO. India wants BIS registration. Korea wants KC. Russia and Eurasia want EAC. These are national approvals held by a local entity, and they are a different class of thing from CE or FCC. They are usually applied for by the importer, not the factory, and they take months.

A supplier's honest answer here is narrow. For a smart switch or a doorbell, standard support typically covers CE, FCC, RoHS, UKCA, CCC and PSE. That list has apparent breadth and one conspicuous gap: it does not include UL or ETL, which is what the North American market actually asks for on electrical safety.

So the right way to handle a market outside that list is not to assume coverage. It is to ask, per market, whether there is a live file, and to treat "we can support the testing" as a different sentence from "we hold the approval."

What to put in the RFQ

Four lines, and they belong in the first email rather than after the sample.

The target market list, given before sampling and not after. This single change moves every downstream decision, because plug type, cord approval, label format and manual language all derive from it.

Plug type and cord length, named per market. "EU plug" is not a specification; there are several, and the length is a separate choice that affects both cost and the carton.

Whether the cord is fixed or detachable, and if detachable, which inlet. This decides whether market variants are five SKUs or one SKU with five cords.

Nameplate and manual localisation. The rating label has to show the voltage and frequency for the market, and several markets require the manual in the local language before the product can be sold.

Why this saves more than it costs

Every one of these is a decision that costs almost nothing at the tooling stage and costs a reprint, a re-test, or a rejected shipment later.

The pattern that produces the most damage is a buyer who specifies the product in detail, approves the sample, places the order, and only then mentions the countries. At that point the cord is the least of it, because the label, the carton, the manual and possibly the certification file were all built for an answer nobody had given yet.

It is a small subject with an outsized bill attached to it, and it is settled in one email if it is sent early enough.

Sourcing references: XingLiao ODM/ODM and the manufacturing programmes.

The Mould Is the Only Part of the Tooling Bill That Anyone Mentions

Xingliao Smart Home — Mon, 02 Nov 2026 06:45:25 +0000

Ask a buyer what an ODM project costs to start and the answer is usually one number: the mould. It is the only tooling item that gets a price in the first email, and it is not the largest one. It is simply the only one that shows up as a line.

Everything else that makes a new product exist is paid for somewhere else, and most of it is paid without being named.

What actually sits inside non-recurring cost

Non-recurring engineering, or NRE, is whatever you pay once so that a product can exist at all. On a smart home device it usually holds five things.

The mould, which everyone expects.

The PCB layout and the firmware work that make the electronics match the enclosure you just tooled. They are not the same team and they are not the same invoice.

Certification testing. Not the certificate you already have for a standard platform, but the test reports and filings for this product in these markets. Each market is a separate file with its own fee and its own lab slot.

Packaging tooling: printing plates, die-cut forms for the insert, and the artwork that has to be laid out to the printing house's template rather than to your brand book.

Production aids: jigs, test fixtures, and the first article run that proves the line can hold tolerance before you commit to volume.

None of these is optional. Several of them are invisible on a quote.

The two ways factories bill it

There are only two patterns, and they behave very differently for you.

In the first, NRE is billed separately and up front. You see the number, you pay it, and the unit price afterwards looks attractively low. This is honest and it is also the version that surprises buyers who budgeted only for the mould.

In the second, NRE is amortised into the unit price. There is no tooling line at all, and the price per piece is slightly higher than it would otherwise be.

The second pattern is not a trick, but it does two things you should notice. It makes the unit price depend on the order size, because the factory has to recover a fixed amount across however many pieces you actually buy. And it means the low-MOQ price and the volume price differ for a reason that has nothing to do with the parts.

If you are quoted only a unit price, ask what is inside it. If you are quoted only a mould, ask what is not.

Why this decides whether you can leave

Here is the part that rarely gets said in a sourcing conversation.

NRE is sunk cost. Once it is paid, it does not follow you. If you move the product to another factory in year two, you pay the mould again, the layout and firmware work again, the certification for the new production site again, and the production aids again.

That is the real reason brands stay with a factory they are unhappy with. Not loyalty, and not the inconvenience of moving. The reason is that moving means paying the entire non-recurring bill a second time for a product you have already paid to create once.

So the question to settle before you sign is not how much the tooling costs. It is what you own afterwards.

What to settle in writing

Four things, and they are all contract questions rather than engineering ones.

Mould ownership and release. You can pay for a mould and still not be allowed to take it. A usable clause says the tool is yours, held at the factory for your programmes, and released to you or to a nominated mover if the relationship ends.

Responsibility for modification. A mould will be modified. The first shot rarely gives the fit you wanted on a glass panel or a clip. Say now who pays for revisions, and how many are included, because this is where a fixed tooling quote becomes an open one.

Certification file ownership. Test reports can be held in the factory's name, which means the approval does not travel with you either. If you need the approval in your name, say so at quotation time and not after the test is booked.

What happens to the amortised version. If NRE is inside the unit price, ask whether there is a settling point: a volume at which the tooling is considered paid and the price is reviewed. Without that sentence, the amortisation never ends.

What this tells you about the cheap quote

Two quotes for the same product can differ by a few percent on unit price and by a large multiple on what you will ultimately spend, and the quote will not show it.

The low unit price with no separate tooling usually means the recovery is spread across a volume you have not committed to yet, and a price that only works if you reorder at scale. The separate-tooling quote is more expensive in month one and cheaper by the third order.

Neither is wrong. They are bets on different futures, and the only way to read them is to know which one you are being offered.

