Most foreign companies sourcing from China believe they have made a choice between two contracts. In practice they have made a choice between two sets of default legal rules, and the document they sign determines which set applies when the relationship goes wrong. That choice is usually made by whoever sends the template, at a point when nobody is thinking about title, risk, tooling or the right to walk away.

The commercial stakes are easy to state. Under one structure the company is a buyer of goods: it takes what the factory makes, pays for it, and owns a warranty. Under the other it is a commissioning party: it brings a design, pays for tooling, and expects to own the output and the means of producing it. The second structure costs more to negotiate and buys protections the first cannot provide. The first is cheaper and faster, and is the correct answer more often than founders assume.

The expensive mistakes run in both directions. A company doing custom production on a purchase order has bought a commodity contract for a bespoke relationship, and discovers at the point of dispute that it does not own the mould, cannot inspect the line, and has no answer to the factory selling its design to a competitor. A company that imposes a fifteen-page manufacturing agreement on a supplier of standard power adapters has spent legal fees on protections it will never use and has slowed a relationship that needed a price schedule.

The label on the document does not decide the regime

Chinese law treats a sale of goods and a contract for work as different contract types, with different default rules on quality, risk, inspection, subcontracting and termination. The Civil Code contains a chapter on sales contracts and a separate chapter on work contracts, the type used for processing, customisation and repair. Which chapter applies is not determined by the title on the cover.

A Chinese court or tribunal looks at substance. Who supplied the materials? Who controlled the process? Is the output a standard product the maker sells to others, or a bespoke item made to the customer’s specification? Did the customer pay for tooling? A document called a purchase agreement under which a factory makes a bespoke product to the buyer’s drawings, using a mould the buyer paid for, may be treated as a work contract whatever its title. A document called a manufacturing agreement under which the factory supplies its own catalogue product in the factory’s packaging may be treated as a sale.

That matters because the defaults are not neutral. Where the contract is silent, the statutory regime fills the gap, and the statutory regime was designed for a different transaction from the one the parties thought they were doing. The practical response is to draft to the substance and to say expressly which regime is intended, rather than to rely on a title. A clause that states the parties intend a contract for work, that the buyer supplies the tooling and the design, and that the maker is paid for processing rather than for the goods, does real work if the relationship is later litigated.

Mixed arrangements are common and are handled by reference to the contract type that characterises the relationship. A contract under which the factory supplies some materials, the buyer supplies others, and the factory performs a substantial transformation is closer to a work contract than to a sale. The clause that states the intention is worth more than the argument about the label.

There is a practical test that a founder can apply before the lawyers are involved. Ask who owns the drawings, who owns the mould, who decides the bill of materials and who carries the inventory if the order is cancelled. If the answers point to the factory, the relationship is a supply relationship and a supplier agreement is the right instrument. If the answers point to the buyer, the relationship is a manufacturing relationship and a supplier agreement will leave the buyer’s most valuable assets undocumented.

Title, risk and control

Four consequences follow from the choice, and each has a price.

Title. In a sale, ownership passes when the contract says, and in default when the goods are delivered. Retention of title is recognised, and a seller can provide that ownership does not pass until payment, but an unregistered retention clause may not bind a good-faith third party, which means it is a weak security against a factory’s bank. In a work contract, the materials the customer supplies remain the customer’s property throughout, and the finished work is produced from them. That is the legal foundation for the proposition that a buyer who pays for tooling and supplies it owns the tooling.

Risk. In a sale, risk passes on delivery, and where the contract requires carriage, when the goods are handed to the first carrier. That is why an FOB or FCA term leaves the buyer bearing transit risk while the question of conformity remains open: the risk of loss in transit and the question whether the goods conform are separate questions, and a contract that treats shipment as acceptance has answered the second by reference to the first. In a work contract, the maker must keep the customer’s materials and the work in progress safe and is liable for loss caused by its failure to do so, which puts the custody obligation on the party that physically holds the goods.

Control. A work contract gives the customer a statutory right of necessary supervision and inspection over the progress and quality of the work. A sale gives the buyer no equivalent right: it can inspect the goods, not the process. For a company that needs to see a production line, audit incoming material records or place an engineer on site during ramp-up, that difference is the whole point, and it has to be recreated by contract if the structure is a sale.

