A hardware company with a Chinese supply chain is exposed in a way that is easy to describe and hard to fix. One factory, one mould, one battery cell supplier and one port can each stop revenue, and none of them appears on the balance sheet as a risk until it fails. The company’s financial statements show inventory and fixed assets. They do not show that the mould is in a building it does not control, that the second source exists only in a slide, or that the component with the longest lead time has a single approved manufacturer.

Risk management in this setting is not a compliance exercise. It is the work of deciding which failures the company can absorb, which it must prevent, and which it must be able to survive for a defined period. That decision has a cost, and the cost has to be compared with the loss it avoids rather than with zero.

The framework below is built for a company that manufactures in China and sells elsewhere, typically with one contract manufacturer, a bill of materials of a few hundred lines, and a product cycle measured in quarters. It covers how to map the chain, how to score what the map reveals, how to make dual sourcing real, how to size buffer stock, how to protect tooling through escrow and access rights, and how to draft step-in rights that a substitute manufacturer can actually use.

Six exposures, one balance sheet

The risks that matter to a hardware company fall into six groups. Each reaches the profit and loss account through a different mechanism, and each is detected by a different signal.

ExposureHow it reaches the P&LEarly signalPrimary instrument
Supplier financial and operational failureLost output, expediting cost, requalificationDelayed payments to sub-suppliers, sudden discounting, staff departuresFinancial monitoring, guarantees, dual sourcing
Quality failureRework, scrap, warranty, recall, brand damageRising defect rates, process deviations, supplier substitutionsSpecification control, inspection, change control
Delivery and lead time failureMissed launch, air freight, customer penaltiesLate material, capacity booking, queue at the factoryLead time mapping, buffer stock, allocation clauses
IP and trade secret lossLost margin, competitor entry, legal costTooling reuse, unexpected clones, staff poachingNNN, tooling control, registration, compartmentalisation
Geopolitical and regulatory changeTariff cost, blocked shipments, lost market accessPolicy consultation, licence denials, customs queriesScreening, classification review, scenario planning
Physical disruptionIdle capacity, destroyed tooling, months of recoveryWeather alerts, power notices, labour disputesGeographic spread, duplicate tooling, insurance

The six are not independent. A quality failure that requires a recall consumes the cash that would have funded the next production run, which turns into a supplier financial problem. An export control change forces a component substitution, which creates a quality risk because the new part has not been through a full qualification. A port closure delays a shipment, which triggers a customer penalty, which strains the relationship with the factory. The map has to show the links, not just the nodes.

Building the risk map

A risk map is a list of nodes with attributes, not a diagram. The diagram is useful for a board presentation; the list is what an operations team can act on.

Start from the bill of materials rather than from the supplier list. Every line item has an approved manufacturer, and many have only one. The factory’s purchasing team knows the approved manufacturer list, but it may not disclose the sub-suppliers behind it, and the sub-suppliers are where the single points usually sit: the battery cell, the connector, the specific grade of steel, the anodiser, the PCB house, the firmware toolchain.

For each node, record seven attributes: annual spend, lead time, number of approved sources, whether an alternative is qualified, geographic location, regulatory exposure and IP exposure. The last two are the ones companies omit. A component made in one region may be exposed to a tariff change; a component whose design the buyer supplied may be exposed to copying if the sub-supplier also serves competitors.

Tier two visibility is the hard part and the part that produces the surprises. Three mechanisms make it practical. The first is contractual: the manufacturing agreement should require the factory to disclose its approved manufacturer list and any change to it, and to flow the buyer’s confidentiality and change-control obligations down to sub-suppliers. The second is a site visit: a day at the sub-supplier’s plant answers questions that a spreadsheet cannot. The third is a bill of materials audit: compare the components in the units received against the approved list, because substitution is common during shortages and is usually invisible in the paperwork.

The map should also include the non-material nodes that stop production when they fail: the moulding shop that holds the tool, the test laboratory that performs the qualification, the certification body that holds the certificate, the freight forwarder that controls the booking, and the port of departure. These are not suppliers in the purchasing system, and they are often the weakest links.