The shortest useful version of the question

Before you compare two suppliers, send both the same four lines and make them answer in writing.

Is NRE billed separately or inside the unit price, and what exactly is in it.
Who owns the mould, and what releases it.
Who pays for tooling modification, and how many rounds.
Whose name is on the certification files, and what a second production site would cost.

Four answers tell you more about the next three years than the unit price does.

Sourcing references: XingLiao ODM/ODM and the manufacturing programmes.

Ask How Long the Platform Is Meant to Live, Not Just What It Costs

Xingliao Smart Home — Mon, 02 Nov 2026 06:39:45 +0000

Products stop being made, and not always on purpose

Every electronic product sits on top of components with their own life cycles. A wireless module, a display driver, a specific connector, a particular power IC — each of those has a manufacturer who will eventually end it. When that happens, the product that used it does not stop selling. It stops being makeable.

This is the least discussed and most expensive thing in a sourcing conversation, and it is invisible at quotation time because on the day you ask, everything is available.

There are two kinds of end. The planned kind: the component maker issues an end-of-life notice with a last-time-buy window, and there is a date. The unplanned kind: the part simply becomes unobtainable, usually because demand moved and the line was redeployed, and nobody sent anyone a notice.

The planned kind is manageable if somebody is reading the notices. The unplanned kind is what turns into a redesign under time pressure.

What happens downstream

When a component ends, there is a substitution. A substitute that is truly identical is rare; usually it is equivalent with different firmware behaviour, a different footprint, or a marginally different electrical characteristic.

That is where the three hidden costs land. Firmware work: the new part needs a driver, and the firmware that used the old one needs to be revalidated. Documentation: the certification file covers a construction, so a substitution changes what the file covers, and some markets require the shipped model to appear on it. And requalification: the product has to be retested, at least partially, which costs time more than money.

The part that hurts most is that this happens on a product that is selling. The redesign is not a project you scheduled, it is an interruption to one that is running.

Why it matters more for some buyers than others

A distributor moving consumer volume through retail can treat a model change as normal: the old SKU ends, a new one replaces it, the listing changes.

A project buyer cannot. If the devices are installed in buildings — in apartment blocks, hotels, a senior-care facility — the requirement is not "the current model," it is "the same model as the three hundred already on the walls." A revision that changes the physical interface, the protocol behaviour or the spare parts means the installed base is now two different products, and the maintenance contract has become complicated.

This is the buyer for whom a platform's remaining life is a real line in the specification, and it is the question almost nobody asks.

What "lifetime support" has to mean to be worth anything

Any supplier can say they will support a product. The useful version has three parts and all three are specific.

Spare parts: for how long after the last production run, and whether spares means complete units or service parts. Some support arrangements keep finished units, which is what a building owner actually needs; others keep components, which is what a repair shop needs.

The substitution policy: when a part ends, does the buyer get told, and is the replacement validated as the same product or sold as a revision. A revision that changes the model number is a different answer from one that does not.

And the notice itself: a last-time-buy window is a date. Whether anyone is watching for it is the practical question.

Stated plainly for us: we give a one-year warranty with lifetime support across 172 production platforms, and NDA as the default on ODM work. The second half is the part that matters — lifetime support means spares and technical answers remain available after the warranty window closes, which is what an installed base actually consumes.

The honest limit

No supplier can promise that a component will never end. The supply chain is not theirs. What can be promised, and should be asked for, is the process: whether end-of-life notices are monitored, whether the buyer is informed before a substitution happens, and whether the substitute is presented as the same product or as a change.

There is also a design-side answer that is more robust than any promise. Platforms built on widely-used, second-sourced parts survive longer than platforms built on a single attractive component, because the substitution has somewhere to go. That is a choice made at the design stage, and it is worth asking about when the product being sourced is going into buildings.

Five questions for the RFQ

How long this platform is planned to remain in production. Whether spares after the last run are finished units or parts. What happens when a component is discontinued — notice, or a change discovered on arrival. Whether a substitution keeps the model number, which decides whether the installed base stays one product. And whether the firmware is maintained for the life of the product or only until the next one ships.

None of these add cost to negotiate. They convert a future interruption into a known number, which is the only form of it that can be planned around.

Sourcing references: XingLiao ODM/ODM and the ODM programmes.

The Quote Is Not the Landed Cost, and the Gap Has a Name

Xingliao Smart Home — Mon, 02 Nov 2026 06:34:34 +0000

Three letters decide who pays for what

A factory quote arrives as a unit price and three letters. EXW, FOB, CIF. Those letters allocate cost and risk between two parties, and the allocation is the part buyers read too fast.

EXW means the goods are available at the factory door. Everything after that — loading, inland haulage, export clearance, freight, insurance, import clearance, duties, delivery — is the buyer's. A supplier quoting EXW is not being unhelpful, they are pricing exactly what they control. What it means in practice for a buyer without their own freight forwarder in the origin country is that they are taking on a job they cannot see.

FOB means the seller gets the goods on board the vessel at the named port, and clears export. From the moment the cargo is over the rail, it is the buyer's. This is the one most buyers think they understand, and it is the one with the sharpest edge: under FOB the buyer books the vessel. If the buyer's nominated forwarder is slow, or the space is not there, the goods sit at the port and storage starts accruing, and the seller is entitled to be paid.

CIF means the seller also pays freight and insurance to the destination port. Which sounds like the buyer has bought an all-in price, and is where the expensive misunderstanding lives.

