Most sourcing problems are not manufacturing problems. They are decisions taken at one stage without knowing what they cost at a later one — a fabric chosen before anyone checked which factories can buy it at your quantity, a factory chosen before anyone priced the freight its location implies, a sample approved before anyone wrote down what it is evidence of.
This guide walks the lifecycle in the order it happens and treats it as one chain rather than eight departments. Per stage: what happens, what gets decided, who decides, what it produces, and what happens when it is skipped. That last part carries the most weight, because the cost of a skipped stage is almost never paid at the stage that skipped it.
The eight stages, and the one decision each owns
Sourcing is usually drawn as a pipeline of activities. It is more useful as a sequence of decisions, because activity can be delegated and a decision cannot: somebody owns it, and somebody has to be able to point at the artefact that records it.
| Stage | Decision it owns | Who decides | What it produces |
|---|---|---|---|
| Brief | What the product is, and what you will trade to get it | The brand | Target landed cost, quantity, market, non-negotiables |
| Product development | How the product is constructed | The brand, advised by product expertise | A tech pack with measurements and tolerances |
| Materials | Which material, from which supplier, on what price basis | The brand, from a tested shortlist | Approved references with supplier, price basis, test results |
| Factory matching | Which factory makes it, on which terms | The brand, from a vetted shortlist | Normalised quotes, a shortlist, a signed order |
| Sampling | Whether the factory has proved it can make this | The brand, on documented review | A sealed approved sample and the comments behind it |
| Production | What to do about each deviation from plan | The factory runs it; the buyer decides exceptions | A dated plan, progress against it, an exception record |
| Quality | Whether this shipment may ship | The buyer, on inspection evidence | Inspection reports, a pass or fail, shipment approval |
| Logistics | How the goods move, and what they cost delivered | The buyer, on a current quote | Booking, customs documents, delivered goods, a landed cost |
One decision per row. Where two stages look like they share a decision, one is reviewing the other — and that seam is where fragmented processes lose things.
Stage 1 — the brief
The brief is the only document written before money is committed, which makes it the cheapest place to be precise. A brief that works states what the product is and what it competes with, the target retail price and the target landed cost, the first order quantity and whether a reorder is realistic, the market it enters, the hard non-negotiables — a certification the channel requires, a labelling rule, a packaging format, a date tied to a season — and, crucially, what is genuinely flexible.
The decision the brief owns is not "what do we want". Everyone wants the best version at the lowest price. It is what you are willing to trade: a heavier fabric for a shorter lead time, a simpler closure for a lower unit price, a smaller first order for full sampling. That is the information a supplier needs before it can offer you the cheaper option that would have been fine.
Skipped, and what it costs later. When the brief is a conversation and a reference image, every supplier answers a slightly different question. You find out at quote comparison, where the numbers cannot be made comparable after the fact — so you either choose on a false comparison, or spend a second full round of supplier time re-quoting against the specification you should have written first.
Stage 2 — product development
Product development turns the brief into a specification a factory can quote from and be held to. The output is a tech pack: dimensioned drawings, a bill of materials naming every component, construction detail, a measurement chart with tolerances graded across the size range, colour references and the method by which colour will be judged, labelling and packaging requirements, and artwork with its placement and application method.
The commercially important part is the tolerances, and it is the part most often missing. A tolerance is not a technical nicety; it is the range you have agreed to accept. State none and the factory applies its own, which you will meet for the first time in an inspection report you cannot dispute. For apparel and soft goods the grading rules matter as much as the base measurements — the apparel sourcing guide works through how a size set and a fit block change what you are actually specifying.
Skipped, and what it costs later. Without a tech pack a factory prices its own default construction against your photograph. The quote is real; it is just not a quote for your product. The gap surfaces at the first sample, usually after the order is placed, so the price moves upward with the leverage on the other side. The quieter cost: inspection has no criteria. You cannot fail a shipment against a specification that does not exist.
Stage 3 — materials
Materials decide more than most brands expect. On many products the material is the largest single line in the unit cost. It also sets weight and volume, which set freight; composition, which drives tariff classification and testing; and availability, which sets the earliest date production can start. A casual material decision is a cost, compliance and schedule decision taken at once.