RiskNode in the chainLikelihoodImpactOwnerInstrument
Single battery cell sourceTier two cell makerMediumCriticalProcurementDual qualification, twelve-week buffer
Mould held at the factoryToolingLowCriticalOperationsAsset register, power of attorney, data escrow
Fire at the contract manufacturerTier one plantLowCriticalOperationsDuplicate tool, insurance, alternate site
Component substitutionTier one processHighHighQualityApproved BOM, change control, incoming inspection
Export control changeTier two semiconductorMediumHighLegalScreening, alternate part, licence review
Port congestionLogisticsMediumMediumLogisticsAlternate port, air freight agreement
Certification lapseComplianceMediumCriticalComplianceCertificate schedule, factory audit
Loss of a key process engineerTier oneMediumMediumSupplier managementNamed deputy, documented process

The register is only useful if each row has a name against it. A risk owned by a committee is a risk nobody checks. The owner should be the person who would have to run the response, not the most senior person available.

Scoring and prioritisation. Scoring turns a long list into a short one. The method does not need to be sophisticated; it needs to be consistent enough that two people scoring the same risk reach a similar answer.

Score likelihood on evidence rather than instinct. Low means the event has not occurred at this node in the recent past and the conditions for it are absent. Medium means it has occurred in the industry or at a comparable node, or the conditions are present. High means it has occurred at this node or the conditions are clearly present now.

Score impact in days of lost revenue and in customer commitments, not in adjectives. A critical impact is one that stops a product line for longer than the buffer can cover, or breaches a customer contract with a penalty, or creates a safety exposure. A high impact stops a line within the buffer. A medium impact delays a shipment without stopping the line. A low impact is absorbed by existing stock.

ImpactDefinitionTypical response
CriticalLine stops beyond buffer, or safety or regulatory exposurePrevent, duplicate or insure; board visibility
HighLine stops within buffer, or a key customer is affectedDual source or increase buffer
MediumShipment delay absorbed by stock or reschedulingMonitor, qualify an alternate
LowNo customer effectAccept and review annually

The priority bands follow from the combination. High likelihood with critical impact is a project, not a monitoring item. Low likelihood with critical impact is an insurance and redundancy question, because prevention is uneconomic and the only sensible response is to survive the event. High likelihood with low impact is a process problem for the quality team. Low likelihood with low impact is accepted and reviewed.

Two concentration metrics belong beside the scores. The first is the share of spend or volume at the largest factory, and the share at the largest component source. The second is the share of revenue that depends on a single node, which is usually much higher than the share of spend. A connector that costs a few cents can carry a large share of revenue if it is the only approved part.

Dual sourcing: the second source that actually works

Dual sourcing is the most recommended and least implemented control in Chinese manufacturing. The reason is economic: a second source that receives no volume is not a source, it is a supplier record. It has no line time, no trained operators, no recent quality data and no reason to prioritise the buyer when capacity is tight.

The first decision is where to apply it. Dual sourcing every line item is wasteful. Apply it to the components and nodes that can stop the line: custom semiconductors, battery cells, displays, moulded parts with a single tool, and any part with a lead time longer than the company’s buffer. For commodity parts, approved alternates in the bill of materials are usually enough.

The second decision is the split. A common structure is a primary share of 60 to 80 per cent and a secondary share of 20 to 40 per cent, held on a running order rather than a dormant qualification. The exact split depends on volume and price, and the principle is that the second source must have enough volume to stay competent and enough revenue to care. A five per cent allocation satisfies a procurement policy and fails when it is needed.

The third decision is tooling. A second factory usually needs its own tool, because moulds are rarely portable between machines and a mould moved to a new press needs re-qualification. Duplicate tooling doubles the fixed cost and is the single largest line in a dual-sourcing budget. Where duplication is uneconomic, the alternative is a second moulder that can run the same tool, which requires the tool to be designed for portability and the first factory to release it.

The fourth decision is design. Dual sourcing is much cheaper if the product is designed for it: standard interfaces, modular assemblies, approved alternates at the component level, and no custom part where a standard part would do. A design that depends on one proprietary module cannot be dual sourced at a sensible cost, and the decision has already been made by the engineering team.