Subcontracting and termination. A work contract requires the maker to perform the main work itself, and where the main work is subcontracted without the customer’s consent, the customer has a right to terminate. A sale imposes no such limit: the seller may buy the goods from anyone. A work contract also allows the customer to terminate at any time on compensating the maker for its loss. That is a genuine exit right that a sale does not confer, and it is worth having when a product is cancelled. It is not free: the compensation can include the maker’s lost profit on the unperformed work, and it does not by itself get the tooling back.

QuestionSale of goodsContract for workWhy it matters commercially
Who owns the designSeller, unless assignedBuyer, where the buyer supplied itDetermines whether the factory can sell the product to a competitor
Who owns buyer-funded toolingNot addressed by defaultBuyer, where the buyer supplied the toolingDetermines whether the mould can be recovered
When risk passesOn delivery or handover to the carrierOn acceptance of the workDetermines who bears transit loss and who bears rework
Right to inspect the processNone beyond the goodsStatutory supervision rightDetermines whether the buyer can see how the product is made
SubcontractingPermittedMain work requires consentDetermines whether the product can be made in an unapproved plant
Termination for convenienceNot available unless agreedAvailable on compensationDetermines whether a cancelled product can be stopped
Quality benchmarkContract, then national standard, then industry standard, then usageThe customer’s specificationDetermines whether the factory can fall back on its own yield
Defect notificationBuyer must inspect and notify within the agreed or reasonable periodAcceptance against the specificationDetermines whether a late complaint is time-barred

The asymmetry is worth stating plainly. Custom production is the case in which the buyer most needs control, and it is the case in which the work contract regime supplies control by default. Commodity purchasing is the case in which the buyer needs speed and price, and the sale regime supplies that. The mistake is to buy custom production on a sale contract and then try to negotiate the control back clause by clause, which produces a long document that is still, in its defaults, a sale.

The supplier agreement: what it is good for

A supplier agreement governs the purchase of goods the factory already makes and sells to other customers. Its natural contents are the product description, price and price adjustment, order mechanics and forecast, delivery terms, payment, minimum order quantity, warranty, confidentiality, term and termination, and dispute resolution. It is a good document for its purpose, and its purpose is narrow.

It works when the product exists in the factory’s catalogue, the design belongs to the factory, customisation is limited to branding and packaging, no tooling investment is required, and the buyer’s exposure is a warranty claim rather than an IP claim. Standard modules, passive components, cables, power supplies, packaging and spares are all in this category. A company assembling its own product from bought-in parts should be buying on supplier agreements and spending its legal budget on the specification annex and the change-control clause.

Where it stops is predictable. There is no IP ownership provision, because the factory owns what it makes. There is no tooling provision, because there is no buyer tooling. Confidentiality is thin, usually a mutual NDA bolted on and never operationalised. There is no exclusivity and no non-compete, so the factory can sell the same product to a competitor, which it will. The warranty is short and excludes consequential loss. Quality is defined by reference to the factory’s specification, which is the specification the factory can meet at its yield.

One point is often missed. A supplier agreement does not allocate the regulatory role. If the foreign company is the importer of record, it carries the certification and labelling obligations, and it carries product liability to end users, whatever the supplier agreement says. A warranty from the factory is a contractual claim; it does not answer a regulator or an injured customer. Companies buying catalogue goods for resale or incorporation need a supplier agreement that deals with conformity evidence, certificates of compliance and the documents needed to defend a product liability claim, and most templates do not.

Branding is the specific trap. Adding a logo to a catalogue product does not transfer the design, and it does not make the product the buyer’s. It does, however, put the buyer’s name on the goods, which affects how a Chinese regulator and a Chinese court will treat the buyer if the product injures someone. A white-label arrangement sits between the two models and needs clauses from both.

A supplier agreement can be improved without turning it into a manufacturing agreement. Four additions do most of the work: a specification annex with revision numbers and a golden sample, an acceptance and notification procedure, a change-control clause requiring written approval for any substitution affecting form, fit or function, and a compliance pack obligation covering certificates, test reports and material declarations. These clauses cost little and close the gaps that matter for a catalogue purchase.