What gets recorded has to be a specific material: named supplier, article reference, and a price with its full basis — currency, unit, quantity band, delivery term, and the date it is valid until. A material price without that basis is not a price, and two of them cannot be compared. Test before the order rather than after, against how the product will be used and against what the destination market requires.
Two different things get called compliance here, and blurring them is expensive. Regulation is an obligation. REACH — Registration, Evaluation, Authorisation and Restriction of Chemicals — is an EU regulation that entered into force on 1 June 2007, and ECHA is explicit that it places the burden of proof on companies: they must identify and manage the risks of the substances they manufacture and market in the EU. ECHA also notes it reaches articles such as clothes, furniture and electrical appliances, not only industrial chemicals. A certification scheme is a choice: OEKO-TEX, GOTS or GRS are things a buyer or supplier elects to do, and holding one does not by itself discharge a legal obligation. EU manufacturing compliance covers the distinction properly.
Skipped, and what it costs later. The classic failure is substitution: a material approved from a photograph, the factory buys the nearest available thing when the order runs, and nobody notices until final inspection or until customers do. Every option then is bad — accept it, remake and pay for material and labour twice plus the lead time, or discount. What prevents it is dull and cheap: a physical reference, a named supplier and a test report, filed against the order.
Stage 4 — factory matching
This stage is usually described as finding suppliers. Finding them is the easy half. The work is issuing one request to a shortlist, checking capacity exists in the window you need, and making the quotes genuinely comparable — same quantity, construction, material, delivery term, currency and validity date — before anyone compares them. Most of the value here is created before the first negotiation, in normalisation.
Capability has to be assessed where it is real. No factory is simply "good at apparel" or "good at homeware". A factory is good at a product type, in a material family, at a quantity band, on machinery it owns. Success in fine-gauge knitwear is not evidence for technical outerwear, and a reference for a 20,000-unit programme says little about how your 800 units get treated. Ask for a comparable product at a comparable quantity, and ask which line will run it.
Vetting is a separate exercise from quoting and runs in parallel: legal existence and ownership, whether the counterparty produces or resells capacity, audit and certification status with issue and expiry dates rather than a logo, references you actually contact, and what a live video walk-through of the relevant line shows. How to vet an overseas manufacturer covers the sequence, the factory vetting guide goes deeper on evidence, and how to avoid Alibaba scams covers the failure modes specific to platform-sourced suppliers. For a private-label version of something a factory already makes, finding a private label manufacturer is the narrower path; finding a manufacturer for your brand is the general one.
Country choice belongs here rather than in logistics, because it moves duty, transit time and the practical cost of being wrong at once. Sourcing from Vietnam works through one region in that shape.
Intellectual property terms belong here too, and specifically before the tech pack is sent. A tech pack is the transferable version of your product; once it has gone out, the leverage to agree terms has gone with it. NNN agreements in sourcing explains why non-disclosure alone is usually the wrong instrument for manufacturing, and what non-use and non-circumvention add.
Skipped, and what it costs later. Choosing on unit price is the default failure, and it is rarely a pricing mistake — it is a comparison mistake, because the cheapest quote usually includes the least. The real cost of the wrong factory is the extra sampling rounds, the defect rate you then inspect against, and the re-sourcing you do under a date you have already promised. Relationships compound the other way, which is the argument in relationship-based sourcing.
Stage 5 — sampling
Sampling is where the specification becomes real, and it has a shape: request, sample received, measured against the chart, fit or function review, written comments, revision, approval. Several rounds are normal. Unlimited rounds are a signal — usually that the tech pack is ambiguous or the factory is guessing.
The sample types are not interchangeable, and treating them as though they were is the commonest error in the stage. A proto proves the shape and the idea. A fit sample proves the measurements and the grading. A pre-production sample proves the product in the correct material, with the correct trims, from the factory and ideally the line that will run the order. Approving a proto as though it settled the material question is how a specification quietly loses its authority.