What CIF actually covers

CIF covers carriage and minimum insurance to the destination port. Not to the buyer's warehouse, not through customs, and not including duties or taxes. The cargo arrives at a port, and from there it is the buyer's problem exactly as under FOB — just later, and with less visibility, because the buyer did not book the shipment and may not know which vessel it is on.

The insurance under CIF is minimum cover, typically the lowest institute cargo clause. It is not full-peril cover. If the claim is for something outside that minimum, the buyer discovers the gap at the moment they need it.

There is also the cost you cannot see. A CIF price has the seller's freight and insurance inside it, priced by whoever the seller uses. Buyers with a good freight contract frequently pay less under FOB with their own forwarder than under CIF with the factory's — and they get to see the booking.

The costs that appear after the price is agreed

A handful of items are missing from almost every first RFQ, and they are the ones that turn a good unit price into a mediocre landed cost.

Origin charges: inland transport to the port, terminal handling, documentation, export customs. Under FOB these are inside the seller's number. Under EXW they are all yours, and in some origins they are not small.

Destination charges, which are the ones that surprise people most: terminal handling at arrival, customs clearance, duties, import VAT or GST, and the delivery leg to the warehouse. Duty rates are set by the importing country and the classification of the product — the same product can carry a different duty rate in two markets. No factory anywhere can tell you what your duty is going to be.

Then the timing costs. Demurrage and detention when containers sit too long. Storage when production finishes before the vessel does. These are not mistakes, they are the normal consequence of two calendars that were never synchronised.

Why the unit price comparison is misleading

Two suppliers quoting the same product at 10.00 and 10.40 are not necessarily 4% apart. If the first is EXW and the second is FOB, the comparison depends entirely on what the origin costs turn out to be. If one quote includes packaging the buyer will have to redo, or a certificate the buyer will have to re-obtain, the gap moves again.

The honest way to compare is to convert every quote to the same basis before comparing: same incoterm, same port, same packaging specification, same certificate scope, same tooling amortisation. Doing that arithmetic usually reorders the list.

What to settle before agreeing the number

Four lines. The incoterm and the named place — not "FOB China" but the port, because inland cost to one port can differ from another. Who books the vessel. What the quote explicitly excludes. And who owns the risk at each handover point, which matters most when something goes wrong in transit and both parties discover they thought it was the other one's.

There is also a practical sequence worth knowing: a buyer who does not have a forwarder in the origin country, and has never imported before, is usually better served by asking for CIF to their nearest port than by nominating FOB and inheriting an unfamiliar job. The CIF price is higher and the surprises are fewer. Later, with volume and a forwarder relationship, FOB becomes the cheaper and more controllable option.

Where this sits against the other costs

Incoterms are one of four groups that separate a unit price from a landed cost: the freight term, the tooling amortisation, the certification file, and the payment structure. They are decided in different documents by different people inside the same order, which is exactly why they are optimised separately and discovered together.

Pricing the whole chain once, in one conversation, is worth more than another 2% off the unit price.

Sourcing references: XingLiao ODM/ODM and manufacturing capability.

A Passed AQL Inspection Does Not Mean the Shipment Is Good

Xingliao Smart Home — Mon, 02 Nov 2026 06:29:23 +0000

AQL is an acceptance number, not a quality standard

AQL stands for acceptable quality limit, and the phrase itself contains the misunderstanding. It is not a promise that a shipment is good. It is the defect rate at which a sampling plan will still accept the lot. A common general inspection level with AQL 2.5 for major defects means a shipment can contain a certain number of defective units and still pass on paper.

That is not a loophole, it is the design. Sampling is a statistical compromise: nobody opens ten thousand cartons, so you inspect a random sample and accept the lot if the sample's defects fall under a threshold. The inspection tells you the lot was not obviously worse than the threshold on the day it was checked. It does not tell you the lot was good, and it says nothing at all about the units that were not opened.

The two different things buyers confuse

The first is a factory audit. It looks at the supplier: capability, equipment, process control, traceability, whether they own the line or subcontract it. It answers "can this plant make this repeatedly."

The second is a pre-shipment inspection. It looks at one batch: cartons are opened at random, units are counted, checked against the specification, checked for workmanship and function, sometimes drop-tested. It answers "is this particular batch acceptable."

They are bought by different people for different reasons and they do not substitute for each other. A plant with an excellent audit record can produce a bad batch. A batch that passes inspection can come from a plant you would not want to depend on. When a buyer says "we did the inspection," the question is which of the two happened.

What the inspection cannot see

Sampling happens at one moment, on finished goods, usually packed. There are whole classes of failure it structurally cannot catch.

Failures that need time. A contact that frets after two thousand insertions, a capacitor that drifts after three hundred hours at temperature, a plastic part that crazes under UV on a south-facing wall. These are the failures that create warranty cost, and none of them exist yet when the cartons are opened.

Failures that need the installed environment. A panel that behaves differently when it is screwed into a slightly bowed box, or in a bathroom with condensation, is not a workmanship defect the inspector can see. The wrongness only appears in the wall.

Consistency across batches. A single inspection says this lot is acceptable. It does not say the next lot will match it, which is why the sample retention conversation matters more than the inspection report.

Inspection is a cost line, and it belongs in the quote

Pre-shipment inspection costs money and somebody has to pay for it. On a small order the fixed cost of a man-day is a meaningful percentage of the order value, which is exactly why first orders skip it and exactly when it is most useful.