The manufacturing agreement: what it actually buys

A manufacturing agreement is the document for custom production. It builds on the supplier agreement and adds the provisions that determine who owns what.

IP ownership. The agreement should separate background IP, which each party brings, from foreground IP, which is created during the relationship. Background IP stays with its owner, with a licence to the other party limited to the purpose of the agreement. Foreground IP, meaning design improvements, new tooling designs, custom firmware, test fixtures and production know-how developed for the product, should be assigned to the buyer where the buyer has paid for the development, and the clause should be an assignment rather than a promise to assign, with an obligation to execute further documents.

Chinese copyright law adds a wrinkle that surprises foreign drafters. Economic rights in copyright can be assigned, but moral rights cannot be transferred outright. A clause that assigns the economic rights and grants a broad waiver, consent and exclusive licence in respect of the rest is the workable structure, and it should be in the Chinese text as well as the English.

Three IP questions are usually left unanswered and later litigated. The first is improvements made after the project ends: a factory that continues to refine a process for the buyer’s product will argue that the refinement is its own. The clause should state that improvements to the product, the tooling and the test method belong to the buyer, and that improvements to the factory’s general process remain the factory’s, with a licence to the buyer for the product’s life. The second is employee inventions. The factory should warrant that its employees’ contributions have been assigned to it and that it has the right to assign them onward, because a Chinese employee-invention claim can sit between the buyer and the asset. The third is the factory’s pre-existing obligations to other customers. An ODM factory that has granted exclusivity to another brand may be unable to grant the buyer anything, and the warranty should say so.

Tooling. Tooling is the asset most often lost and the cheapest to protect. The agreement should state who owns each mould, fixture and jig; that ownership is not affected by who paid for it in instalments or by where it is stored; that the tooling is marked with the buyer’s name and an asset number; that the factory maintains it, insures it and does not modify it without approval; that the buyer may inspect it and take photographs; and that on termination the buyer may remove it within a defined window, with the factory obliged to release it and to cooperate with the removal. Where the moulds sit at a third-party moulding shop, the obligation has to reach that shop, because the buyer has no contract with it.

The removal right is worth little without two supporting instruments: a power of attorney that lets the buyer or its agent collect the tooling, and a liquidated sum for each day the factory retains it. A factory that is owed money will use the mould as leverage, and a clause that says the buyer may remove it does not answer the question of who is entitled to possession on the day.

A tooling schedule turns the clause into an operational record. For each tool it should list the asset number, the part it makes, the machine it fits, the number of cavities, the location, the replacement cost, the insurance value, the date of last maintenance and the ownership basis. A buyer that can produce that schedule in a dispute is in a different position from one that has a clause and no list.

Where the tooling is paid for through an amortised unit price rather than upfront, the agreement should say when title passes. The common structure is that title passes on the first payment and the factory holds the tool as bailee until it is collected. Without that sentence, the factory’s accounts will treat the tool as its fixed asset, and a receiver or a bank will do the same.

NNN. Non-disclosure, non-use and non-circumvention is the China-specific upgrade on an NDA. Non-disclosure stops the factory revealing the design; non-use stops it using the design for anyone else; non-circumvention stops it going around the buyer to the buyer’s customers or suppliers. Each limb needs a defined confidential category, a defined permitted purpose and a liquidated sum that is tied to a plausible loss, because a liquidated sum that is disproportionate to the actual loss can be adjusted by a court or tribunal.

Exclusivity and non-compete. If the buyer is paying for tooling and development, it will want the factory not to sell the same product to competitors. Scope it by product, by market and by duration, and tie it to the buyer’s minimum commitments. A non-compete that reaches the factory’s entire business is both commercially unrealistic and exposed to competition law scrutiny, since Chinese competition law treats vertical restraints by reference to their effect on competition and their necessity.

Capacity and allocation. A manufacturing agreement should reserve capacity, state the buyer’s priority in a shortage, and define how forecasts become binding orders. Without an allocation clause, a factory facing a component shortage will serve its largest customer first, and a small buyer will be told that the line is full.