Component typeSingle source riskRecommended postureMain cost
Custom processor or system on chipHighDesign an alternate or hold a long bufferNRE and twelve to sixteen weeks of stock
Battery cellsHighDual qualification with separate cell makersTest, certification and safety review
Injection mouldingsMediumDuplicate tool or a second moulder able to run the toolTool cost and re-qualification
PCB assemblyMediumSecond EMS provider with the same test fixturesStencils, fixtures and first-article approval
Connectors and passivesLowApproved alternates in the bill of materialsEngineering time
Display panelsMediumSecond panel source or a buffer at the module levelRe-qualification and yield learning
Firmware toolchainHighSource code escrow and a build environment the buyer controlsLicence negotiation and maintenance

Dual sourcing across countries is a separate decision with a different cost base. A second site in Vietnam, India, Mexico, Malaysia or Indonesia reduces concentration, and it also introduces a new certification programme, a new customs classification, a new quality system and a new set of lead times. The sensible sequence is to move final assembly or a low-complexity subassembly first, prove the quality system, and then move more. Moving the whole product at once transfers the risk rather than reducing it.

The contract has to support the strategy. Three clauses do most of the work: an allocation clause that states the buyer’s priority during a shortage, a capacity reservation that commits the factory to hold line time, and a transfer clause that obliges the factory to release tooling and technical data to a nominated alternate. Without the transfer clause, the second source is a plan the buyer cannot execute.

Buffer stock: sizing, location and the cost of holding

Buffer stock is the control that works when prevention fails. It is also the control that companies size by habit rather than by analysis.

The size should follow from two variables: how long it takes to replenish, and how much the replenishment time varies. A part with a four-week lead time that reliably arrives in four weeks needs less cover than a part with a four-week average that sometimes takes ten. The buffer covers the difference between the normal and the bad case, not the normal case itself.

A workable tiering is four to six weeks for standard components with reliable suppliers, eight to twelve weeks for critical or long-lead components such as custom semiconductors and precision machined parts, and twelve to sixteen weeks for components with a single source or high geopolitical exposure. The tiers should be reviewed when the lead time, the source count or the trade environment changes, not once a year.

Location matters as much as quantity. Stock at the factory is cheap to hold and useless if the factory is the node that fails. Stock at a third-party warehouse in China is accessible but sits behind the same border. Stock at a bonded warehouse defers duty but not the physical risk. Stock at the destination market protects the customer but costs more and may be obsolete before it is used. Most companies need a mix: consignment at the factory for high-runner parts, a buffer at an independent warehouse for single-source parts, and finished goods at the destination for the products with the shortest selling season.

Consignment and vendor-managed inventory shift the ownership question but not the risk. If the supplier owns the stock until it is consumed, the buyer avoids the working capital charge and the obsolescence, and the supplier prices both into the unit cost. The contract should say who insures the stock, who bears shrinkage, how the stock is counted and what happens on termination. A consignment arrangement without a count procedure is an argument waiting for an audit.

The cost of holding is real and should be stated in the business case: working capital, storage, insurance, handling, obsolescence and the risk of a design change that makes the stock unusable. Against that cost is the expected loss from a stock-out, which is the revenue at risk multiplied by the probability and duration of the interruption. For a single-source part on a product with a fixed launch, the expected loss usually dwarfs the carrying cost; for a commodity part with three qualified sources, it usually does not.

TierTypical examplesCoverUsual locationReview trigger
A, critical single sourceCustom chip, battery cell, custom displayTwelve to sixteen weeksIndependent warehouse or bonded storeAny change in source, tariff or lead time
B, long leadPrecision machined parts, custom connectorsEight to twelve weeksFactory consignment plus warehouseLead time moves by more than two weeks
C, standardPassives, fasteners, standard cablesFour to six weeksFactorySupplier performance or price change

Tooling: escrow, access and recovery

Tooling is the asset that companies believe they own and discover they cannot reach. The legal position is usually clear enough: a buyer that paid for a mould and supplied the design owns it, and a work contract supports that conclusion. The practical position is different, because the mould is a heavy steel object in a building controlled by someone else.