There are three workable arrangements and they should be settled before the purchase order, not after. Buyer arranges and pays: highest confidence, and the report belongs to the buyer. Factory arranges and pays: cheaper for the buyer, and the report belongs to the factory. Third, and most useful for a first order: the buyer specifies the level and the acceptance criteria in the order, and the factory's internal QC reports against those criteria, with the buyer reserving the right to send an independent inspector.

The third one only works if the criteria are written down. "Inspect before shipment" is not a specification. Level, AQL per defect class, and what counts as critical versus major versus minor is a specification.

Four things to put in the order

The inspection level and the AQL for each defect class — critical is normally zero, not 2.5. The definition of a critical defect for this product, in words: wrong plug type for the market, missing earthing, a relay that will not break the rated load. What happens on a failed inspection: rework and re-inspect, or rejection, and who pays the delay. And who keeps the retained samples, for how long.

The last one is the cheapest and most skipped. A retained unit from the accepted lot, kept by both sides, is what makes a later argument about finish or feel resolvable instead of being two people describing the same object from memory.

Why this is a commercial question, not a quality-control one

Because the result of getting it wrong is not a defective shipment, it is a surprise. A brand that has never seen a 1.5% field failure rate will meet it for the first time in the second month after launch, at which point the cost is not the units, it is the returns handling, the marketplace ratings, and the relationship with the distributor.

The right frame for a first order is not "is this supplier good." It is "what will I do when a batch is bad, and can I prove it." A written inspection specification plus a retained sample is most of that answer, and it costs one paragraph in a purchase order.

Sourcing references: XingLiao ODM/ODM and the manufacturing programmes.

The Cheapest Line on the Quote Is the One That Decides Your Warranty Cost

Xingliao Smart Home — Mon, 02 Nov 2026 05:35:09 +0000

What a warranty actually costs

Every quote carries a warranty line, usually one sentence, and nobody negotiates it. It reads like a promise. In practice it is a cost model, and the model is decided by three numbers that rarely appear in the document: the failure rate, who pays the freight on a return, and how many spare units exist.

A one-year warranty on a product with a 0.5% field failure rate is a rounding error. The same warranty at 3% is a line item that can exceed the profit on the order. The difference between those two numbers is not luck — it is decided during design and assembly, and the buyer has almost no visibility into it.

The DOA question and the spare parts clause

Two mechanisms exist and they do the same job in different ways. A spare parts clause ships a small percentage of extra units with the order — commonly around 1% — so the distributor can swap a failed unit immediately without an international return. A DOA credit reimburses the value of units that are dead on arrival or fail within a short window.

The spare parts route is faster for the end customer and cheaper per event, because no one ships anything back across an ocean to resolve a 20-dollar failure. The credit route looks better on paper and is slower in practice, because it requires proof, paperwork, and usually a threshold before anything is settled.

For a distributor with an end customer waiting, the 1% spares clause is almost always the better deal, and it is the one buyers forget to ask for.

Who pays the freight is the whole argument

A warranty that says "we will replace it" is incomplete. Replace it from where, with shipping paid by whom, and within how many days of the claim? A unit that costs 25 dollars does not survive a 40-dollar return shipping cost, and the only honest resolution in that situation is to replace locally from spares and scrap the failed unit.

This is why the warranty clause should be read together with the incoterm. If goods ship FOB, the buyer owns the freight from the port onward, including the return leg. A warranty that sounds generous on FOB terms can be worthless in practice because the return costs more than the part.

What the failure rate is made of

Most field failures are not random. They cluster into three groups, and each has a signature.

Early failures, in the first days, are assembly: a connector not seated, a solder joint, a screw that backed out during transport. These show up as DOA and they are the reason a burn-in period matters.

Mid-life failures, in the first months, are component selection running closer to its limit than the datasheet suggested — a capacitor at the edge of its temperature rating in a sealed enclosure, a relay switching a load type it was rated for on paper only.

Late failures are environmental: moisture ingress in a bathroom installation, thermal cycling on a panel that never gets turned off, contact fretting on a socket that gets plugged and unplugged daily.

A factory that tests for the first group and not the second will quote a lower price and a higher real warranty cost.

What we state and what we actually do

We give a one-year warranty with lifetime support, 172 production platforms across 16 categories, and NDA as the default on ODM work. Stated plainly, because the useful part is the second half: lifetime support means spare parts and technical answers remain available after the warranty window, which is what actually matters when a product has been installed in a building for six years and one component needs replacing.

The honest limit is that a warranty cannot fix a specification error. If the load type was wrong, if the installation environment was outside what the product was designed for, or if the market's electrical standard differs from what was stated at quotation, no warranty clause covers that. Those are settled at the specification stage.

Four lines to settle before signing

What the warranty period is and what it excludes, in words. Whether spares ship with the order and at what percentage. Who pays freight on a return, and whether local scrap-and-replace is accepted instead. And what the response time on a technical claim is, because a warranty that takes eight weeks to answer is a warranty in name only.

None of these add cost to negotiate. They just move the cost from a surprise into a number you can price.

Sourcing references: XingLiao ODM/ODM and manufacturing programmes.

Your Packaging Decision Is a Freight Decision

Xingliao Smart Home — Mon, 02 Nov 2026 05:29:59 +0000

The box is priced by volume, not by weight

Most buyers treat packaging as a marketing item that gets settled after the product is finished. It is not. It is the single variable that decides how much air you ship, and on consumer electronics air is usually what you are paying for.