Change control and quality. These are the operational clauses, and they are treated separately below, but their legal function belongs here: they define what the factory may change without permission, and what happens when the goods do not conform.

OEM, ODM, JDM and toll manufacturing. The industry labels describe who designed the product and who owns the relationship with the end customer. They do not have fixed legal meanings in Chinese law, so the contract has to supply the meaning.

In an OEM arrangement the buyer supplies the design and often the tooling, and the factory manufactures to specification under the buyer’s brand. The buyer owns the design and normally applies for certification in its own name or through its Chinese entity. The factory owns its process know-how, which is why the agreement should say so rather than leaving the factory to argue that the process improvements are its background IP.

In an ODM arrangement the factory designs the product, or designs it with the buyer’s input, and the buyer rebadges it. Three issues dominate. Ownership of the design, which the factory will say is its own because it employed the engineers; exclusivity, because the factory’s business model is to sell the same design to several brands; and certification, because the factory usually holds the certificate, which means the buyer’s product stops being sellable if the factory’s certificate is suspended. An ODM agreement should buy exclusivity for the buyer’s market, secure the right to obtain the technical file, and require the factory to maintain the certificate for a defined tail after the relationship ends.

JDM, joint development, is the arrangement in which both sides contribute engineering. It is the one that most often ends in a dispute about ownership, because neither side documented who contributed what. The agreement should identify each party’s background IP, define the joint output, decide who files patents and in whose name, and set out the licence each party retains if the relationship ends. A royalty on the other party’s future sales is a common compromise and should be drafted as a payment obligation with an audit right, not as a handshake.

Toll manufacturing is a different axis. In processing with supplied materials, the foreign party supplies the materials and pays a processing fee; in processing with imported materials, the factory imports the materials and sells the finished goods. The customs and tax treatment differs between the two, the party that is the exporter of record differs, and the ownership of the materials during processing differs. The structure affects which entity applies for certification, who can claim a VAT refund and who bears the duty. These are tax and customs questions, and they should be settled with an adviser before the contract is drafted, because changing the structure afterwards means changing the customs declarations.

Liability, quality and warranty

Liability is the part of the contract that is negotiated last and read first after an incident. The two structures allocate it differently, and the difference is larger than the price difference between the documents.

ExposureSupplier agreement defaultManufacturing agreement positionWhat to draft
Defective goodsConformity to contract or statutory standard; buyer must inspect and notifyConformity to the buyer’s specification and approved sampleAcceptance procedure, latent defect carve-out, repair, replace or credit hierarchy
Consequential lossCommonly excluded or capped at invoice valueNegotiable, and should reach recall and line-down costs for manufacturing defectsCap as a multiple of annual spend, with carve-outs for IP, safety and wilful misconduct
Product recallFalls on the entity treated as producer, usually the brand ownerFactory indemnity for a manufacturing defectRecall protocol, cost allocation, notification duty, insurance
IP infringementThin warranty of quiet possessionBuyer indemnity for its design; factory indemnity for its processSeparate indemnities by source of the claim
Tooling loss or damageNot addressed, because the factory owns the toolBuyer owns, factory is custodianCustody, insurance, removal, liquidated damages for retention
Late deliveryDate stated, no consequence unless draftedLiquidated damages for delayRate per week, cap, definition of excused delay
Personal injuryProduct liability attaches to the producer or brandIndemnity and insuranceProduct liability insurance naming the buyer, incident notification
Unapproved subcontractingPermittedConsent required for the main workApproval right and flow-down of obligations
Confidentiality breachMutual NDA, rarely operationalisedNNN with a liquidated sumDefined information, permitted purpose, remedy
TerminationNotice, no transition dutyTermination for convenience on compensationExit schedule and transition assistance

The first drafting point is the cap. A supplier that offers a cap at the invoice value has offered a remedy that is proportionate to the order and irrelevant to the loss. A manufacturing agreement should set the cap by reference to the buyer’s exposure, commonly a multiple of the annual spend or the value of the affected production, with carve-outs that survive the cap for IP indemnities, recall costs, personal injury and wilful misconduct. The carve-outs matter more than the multiple, because the events that hurt are the ones a court would treat as falling outside an ordinary cap.