The first instrument is the asset register. Every tool should have an asset number, a description of the part it makes, the machine it fits, the number of cavities, its location, its replacement cost, its insurance value and the date of its last maintenance. The register should be agreed with the factory and updated when a tool is moved or modified. A buyer that cannot list its tools cannot prove what is missing.

The second is marking. A mould marked with the buyer’s name and asset number is harder to treat as the factory’s own asset, and easier to identify in a warehouse or at a third-party moulder. Marking should be permanent and recorded in photographs.

The third is the power of attorney. A clause that says the buyer may remove the tool is a right; a power of attorney is the instrument that lets the buyer’s agent collect it. The two should be signed together, and the power should name the factory, the tool and the person authorised to collect.

The fourth is escrow, which in this context means two different things. Physical escrow places the tool with an independent third party, typically a moulding shop with spare capacity, under a tripartite agreement that gives the buyer access on a trigger event. Data escrow places the technical package needed to remake the tool with an escrow agent or the buyer’s own adviser. Physical escrow is expensive and rare; data escrow is cheap and common. Most companies need the data package more than they need a second physical mould, because a replacement tool can be made in weeks if the drawings, the steel specification, the cavity layout, the hot runner details, the CAM programmes and the trial reports exist. Without them, remaking the tool means re-engineering the part.

The fifth is the tripartite arrangement with a sub-tier moulder. Where the tool sits at a moulder that the factory appointed, the buyer has no contract with the moulder and no right to demand anything. The manufacturing agreement should oblige the factory to procure a written undertaking from the moulder acknowledging the buyer’s ownership and agreeing to release the tool on the buyer’s instruction. That undertaking is worth more than a clause in a contract the moulder has never seen.

InstrumentWhat it gives the buyerTriggerCostWhere it fails
Asset register and markingEvidence of ownership and locationAny disputeLowThe factory denies possession
Power of attorneyRight to collect the toolTermination or defaultLowThe factory physically refuses
Duplicate toolPhysical redundancyLoss of the first toolHighBoth tools sit at the same site
Data package escrowAbility to remake the toolInsolvency or exitMediumThe package is stale after engineering changes
Tripartite storage agreementTool held by an independent shopDefault or insolvencyMediumThe shop has no spare capacity
Step-in licenceRight to make the product elsewhereSustained failureMediumNo qualified alternate exists

The failure mode that defeats all five is a stale data package. A tool changes during production, the drawings are not updated, and the escrowed package describes a version that no longer exists. The agreement should require the factory to deposit an updated package after every engineering change, and the buyer should audit the deposit annually.

Step-in rights and the continuity contract

A step-in right is the contractual ability to take over production, or to have someone else take it over, when the factory can no longer perform. It is not a single clause. It is a bundle of rights that only works when the surrounding contract supports it.

The first element is a trigger. Workable triggers include insolvency or administration, cessation of production for more than a defined period, failure to supply for a defined period, loss of a certification required for the product, a change of control to a competitor, and a force majeure event that continues beyond the buffer. Each trigger needs an evidential test, because a right that can be exercised on suspicion will be challenged and may be restrained.

The second element is access. The buyer, or a nominated substitute manufacturer, needs a right to enter the premises, inspect the tooling and the work in progress, and remove what it owns. Access should be supported by the power of attorney and by a list of the assets to be removed.

The third element is a licence. If the substitute is to make the product, it needs the factory’s process know-how, its test methods and its fixtures, as well as the buyer’s design. The buyer owns the design, but it may not own the process. The step-in licence should grant the buyer and its nominated substitute a non-exclusive licence to the factory’s background process IP, limited to making the product for the buyer, for as long as the step-in continues.

The fourth element is cooperation. A factory in default has little incentive to help. The contract should attach a liquidated sum to each day of non-cooperation, and should make cooperation a condition of receiving the final payment or the return of the tool.

The fifth element is a qualified substitute. A step-in right is worthless if no one can step in. The substitute should be identified before the event, qualified on the product, and given enough volume to stay competent. That is the same work as dual sourcing, which is why the two controls belong in one programme.