Freight is billed on chargeable weight: the greater of the actual weight and the volumetric weight. A light product in a slightly oversized carton is billed as if it weighed several times more. A switch that weighs 200 grams in a box that is 4 centimetres too tall across a pallet can cost more to move than the difference between two suppliers' unit prices.

This is why the packaging drawing belongs in the cost conversation, not in the artwork conversation. A two-centimetre reduction on one dimension, multiplied across a container, is a real number, and it is a number that repeats on every reorder for the life of the product.

Custom packaging has its own minimum, and it fights your first order

On our ladder, custom packaging starts at 500–1,000 pieces. Standard models start at 500 pieces per colour. Those two numbers are the same order of magnitude, which is the problem: on a first order the packaging minimum and the product minimum arrive together, and if you are testing the market with 500 units you are committing to a full print run before you know whether the product sells.

There is a cheaper path and it is not a compromise. Plain white box plus a printed label or a sleeve gives you branded retail presence with a fraction of the minimum and a fraction of the lead time. You move to a printed carton at the second or third order, when the volume justifies the run and the artwork has stopped changing.

The buyers who insist on a full colour carton at 500 pieces usually discover two things: the unit cost of the box is high at that quantity, and the artwork is wrong by the time the second order arrives.

Lead time lives in the box too

Sampling runs 7–15 days and tooling 25–40 days, and those numbers describe the product. The packaging has its own clock: artwork approval, plate or plate-less digital setup, print run, and then assembly. A print house will not start without approved artwork, and artwork revisions are the most common cause of a shipment slipping after the product itself is ready.

Practically this means packaging artwork should be locked while the product is still in sampling — not after the confirmation sample is approved. Doing it in parallel removes a week or two that otherwise lands at the worst possible moment.

Three things that fail on the packing line

Packaging failures are almost never aesthetic. They are mechanical. The carton closes in the design file and does not close on the line because the cable bundle is thicker than the drawing assumed. The insert holds the product in the render and lets it shift in the drop test. And the master carton quantity does not divide the pallet, so the last layer is half full and ships badly.

None of these show up in a photograph of a nice box. All of them show up in a damage rate three months after launch.

What packaging actually has to do

Four jobs, in order: survive the drop and vibration test for the channel it sells through, hold the product and its accessories without movement, present the brand at shelf or in the listing photograph, and carry the compliance markings the destination market requires.

That last one is the one people forget. Country of origin, disposal symbol, importer address, rating labels, and in some markets a local-language requirement. The markings have to be on the box, and they have to be correct for each destination, which is another reason to keep the first run simple: a label is easy to change per market, a printed carton is not.

Where this sits against the other cost lines

Packaging is one of four costs that quietly decide landed cost, together with the tooling amortisation, the certification file, and the freight mode. Buyers optimise the unit price and then discover the box, the mould, the file and the air freight. None of those are hidden — they are just decided in different documents by different people.

Settling packaging early, in the same conversation as MOQ and lead time, is the cheapest way to avoid a container of product arriving with a carton that costs more to ship than it did to print.

Sourcing references: XingLiao ODM/ODM and custom programmes.

Payment Terms Are Financing, Not Paperwork

Xingliao Smart Home — Mon, 02 Nov 2026 05:24:48 +0000

The line everyone negotiates last

Payment terms usually get settled at the end of a negotiation, after the price, the MOQ and the delivery date. It is treated as administration. It is not. It is the largest financing decision in the order, and the party who does not notice that is the party who funds the other one for free.

A deposit is not a formality. It is working capital. On a tooling project the deposit is what pays for steel. On a packaging order it is what pays the print house, who will not start without it. On a first order with a new supplier it is the only thing standing between you and a factory that has already bought your materials against a promise.

What the common structures actually do

A deposit against balance before shipment is the default for a reason: it splits the risk roughly in half. The factory carries the material risk, you carry the finished-goods risk, and neither side is financing the whole cycle. Typical structures land around a third down, the rest against bill of lading or before loading, and the exact split is a negotiation about who trusts whom more at that point in the relationship.

Open account — net 30, net 60 — is not cheaper, it is a loan. Someone has to fund the gap between materials leaving the factory and money arriving, and if that someone is the factory, the cost of that loan is inside your unit price. Buyers who push hard for long terms on small volumes should expect the quote to move, because the factory is pricing the credit, not the product.

A letter of credit moves risk to the bank and charges for it. On small orders the bank fees are a meaningful percentage, and the documentary requirements are strict enough that a single mismatch in a date or a description can delay payment for weeks. It makes sense above a certain order size and for a first transaction with an unknown counterparty. Below that, it costs more than the risk it removes.

Why terms and lead time are the same conversation

The money and the calendar are linked. Sampling is 7–15 days, tooling 25–40 days, and sea freight from Shenzhen is typically 30–40 days to Europe, 7–10 by air. A full cycle from deposit to shelf is easily ninety days or more for a custom project.

That gap is the reason the deposit exists, and it is also the reason a buyer's own cash cycle matters. If you are selling through a channel that pays you sixty days after delivery, and your supplier wants the balance before loading, you are carrying a gap of roughly five months on a product you have not sold yet. That is a real cost and it belongs in your landed cost, not in your margin surprise.

The honest version of what we can say

We work on T/T: deposit to start production, balance before shipment, with the exact split given with the quotation. That is the standard and it is what a factory of our size can actually offer. Payment terms beyond that are a function of order history, not of negotiation skill — they widen after a few clean orders, not before.