The second is the remedy hierarchy. Repair, replacement, credit and termination should be sequenced, with time limits at each stage, so that a buyer is not left arguing about whether it must accept a rework that will miss the season. A clause that gives the factory the right to repair “within a reasonable time” gives it the right to choose the buyer’s launch date.

The third is insurance. An indemnity from a factory with no balance sheet is a promise, not a recovery. The agreement should require product liability insurance, with the buyer named as an additional insured or loss payee where local practice permits, and should require evidence of cover at each renewal. For a consumer product, the recall cost is the exposure that most often exceeds the factory’s net assets.

Quality is where the two regimes diverge most sharply in daily operation, and where the contract has to do the most work.

In a sale, the seller must deliver goods that conform to the agreed quality. Where the contract is silent, the law fills the gap by reference to the national standard, then the industry standard, then trade usage. The practical effect is that silence hands the specification to the factory, because the factory’s own standard is the standard it can meet. The sale regime also places an inspection burden on the buyer: the buyer must inspect within the agreed period, or a reasonable period where none is agreed, and must notify non-conformity within it. Silence can be treated as acceptance, subject to defects that a reasonable inspection would not reveal and to defects the seller knew of. A buyer who receives a container, stores it, and complains six weeks later may find that acceptance has already occurred.

In a work contract, conformity is measured against the customer’s requirements, which is both a strength and a risk. The strength is that the specification is the buyer’s, so the factory cannot fall back on its own standard. The risk is that a defective specification is the buyer’s problem: a factory that builds exactly what the drawing says has performed, even if the product does not work. That is why the specification set, the golden sample and the design-for-manufacture review matter more in a work contract than in a sale. Approval of a sample is evidence of conformity; it is not, without an express clause, a waiver of defects that a reasonable examination would not reveal.

Warranty sits on top of both. The statutory quality obligations and the contractual warranty are different things, and the contract should state the warranty period, its start date, its scope, its exclusions and its remedy. The gap that matters commercially is between the warranty the buyer gives its own customers and the warranty it receives: a buyer that promises twenty-four months and receives twelve has accepted a year of exposure, and a buyer that promises a seven-day replacement and receives a forty-five-day factory turnaround has accepted a service failure it cannot remedy.

Product liability is the obligation neither regime allocates by default. Chinese product liability looks to who put the product into circulation and under whose name. Where the brand owner commissions the manufacture and the product carries only the brand owner’s name, the brand owner is exposed as though it were the producer, and its recourse against the factory is only as good as the factory’s balance sheet and the indemnity in the contract. A defective consumer product can also trigger a mandatory recall administered by the market regulator, and the cost of a recall falls on the party treated as the producer. The manufacturing agreement should require the factory to carry product liability insurance naming the buyer, to notify the buyer of any safety incident, and to cooperate in a recall at its own cost where the defect is a manufacturing defect.

Acceptance deserves a procedure rather than an adjective. The workable sequence is a pre-shipment inspection against a defined sampling plan, a documented acceptance decision within a fixed number of days of arrival, a cure period for non-conforming batches, and an express reservation of claims for latent defects that a reasonable inspection would not reveal. Without the reservation, the acceptance signature becomes a waiver, and the buyer discovers the defect after the warranty has run.

Choosing the structure

Five facts decide the answer, and they can be established in a single meeting.

FactPoints to a supplier agreementPoints to a manufacturing agreement
Who owns the designThe factory’s catalogue productThe buyer’s design, or a design the buyer has commissioned
Who pays for toolingNo buyer toolingBuyer-funded moulds, fixtures or test equipment
How much IP is sharedNone beyond forecasts and pricesDrawings, firmware, process know-how, test methods
Volume and customisationStandard runs, logo and packaging onlyCustom configuration, dedicated line, ramp commitments
Regulatory and brand roleBuyer is a reseller of the factory’s certified productBuyer is the brand owner, the applicant for certification, or the importer of record

Where the answers split, the structure should follow the fact that carries the most risk. A company buying a catalogue product but applying its own brand has a supplier relationship with a manufacturing exposure: it needs the supplier agreement’s commercial terms and the manufacturing agreement’s certification, indemnity and change-control clauses. A company commissioning a custom product from a factory that also sells a similar catalogue item needs the manufacturing agreement, because the design boundary is the thing most likely to be disputed.