TriggerEvidenceImmediate actionPractical limit
Insolvency or administrationCourt or registry noticeRemove tooling, activate substituteAdministrator may dispute asset ownership
Cessation of productionSite visit, written noticeMove work in progress and toolingThe factory may withhold records
Sustained failure to supplyDelivery records over the defined periodReallocate orders, claim liquidated damagesThe factory may prioritise other customers
Loss of certificationCertificate register, regulator noticeSuspend shipments, re-certify elsewhereRe-certification takes months
Change of control to a competitorPublic filing or notice obligationSuspend disclosure, review the relationshipThe clause must define control

Chinese courts enforce clear contractual rights, but specific performance of a positive obligation is not always available, and a factory that refuses access can delay enforcement for months. The practical design is therefore to rely on things the buyer can do without a court: hold the data package, hold the power of attorney, keep the substitute qualified, and structure payments so that the final tranche is payable only on cooperation. Litigation is the backstop, not the plan.

Quality, IP and regulatory exposures in the same map

The six exposures are usually managed by different teams, which is why the links between them are missed. Three examples show why they belong on one register.

A quality failure is often a supply chain event before it is a defect. A substituted component, a new sub-supplier or a rushed batch during a shortage produces the defect, and the root cause sits in procurement rather than in production. The map should record which components have approved alternates and which substitutions require approval, and the quality plan should test the boundary.

An IP loss is often a tooling event. The factory runs the buyer’s mould for another customer, or sells excess production, or moves the tool to a sister plant. The controls are the ones described above: marking, asset register, access, NNN and a liquidated sum. Registration of the design, the trademark and the software copyright in China before disclosure is the other half, because China is a first-to-file jurisdiction and a registration obtained after a leak is a remedy that arrives late.

A regulatory event is often a factory event. CCC attaches to the production premises, and both CCC and SRRC attach to the tested configuration, so a change of factory or a component substitution can invalidate a certificate that the commercial team believes is in place. The risk register should carry the certificate schedule, with expiry dates and factory names, and the change-control procedure should require a certification impact assessment for any change to the bill of materials or the production site.

The common thread is that the factory controls information the buyer needs. The contract should make that information a deliverable: the approved manufacturer list, the sub-supplier list, the certificate schedule, the tooling register, the process change notices and the quality records. A buyer that receives these documents quarterly has a map. A buyer that receives them on request has a hope.

The crisis playbook and the review cadence

A crisis is a test of decisions made earlier. The playbook exists to shorten the period between the event and the first useful action.

The first phase is assessment, within the first day or two. The company needs facts: which products are affected, which customers are exposed, how long the interruption is likely to last, and what stock exists. A small team with named roles (operations, procurement, quality, logistics, finance, legal and communications) should meet, decide a severity level, and issue a short written brief. The most common failure in this phase is waiting for complete information, which arrives after the customer has already heard the news from someone else.

The second phase is containment, over the following one to two weeks. The options are to expedite from an alternate source, reallocate existing stock to the most important customers, substitute an approved alternate part, move production to a qualified site, or buy from the spot market at a premium. Each option has a cost and a quality risk, and the decision should be recorded with the reason, because the same decision will be questioned later.

The third phase is recovery, over weeks or months. The factory is repaired or replaced, the product is re-qualified, the certificate is re-issued if the site changed, and the customer relationships are rebuilt. The register is updated with what the event revealed, and the controls are changed. A crisis that does not change the register has been survived rather than managed.

PhaseDecisionsOwnerOutput
Assessment, first one to two daysSeverity, affected products, customer exposureChief operating officerIncident brief
Containment, one to two weeksExpedite, allocate, substitute, moveSupply chain leadRecovery plan with costs
Recovery, weeks to monthsRequalify, re-source, claim, re-certifyProgramme leadUpdated register and controls

The review cadence keeps the map current. A monthly review of the metrics (defect rates, on-time delivery, lead time drift, inventory cover, supplier financial signals) catches drift. A quarterly review with each critical supplier catches relationship problems before they become commercial ones. An annual reassessment rebuilds the map from the current bill of materials, because a map built eighteen months ago describes a product the company may no longer make.

What resilience costs, and how to justify it

Every control in this guide has a carrying cost. Dual qualification consumes engineering time and reduces the volume leverage on the primary supplier. Duplicate tooling is a capital cost with no revenue attached. Buffer stock ties up working capital and risks obsolescence. Insurance has a premium. Audits and site visits consume management time.