Stating this early saves everyone time. A buyer who needs ninety-day terms on a first order of 500 pieces is asking a manufacturer to be a bank, and the answer will either be no or will be a price that includes the loan.

Four questions that make the terms concrete

What is the deposit percentage and exactly which event triggers it — signing, or material purchase, or tooling start. What triggers the balance: completion, inspection pass, or loading. What happens if the shipment slips by two weeks — does the balance date move, and does storage start costing money. And what currency the price is in, because on a ninety-day cycle the exchange movement can exceed the margin you negotiated so hard on.

The currency question is the one most often skipped and most often expensive.

What terms do not fix

Good terms do not fix a bad sample, an unclear certificate holder, or a tooling clause that leaves the mould with the factory. Those are settled in the technical documents. Payment terms only decide who funds the interval and who carries which slice of risk while the order moves.

Buyers who win at this are not the ones who extracted the longest terms. They are the ones who matched the payment structure to their own cash cycle and kept the relationship clean enough for the terms to widen on their own.

Sourcing references: XingLiao ODM/ODM and manufacturing capability.

Three Kinds of Sample, and Only One of Them Settles a Dispute

Xingliao Smart Home — Mon, 02 Nov 2026 05:19:37 +0000

"Sample" is not one object

Buyers ask for a sample as if there is one thing with that name. On the factory side there are at least three, and they are built by different people for different purposes. Confusing them is the single most common reason a bulk shipment gets argued about with no way to resolve it.

The first is an engineering sample. It proves the circuit and the firmware work on a bench. It may be hand-assembled, it may have a 3D-printed housing, and it may be built from parts that are not the production parts. It answers "does this idea function."

The second is a confirmation sample. It is built from production tooling, with production parts, on the line that will run your order. It answers "is this what I will receive."

The third is a shipment sample, sometimes called a golden sample or a sealed sample. It is pulled from the actual production run and kept, sealed, by both sides. It answers "was this shipment what we agreed."

Only the third one settles a dispute. The first two are for making decisions, and neither of them is evidence.

Why the first one is dangerous if you stop there

An engineering sample is optimised to look good. It is usually a hand-picked unit, tested on a bench with a clean power supply, in a room with good lighting, mounted on nothing. A touch panel behaves differently when it is screwed into a slightly bowed wall box in a humid bathroom. A doorbell that focuses perfectly on a bench can be aimed at a reflector once it is under a porch soffit.

This is why a sample that works on a table is a weak signal. It tells you the design is sound. It does not tell you the unit in the box will behave the same way after a week in the customer's house, and it certainly does not tell you what the five-thousandth unit will do.

What a confirmation sample is actually for

A confirmation sample is where you test the things that only show up under production conditions: the glass edge finish after a real polishing run, the colour match between two panels made in different batches, the feel of the relay, whether the silkscreen sits where the drawing said, whether the packaging actually closes.

It is also where you learn the yield story. If the factory can only produce one good unit out of four attempts, you will not see that on the bench, but you will see it in the price and the lead time.

Our own numbers are not secret and they frame this stage: sampling runs 7–15 days, new tooling runs 25–40 days, and standard models start at 500 pieces per colour. A confirmation sample sits inside the first window, before the second one, and skipping it does not save the 7–15 days — it just moves the same discussion into the middle of a production run, where it is far more expensive.

The sealed sample is the only one that ends an argument

Pull one unit from the run, sign it, photograph it, seal it, and keep one copy on each side. If a dispute comes up six months later about glass colour, print position, or button feel, there is a physical reference. Without it, both parties are describing the same object from memory and both are right.

The sealed sample is cheap and almost nobody asks for it, which is exactly why disputes about appearance drag on for weeks.

Where the sample fits the customisation ladder

The ladder we work to, stated plainly: samples from 5 pieces, charged and credited against your first bulk order; logo as sticker or silk print from 10 pieces per model; laser engraving from 300; custom packaging 500–1,000; your own product ID from 1,000; new tooling and ODM design 2,000–3,000.

Each rung of that ladder needs a different kind of sample. A logo sample needs a confirmation sample, because the placement is the whole question. A custom packaging order needs a production sample of the box, not of the product — the box is what fails first, and it fails on the line when the carton will not close. A tooling project needs both a confirmation sample from the first trial shot and a sealed sample from the first real run.

Asking for one sample across all of them is how a project gets to the container stage still carrying unresolved decisions.

What to write into the order

Three lines. Which sample type is being approved, in words — not just "sample approved." What that approval does and does not cover. And that one sealed unit from the production run will be retained by both sides until the warranty period ends.

That is the whole mechanism. It costs one paragraph in a purchase order and it removes the most common category of post-shipment argument there is.

Sourcing references: XingLiao ODM/ODM and the OEM programmes.

The Name on the Certificate Decides Whether You Can Change Factory Later

Xingliao Smart Home — Mon, 02 Nov 2026 05:14:27 +0000

What a certificate actually certifies

A CE declaration, an FCC ID, a UKCA file — buyers read them as a property of the product. They are not. They are a property of a specific combination: this construction, this set of internal parts, this factory, and this applicant. The applicant is the name printed on the paperwork, and it is the one line almost nobody reads before signing.

That name matters because certificates are issued to whoever applied for them. If the factory applied, the factory holds the file. If you applied, you hold it. The physical product is identical either way, and the difference only becomes visible on the day you want to do something the factory does not want you to do.