The structure also has to match the certification position. If the factory holds the certificate for an ODM product, the buyer’s route to market depends on the factory’s compliance record, and the agreement should treat the certificate as a deliverable with maintenance obligations. If the buyer holds the certificate, the factory’s production premises are named on it, and the agreement must prohibit moving production without consent.

For most companies the answer is not one document but two layers. A framework manufacturing agreement governs the relationship: IP, tooling, NNN, quality standards, change control, warranty, liability, term and termination, and dispute resolution. Individual purchase orders, issued under the framework, specify quantity, price, delivery date, configuration and shipping terms for each run. The framework should contain an order of precedence clause ranking the documents and stating that the purchase order is a call-off, not a fresh contract. Without it, the factory’s acknowledgement, printed with its own terms on the reverse, becomes the last document in the chain and the one a tribunal is most likely to treat as the operative bargain.

The two-layer model also solves a practical problem. Product terms change every quarter; IP and liability terms should not. Separating them means the commercial team can issue POs without reopening the protections, and the legal team can review the framework once a year.

Governing law, language and enforcement

The choice of law is usually made for the wrong reason. A buyer that specifies its home law because it feels safer has bought a judgment it cannot enforce. Chinese courts do not enforce foreign court judgments in the absence of a treaty or a reciprocity arrangement, and China has no such arrangement with most Western jurisdictions. The practical consequence is that a judgment from the buyer’s home court against a Chinese factory is close to worthless.

Chinese law with arbitration in China or Hong Kong is the workable combination. CIETAC and SHIAC are the common mainland institutions; HKIAC is common where the parties want a neutral seat. All three produce awards that are enforceable in China and, through the New York Convention, in the counterparty’s home state. The principle is simple: the dispute resolution mechanism must be executable where the factory’s assets are.

Two supporting clauses make the arbitration usable. A preservation clause, because a Chinese court can grant property preservation in aid of arbitration and the window to apply is short. And a service address clause, because a respondent that cannot be served can stall an arbitration for months; naming an address for service and an email address, and agreeing that service there is effective, removes the argument.

Language is a substantive choice, not a formatting one. The convention for China-facing manufacturing contracts is a bilingual document with the Chinese text prevailing. That is the right default, because a Chinese court or tribunal will apply the Chinese text and a clause that exists only in English invites an argument about translation. The consequence for the buyer is that the Chinese text must be checked by someone who would be willing to argue it, not by a translation agency alone, and the English text should be marked as a convenience version so that no one later argues that the English prevails.

Execution formalities deserve a line. A Chinese company binds itself by its company chop, and a document bearing the chop is difficult for the company to disown even where the signatory’s authority is contested. The buyer should verify who holds the chop, obtain a copy of the business licence and the legal representative’s identification, and keep the correspondence that shows the signatory’s authority. Where the counterparty is a group, a parent guarantee is worth more than a promise from a special-purpose subsidiary with no assets.

Documents, steps and responsible parties

The sequence matters as much as the contents. The NNN has to precede the first drawing, the framework has to precede the first tooling payment, and the exit schedule has to be written while the parties still like each other.

StepWho is responsibleTypical timingWhat foreign companies miss
Confidentiality before technical disclosureBuyer legal with the factoryBefore the first drawing is sentAn NDA without non-use and non-circumvention
Counterparty checksBuyer legal and financeBefore the term sheetChecking the factory but not its holding company or guarantor
Commercial term sheetCommercial leads on both sidesBefore the factory quotes toolingNo statement of who owns the design or the mould
Framework agreementBuyer counsel, reviewed by Chinese counselThree to six weeksAn English-only draft with no order of precedence
Specification annexBuyer engineeringBefore the pilot runRevisions not identified or dated
Tooling schedule and tooling orderBuyer operations with legalBefore the tooling paymentNo power of attorney to collect the moulds
Quality agreementQuality teams on both sidesBefore the first shipmentWarranty shorter than the promise to end customers
Change control procedureEngineering and qualityBefore mass productionNo approval right over component substitution
Signature and chopLegal representativesBefore the tooling paymentSignatory authority unverified
Purchase ordersBuyer supply chainEach production runPO terms that contradict the framework
Exit scheduleBuyer legalAt signing, not at terminationNo transition obligation, so the buyer has a right and no product