The justification is not that resilience is good. It is that the expected loss from the unmanaged exposure is larger than the carrying cost. For each critical node, the calculation is the revenue at risk multiplied by the probability of interruption multiplied by the duration, plus the customer penalties and the recovery cost. The number is an estimate, and it is still more useful than no number, because it ranks the nodes and shows which controls pay for themselves.

The calculation usually produces a counter-intuitive result. The controls with the best return sit on the cheapest parts, because a connector that costs cents can stop a product that sells for hundreds. Lighter controls are enough on the parts with several qualified sources, where a buffer covers the residual risk. A programme that spends evenly across the bill of materials is spending too much on the safe parts and too little on the dangerous ones.

Building the programme: steps and owners

The programme is a sequence, and each step has an owner. The order matters because a control chosen before the map is a guess, and a second source qualified before the specification is stable is a rework.

StepWho is responsibleTypical timingOutput
Map the bill of materials to approved manufacturersSupply chain analyst with engineeringWeeks one to threeNode list with spend and lead time
Interview the factory on its sub-suppliersSupplier manager with qualityWeeks two to fourTier two list and gap notes
Score likelihood and impactOperations with financeWeek fourPrioritised register
Select controls and budgetChief operating officer with financeWeek fiveControl plan and business case
Qualify the second sourceEngineering and qualityThree to six monthsQualified alternate on a running order
Set buffer levels and locationsSupply chain planningWeek sixStock policy by tier
Register and mark the toolingOperations with the factoryWeeks one to fourAsset register and photographs
Deposit the tooling data packageEngineering with legalWeeks four to eightEscrow receipt and index
Amend the continuity clausesLegalAt the next contract renewalSigned amendment
Test the playbookChief operating officerAnnuallyLessons log and revised register

The qualification of a second source is the longest step, and it is the one most often started late. A three to six month qualification that begins after the first disruption arrives after the customer has already moved. The work should begin while the primary source is healthy, because that is the only time the factory will cooperate with the transfer of tooling and data.

Budget approval follows the register rather than the calendar. A control plan that asks for money without a ranked list of exposures will be cut at the first review. A plan that shows the revenue at risk behind each request, and the cost of the control against it, survives.

Where foreign companies get this wrong

The first failure is a risk map that stops at tier one. The contract manufacturer is visible, audited and visited. The cell maker, the moulder and the PCB house behind it are not, and they are where the single points sit.

The second is a second source that exists on paper. It has a supplier code, a price and a qualification certificate, and it has not run a production order in a year. When the primary fails, the secondary cannot ramp, and the buyer discovers that qualification is not the same as readiness.

The third is buffer stock sized to the average lead time. The average is the case that does not need a buffer. The buffer exists for the tail, and a stock level set at the average is exhausted on the first late shipment.

The fourth is tooling that is owned but not accessible. The contract says the buyer owns the mould, the factory holds it, and nobody has a power of attorney, a data package or an asset number. Ownership becomes a claim that takes months to enforce.

The fifth is a step-in right with no data package and no qualified substitute. The right is exercised, the buyer arrives at the factory, and there is no one to make the product because the process know-how sat with the factory’s engineers.

The sixth is insurance assumed to cover the event. Business interruption cover may exclude the supplier’s premises, product liability may exclude recall costs, and cargo cover may end at the port. The policy should be read against the map rather than against the broker’s summary.

The seventh is a register with no owner and no review. The risks are listed, the mitigations are described, and nothing changes until the next crisis produces a new list.

The eighth is treating certification as the factory’s problem. A certificate that lapses because the factory failed a follow-up inspection stops the buyer’s shipments, and the buyer discovers the dependency when the goods are already at the port.

Practical checklist

Where to go next

Supply chain risk in China is not a single legal problem, and it is not solved by a single clause. It is a set of commercial decisions about concentration, redundancy and access, recorded in contracts that give the company the right to act when a node fails. The companies that recover fastest are not the ones with the longest risk register. They are the ones that qualified the second source, hold the tooling data, and know which certificate attaches to which factory.