The three things the applicant's name controls

The first is whether you can move. If the file is held by the factory, it does not travel with you. You can take the tooling elsewhere — assuming the tooling release clause says so — but the certificate stays behind. The new plant makes the same product and has no paperwork for it, which means a fresh round of testing, a fresh file, and a fresh lead time that was never in your plan.

The second is whether the certificate survives a component change. Certification covers a construction, not a drawing. When a chip goes end-of-life and the factory substitutes an equivalent part, the file either already covers that family of parts or it does not. If you hold the file, you decide. If the factory holds it, you find out when a shipment is held.

The third is who has to answer when a market authority asks. Import surveillance, a marketplace compliance request, an Amazon or retail portal document check — those ask the applicant. If the applicant is a company you do not control, you are waiting on someone else's inbox to release your goods.

The honest version of what we can offer

We support CE, FCC, RoHS, UKCA, CCC and PSE documentation on the platforms we build. Say it plainly and check it, because the gap matters more than the list: we do not hold a current UL or ETL file, and North America in practice asks for UL or ETL (cUL or CSA in Canada) on the electrical safety side. FCC only covers the radio. So for a North American launch the honest answer is "confirm case by case whether a live file exists for that exact model," not a blanket yes. Buyers who hear that early plan around it. Buyers who hear it at the port do not come back.

For markets where we do hold coverage, the question to ask is not "are you certified" but "whose name is on the file, and what happens to it if I move."

Where the cost actually sits

A certificate is not one cost, it is two. There is the test, which is finite and quotable. Then there is the file maintenance, which is the one that surprises people: model listing updates, changes of applicant, re-testing after a substitution, and in some markets a registration that has to be renewed. The second cost is small per event and endless over a product's life, and it belongs to whoever holds the file.

This is also why "just add our brand" is not free. Adding a model name to an existing listing is documentation work. Adding a new applicant is closer to starting over.

What to settle before the first order

Four lines in the RFQ settle it. Which market, and which scheme applies in that market — not "Europe" but the specific directive and whether self-declaration is allowed for this product class. Whose name will appear as applicant on the file. What happens to the file if the buyer moves production, in writing. And who pays for a re-test triggered by a component substitution the factory had to make.

None of these are adversarial questions. A factory that holds its own files will say so, and the answer is workable as long as you know it is the case before you commit volumes against it.

How this fits the rest of the decision

The certificate holder question sits next to tooling ownership and sample sign-off, and the three are usually negotiated by different people inside the same order. Tooling decides whether you can move. The certificate decides whether moving is even useful. The signed sample decides what "the same product" means once you have moved. Settle all three in the same document, because alone each one sounds minor and together they are the whole of your supply continuity.

Sourcing references: XingLiao ODM/ODM and the manufacturing programmes.

A Doorbell That Will Not Join the Wi-Fi Is Usually on a Different Band Than Your Phone

Xingliao Smart Home — Mon, 02 Nov 2026 03:32:54 +0000

The setup story is always the same and always frustrating. The app says searching for device. The doorbell blinks the way the quick start guide says it should. The phone is two feet away, connected to the same home network, and nothing happens. After twenty minutes the assumption is a defective unit.

Nine times out of ten the doorbell is fine and the two devices are simply not capable of talking to each other yet, because they are operating on different radio bands and the setup process depends on them being on the same one.

Almost every doorbell is 2.4 GHz only

Wi-Fi hardware for battery devices is built around 2.4 GHz for three reasons that have not changed in a decade. It reaches further through walls, which matters for a device mounted outside a front door. It is cheaper in silicon and in power, which matters for something running on a battery. And it is what the low-power chipsets these devices use support.

Your phone is almost certainly connected to the 5 GHz band right now, because that is what modern routers hand to phones, and because band steering actively pushes dual-band clients onto 5 GHz. The phone and the doorbell are on the same network in the way that matters for data, and on different frequencies in the way that matters for device discovery.

Some setup flows use Bluetooth for the initial pairing and only hand over network credentials, which sidesteps the problem. Many others use a direct Wi-Fi handshake between phone and device, and those are the ones that fail when bands differ.

What to do about it, in order

Turn off band steering temporarily, or split the SSIDs. If your router publishes one network name for both bands, the phone has no way to choose. Giving 2.4 GHz its own name, or a temporary separate guest name, lets you put the phone on the same band as the doorbell for the five minutes of setup. You can put everything back afterwards.

On phones that allow it, forget the 5 GHz network during setup so the phone falls back to 2.4 GHz. This is the least invasive option and works on most handsets.

Keep both devices close to the router during setup, not close to each other. The doorbell needs to hear the access point at install quality, and a doorbell that is being set up in a hallway at the far end of the house is being tested at a signal level it will never have in service.

If the app offers an access-point mode or a compatibility mode, use it. Those modes make the phone connect directly to the device rather than relying on the router to broker the handshake, and they exist precisely for this failure.

The router settings that break setup even on the right band

Client isolation, sometimes called AP isolation, prevents wireless devices from talking to each other. It is a sensible setting for guest networks and it makes device discovery impossible. If your doorbell setup network is a guest network, start there.

Hidden SSIDs cause a similar problem. A device that cannot see the network name being advertised cannot join it during a scan-based setup flow, and some devices will not accept a manually typed hidden name at all.

MAC filtering and allow-listing produce a specific signature: the doorbell appears to connect, gets an address, and then fails at the cloud registration step. That looks like an account problem and is actually a router problem.