The annexes carry more weight than the body in most disputes, because they are where the product is defined. A specification annex should include the drawing set with revision numbers, the bill of materials with approved manufacturers for critical components, the golden sample reference, the test procedure and the packaging artwork. A quality agreement should include the sampling plan, the acceptance criteria, the defect classification, the rework rules and the warranty remedy. A tooling schedule should include the asset list described earlier. Each should be signed or initialled by both parties, because an annex that is referred to but never signed is an argument about whether it was agreed.

Where foreign companies get this wrong

The first failure is operating on purchase orders. A PO is a price and a quantity; it says nothing about IP, tooling, change control or the right to inspect. Companies that run a custom product on POs for a year and then try to negotiate a manufacturing agreement discover that the factory has already absorbed the design into its own catalogue.

The second is assuming that paying for the mould creates ownership. Payment creates a claim for money, not a property right, unless the contract says otherwise. The factory’s accounts will show the mould as its fixed asset, and a receiver or a bank will treat it as one.

The third is a specification that lives in email. A drawing sent as an attachment, revised twice, with the final version in a messaging thread, is not a controlled document. When the goods do not conform, the factory will produce a different revision and the argument becomes about which message was the specification.

The fourth is home-country law. It reads as caution and functions as a waiver, because the judgment it produces cannot be enforced where the assets are.

The fifth is an IP clause that assigns “all intellectual property” without defining it. The clause looks strong and is unenforceable in practice, because neither party can say what was assigned. The workable clause lists the categories, attaches the schedule of deliverables, and deals with improvements made after the project ends.

The sixth is treating the NDA as the IP protection. An NDA protects information; it does not transfer ownership, does not stop the factory making the product for someone else, and does not reach the factory’s sub-suppliers. The NNN clause and the IP assignment do that work.

The seventh is a warranty that is shorter than the promise to the buyer’s own customers, or a remedy that depends on a factory response time the buyer cannot enforce. The gap is invisible until the first field failure, and by then the buyer is funding the replacement out of its own margin.

The eighth is signing the framework after the tooling payment. At that point the factory knows the buyer has committed, the price is fixed, and the buyer’s leverage has gone. The agreement should be signed before the money moves, not before the first shipment.

Cost and commercial impact

The legal cost difference between the two documents is real but smaller than the risk difference. A supplier agreement drafted for a specific transaction commonly costs a few hundred to a few thousand US dollars; a manufacturing agreement with IP, tooling, NNN and quality schedules commonly costs several thousand to fifteen thousand or more, depending on the number of schedules and the amount of negotiation. Both figures are trivial against the value of a mould, a product line or a customer relationship.

The larger cost is the cost of the wrong structure, and it appears in three forms. The first is the loss of the asset: a mould that cannot be recovered, a design that appears in a competitor’s product, a technical file the factory will not release. The second is the cost of renegotiation at a point of weakness: a company that needs a second production run and has no framework agreement negotiates its IP position while its customer is waiting. The third is the cost of a dispute conducted in the wrong forum, where a favourable judgment is unenforceable and the only realistic remedy is to stop paying, which exposes the buyer to a counterclaim.

There is also a timing cost that companies underestimate. A manufacturing agreement takes longer to negotiate than a supplier agreement, and the negotiation usually runs in parallel with tooling. Starting it after the tooling purchase order has been placed removes most of the leverage, because the factory knows the buyer has already committed. The agreement should be signed before the first tooling payment, not before the first shipment.

A fourth cost is less visible and often larger. A buyer without a change-control clause cannot tell whether the unit it receives in month six is the unit it approved in month one. The cost is not a lawsuit; it is yield loss, field returns and a product that drifts away from its specification while the buyer’s engineering team looks for the cause in its own design.