Geason Law Offices advises foreign AI and hardware companies on manufacturing agreements, tooling ownership and supply continuity with Chinese suppliers. The commercial contracts practice page describes how the firm structures supply and manufacturing agreements, and supplier agreement versus manufacturing contract explains the liability, tooling and IP terms that underpin the controls described here. For the certification dependency, China product certification sets out which certificate attaches to which factory and who holds it, and how to select and vet factories in China covers the diligence that should precede the first order.

Frequently Asked Questions↑ Back to article

What are the biggest supply chain risks when manufacturing in China?

The biggest risks for hardware companies manufacturing in China fall into six categories: supplier financial and operational risk (factory closures, capacity issues), quality risk (process drift, component substitution), delivery and lead time risk (material shortages, logistics delays), IP and trade secret risk (design leaks, counterfeiting), geopolitical and regulatory risk (tariffs, export controls, sanctions), and disruption risk (natural disasters, pandemics, power shortages). The relative importance of each depends on your product complexity, sourcing concentration, and market exposure.

How do I assess the financial health of a Chinese supplier?

Assess supplier financial health through multiple channels: review official business records (营业执照, annual reports via the National Enterprise Credit Information Publicity System), analyze payment patterns (delayed payments to sub-suppliers can signal distress), check ownership structure and related-party transactions, monitor order acceptance patterns (sudden eagerness for large orders may indicate cash flow pressure), and consider third-party credit reports from agencies like D&B or local credit bureaus. For critical suppliers, request audited financial statements and conduct on-site financial due diligence.

What is China Plus One and should I do it?

China Plus One is a diversification strategy where companies maintain manufacturing in China while adding production capacity in another country (Vietnam, India, Mexico, Malaysia, Indonesia) to reduce concentration risk. Whether you should do it depends on your risk profile: if you have high revenue concentration from China-made products, long tooling lead times, or exposure to tariff-sensitive product categories, China Plus One is likely worth pursuing. Start with a small percentage of SKUs or a secondary assembly line before committing to full-scale diversification.

How much inventory buffer should I hold for China supply risk?

The right inventory buffer depends on your lead time, demand variability, and risk tolerance. A baseline is 4-6 weeks of safety stock for standard components with reliable suppliers, 8-12 weeks for critical or long-lead-time components (e.g., custom semiconductors, precision machined parts), and 12-16 weeks for components with single-source dependency or high geopolitical exposure. Use a risk-adjusted approach: categorize components by criticality and supply risk, then tier your safety stock accordingly. Regularly review and adjust buffers as supply conditions change.

How do geopolitical tensions affect hardware supply chains in China?

Geopolitical tensions affect hardware supply chains through multiple channels: tariff changes that alter landed cost economics, export controls (especially for semiconductors, advanced materials, and dual-use items) that restrict technology transfer and component availability, sanctions that may target specific companies or industries, customs scrutiny that slows port clearance, and reputational risk from consumer or political backlash. Hardware companies are particularly exposed because of complex BOMs that may span multiple jurisdictions and because products often contain dual-use technologies.

What is supplier diversification and how should I approach it in China?

Supplier diversification means having multiple approved suppliers for critical components or production capacity rather than relying on a single source. In China, this can take several forms: dual-sourcing from two different factories in different regions, having a primary supplier and a backup supplier that is pre-qualified but not currently producing, or maintaining production capacity at both a Chinese factory and a China Plus One location. The goal is to reduce the risk that a single supplier failure stops your entire production. The right diversification strategy depends on your volume, the cost of qualifying additional suppliers, and the criticality of the component — for the most critical parts, the cost of dual-sourcing is justified by the risk reduction.

What is a supply chain crisis response plan and why do I need one?

A crisis response plan is a documented playbook for what to do when a major supply chain disruption occurs — such as a factory closure, port shutdown, material shortage, or geopolitical event. It should include a crisis team roster with roles and contact information, pre-vetted alternative suppliers that can be activated quickly, inventory reserve strategies, communication templates for customers and stakeholders, escalation procedures, and decision frameworks for when to take specific actions. Having a plan dramatically reduces response time and improves decision-making during a crisis, when every hour of delay costs money. Companies that have response plans recover faster and with less financial damage than those that improvise.

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