A mesh system with a single combined SSID is the modern version of all of this. Meshes are excellent once devices are joined and awkward during setup for exactly the band-steering reason above. Most mesh apps have a temporary IoT network or a setup mode; use it.

Once it is on the network, the real question starts

Joining is not the same as staying joined. The failure that follows a week later has a different cause and a different fix, and it is worth separating the two.

A doorbell that joined fine and then drops is usually at a marginal signal level. The useful number is the received signal strength, and the practical threshold is around -70 dBm: better than that and the link is generally stable, worse than that and you will see intermittent drops that look random. Measuring it at the door, with the door closed, before you mount anything permanently, is the single most valuable five minutes of the whole installation.

A doorbell on a battery sleeps and wakes. Every wake is a re-association with the access point, and at a marginal signal those re-associations occasionally fail, which is why the symptom is intermittent rather than constant. A wired doorbell with a permanent connection does not have this failure mode, and that difference alone explains a large share of the "my battery doorbell keeps disconnecting" reports.

Channel congestion is the other half. 2.4 GHz has three non-overlapping channels and your neighbours are using them. A channel that was clean at install time can be busy six months later, and the symptom is a doorbell that fails in the evening when everyone is home.

Buying with this in mind

Check the specified Wi-Fi band before you buy, and treat a 2.4 GHz-only specification as normal rather than as a limitation. Ask whether the app supports an access-point or compatibility setup mode, because that single feature removes most of what this article describes.

If your router is a modern mesh with one combined SSID, check whether the vendor publishes an IoT or setup network, and confirm before the doorbell arrives rather than after. And if the installation point is far from the router, plan for the signal rather than the doorbell: the video doorbell range includes battery and wired models, and the wired ones remove the sleep-wake re-association failure entirely.

Where the doorbell is part of a larger system, the same applies to the indoor station and to any chime that joins the network, which is why installers working from a video intercom manufacturer specification tend to plan the network first and choose the hardware second. It is a much cheaper order to get wrong than the reverse.

If you are comparing models, the two specifications worth reading past the headline resolution are the supported band and the stated operating signal level, and both are listed on the doorbell category page for the models that publish them.

Smart Smoke Alarm and Interconnected Are Not the Same Thing

Xingliao Smart Home — Mon, 02 Nov 2026 03:31:14 +0000

"Smart Smoke Alarm" and "Interconnected" Are Not the Same Thing

A Wi-Fi smoke alarm that pushes a notification to your phone is a useful device. It is not an interconnected alarm, and conflating the two is how people end up with a house full of detectors that do nothing for anyone asleep upstairs.

Two different jobs

There are two dependant behaviours that get sold under one word:

Notification — somebody gets told. Usually via phone.
Simultaneous sounding — all alarms in the dwelling sound at once, regardless of which one detected smoke.

A notification is worthless to a sleeping occupant. A downstairs detector alerting a phone two floors away does not wake anyone. Only simultaneous sounding does that, and it is the mechanism every residential fire code that I am aware of is actually built around.

So the hierarchy worth remembering is:

Hardwired interconnection — a dedicated interconnect wire between units, mains-powered with battery backup. Fastest, most reliable, least dependent on anything else.
Proprietary wireless interconnection — RF link between units from the same manufacturer. Usually fine, sometimes excellent. The catch is that it only works within that manufacturer's ecosystem, so a partial upgrade can leave you with two independent sets.
Wi-Fi notification only — tells your phone. This is not item 2, and a product page that emphasises app alerts is often quietly telling you interconnection is absent.

If you are replacing one head in an existing interconnected set, read carefully whether it supports the existing interconnect standard. Many do not, and installing one silently splits your set into two.

Three things worth doing regardless

Do not put relays on dry contact input modules. This comes up constantly in retrofit discussions. A detector's signalling contacts are part of its own circuit. Loading them with external relays and coils changes electrical conditions the detector was not designed for, and it can suppress its own alarm. Use the module's designed signalling output, or choose a detector with the interface built in.

Replace on a clock, not on a symptom. The sensing element — ionisation or photoelectric — degrades. Ten years is the common service life and most units say so on the label, along with a manufacture date. Alarms almost never fail in a way you notice; they fail by becoming slow. If you cannot read the date on the unit, replace it.

Test monthly, whether or not it is smart. The button exercises the full chain — sensor, logic, sounder. Nothing the app tells you substitutes for hearing the horn. A self-test that only verifies network connectivity tells you the Wi-Fi works.

Where the smart part genuinely earns its keep

None of the above makes connectivity useless. It is good at one specific job: telling you when you are not there. Knowing your kitchen detector fired while you were at work is genuinely valuable, and no interconnected-only system can do it.

The correct mental model is that connectivity is an addition to interconnection, never a substitute. Buy the interconnection first, then take the app as a bonus.

One thing to check before you buy twelve units: whether they will all still be purchasable in five years. Proprietary wireless sets are ecosystem-locked. A manufacturer that discontinues a set leaves you with no way to expand it, at which point you replace everything.

I work on the manufacturing side of this category, so take this with the usual grain of salt — we sell connected detectors too. But nothing we build changes the physics of who hears what from two floors away.

Interconnected behaviour comes up across our sensors and alarms range, and our engineering and compliance notes cover what the certifications do and do not tell you.

Xingliao Smart Home builds smart switches, video doorbells and security cameras on Tuya WiFi and Tuya ZigBee, and runs OEM/ODM programmes from Shenzhen. Factory and ODM capability.