Practical checklist

Where to go next

The contract structure decides what the company owns when the relationship ends, and the end is the only moment at which the choice is tested. A company that has thought about title, tooling and the right to inspect before the first payment has bought an option it will probably never need. A company that has not will negotiate all three from a position of dependence.

Geason Law Offices advises foreign AI and hardware companies on manufacturing and supply agreements, tooling ownership and IP allocation with Chinese factories. The commercial contracts practice page describes how the firm structures these agreements, and quality control in Chinese manufacturing deals with the specification, inspection and warranty clauses that sit inside them. For the IP layer, hardware IP strategy in China covers registration and the protection of designs before disclosure, and China tooling cost and ownership explains the mould economics that make the tooling clause worth negotiating. Companies at the factory selection stage should read how to select and vet factories in China, because the counterparty’s ownership and litigation record is part of the contract’s value.

Frequently Asked Questions↑ Back to article

When should I use a purchase agreement vs a manufacturing contract?

Use a purchase agreement (or supplier agreement) when buying standard, off-the-shelf products or components from a factory's catalog with little to no customization. Use a manufacturing contract when you're developing custom products, providing designs or tooling, or engaging in OEM/ODM arrangements where IP ownership, tooling rights, and production specifications are critical.

What are the most important clauses in a China manufacturing contract?

The most critical clauses in a China manufacturing contract include: IP ownership (clearly defining foreground and background IP), tooling ownership and mold return provisions, NNN (non-disclosure, non-use, non-circumvention) protections, quality standards with acceptance procedures, payment milestones tied to deliverables, product change control processes, and dispute resolution with China-executable governing law.

Can a supplier agreement protect my IP in China?

A basic supplier agreement provides limited IP protection — typically only confidentiality provisions. It does not address IP ownership of custom designs, tooling, or manufacturing processes. For meaningful IP protection in China, you need a manufacturing contract with explicit IP ownership clauses, NNN provisions, and non-compete/exclusivity terms that are enforceable under Chinese law.

What is the difference between an OEM agreement and a supply agreement?

An OEM (Original Equipment Manufacturing) agreement is a type of manufacturing contract where you provide the design, specifications, and often the tooling, and the factory produces the product according to your requirements — with the product bearing your brand. A supply agreement (or purchase agreement) is a simpler arrangement where you order existing products from a supplier's catalog, with little customization and no transfer of design IP.

Do I need both a supply agreement and a manufacturing contract?

Not necessarily, but many hardware companies use a hybrid approach: a framework manufacturing agreement that governs the overall relationship (IP, tooling, quality standards, general terms) combined with individual purchase orders that specify quantity, price, and delivery for each production run. This gives you the protection of a manufacturing contract with the flexibility of PO-based ordering.

What is an NNN clause and why is it important for China manufacturing?

An NNN clause is a non-disclosure, non-use, and non-circumvention agreement — three protections in one. Non-disclosure prevents the factory from revealing your confidential information. Non-use prevents the factory from using your designs, technology, or IP for anything other than making your products. Non-circumvention prevents the factory from going around you to deal directly with your customers or suppliers. NNN clauses are critically important in Chinese manufacturing because they are enforceable under Chinese law and provide broader protection than standard NDAs. A well-drafted NNN clause should specify clear penalties (liquidated damages) for violations to give the contract real teeth.

What governing law should I use for my Chinese manufacturing contract?

For maximum enforceability against a Chinese factory, Chinese law with China-based arbitration is generally recommended. This is because Chinese courts and arbitration bodies can directly enforce awards against Chinese companies and their assets within China. Foreign court judgments are extremely difficult to enforce in China because China does not have reciprocal enforcement treaties with most Western countries. Arbitration through CIETAC (China International Economic and Trade Arbitration Commission) or HKIAC (Hong Kong International Arbitration Centre) are common choices because their awards are enforceable in China and internationally under the New York Convention. The key principle is that your dispute resolution mechanism must be executable where the factory's assets are located.

Geason Law Offices AI Industry Research Group | Reviewed by He Qianggao (Lawyer License No. 14403201310092337, 10+ years of practice)