Approval only means something if what was approved is identified: sealed, labelled, dated, and referenced by the inspection protocol that will later be run against it. What a golden sample is covers what the reference sample has to carry to be useful in a dispute.
Skipped, and what it costs later. Production against an unapproved sample makes the shipment your first sample — arriving after material, labour and freight are paid. Compressing sampling to protect a launch date is a trade that usually loses; the honest version is a smaller first order with full sampling rather than a full order with none.
Stage 6 — production
Production is the stage a buyer does least in and needs most visibility into. The order becomes a plan with dates: material in-house, cutting or first operation, assembly, finishing, packing, ex-factory. Progress is checked against that plan, not against a general sense of how things are going.
The decision this stage owns is what to do about each deviation. When production goes to plan there is nothing to decide, which is why the stage looks passive; its whole value is how quickly a deviation reaches somebody allowed to decide about it. A late material delivery is not a quality problem yet. It becomes one when the factory absorbs the delay by compressing finishing to protect the ship date, and nobody was told the plan had changed shape.
Skipped, and what it costs later. With no plan there are no exceptions, only surprises, and they arrive where the remaining options are air freight, a late delivery, or goods finished in a hurry. All three cost more than the phone call a dated plan would have triggered weeks earlier.
Stage 7 — quality
Quality is not one event at the end. In-line inspection during production catches a systematic problem while the remaining units can still be made differently; final random inspection before shipment decides whether this lot may go. Compliance verification — the tests and documents the destination market requires — sits alongside both, and shipment approval is the decision they feed.
Attribute sampling inspection has an international standard behind it. ISO 2859-1 covers sampling procedures for inspection by attributes, part 1 being sampling schemes indexed by acceptance quality limit (AQL) for lot-by-lot inspection. One current detail matters: ISO records the 1999 edition as withdrawn in January 2026 and revised by ISO 2859-1:2026, so a protocol naming the older edition is worth re-reading with your inspection provider rather than re-signing out of habit.
What an acceptance sampling scheme gives you is a defensible rule for accepting or rejecting a lot from a sample. What it does not give you is a promise about the units nobody looked at. That is why the classification of defects — which faults are critical, major and minor for your product — and the acceptance limits have to be agreed in writing before the inspection. Agreed in advance, the report is a decision; agreed afterwards, it is an argument. Quality control checks in manufacturing covers the inspection types, the manufacturing quality control guide covers building the system, and reducing supplier quality risk covers the part that happens before any inspector arrives.
Skipped, and what it costs later. Final inspection is the last moment at which a defect is the factory’s problem. Once the goods ship, the identical defect is your inventory — and you have paid the unit cost, the freight and the duty on it. No other stage changes who pays as sharply.
Stage 8 — logistics
The physical part is a sequence: booking, pickup, export clearance, main carriage, arrival, import clearance, inland transport, warehouse, delivery. The commercial part is one decision taken much earlier — on what terms the goods are sold — and it determines which of those steps you are paying for separately and which the supplier already priced in.
The Incoterms rules are the vocabulary for that. The ICC publishes eleven rules, used in trade contracts since 1936 and updated for Incoterms 2020, which allocate the tasks, costs and risks of delivery between seller and buyer; in the 2020 edition all costs for a rule are collected in article A9/B9. UK customs guidance groups them as seven rules for any mode of transport — EXW, FCA, CPT, CIP, DAP, DPU and DDP — and four for sea and inland waterway transport: FAS, FOB, CFR and CIF. At one end, EXW has the seller merely make the goods available at its own premises, with no duty to load; at the other, DDP has it deliver them cleared for import, duty paid.
The practical consequence is the one sourcing decisions get wrong most often: a quote is a price and a term, and two prices under different terms are not two comparable numbers. An EXW price and a DDP price for identical goods can rank in either order once the missing steps are added. That comparison is the subject of what landed cost is and, at length, the landed cost guide.
Duty is not a fixed percentage of a shipment; it is a consequence of three things about the goods. The European Commission is direct that the duties payable — and related requirements such as import or export certificates — depend on tariff classification, and that tariff treatment is decided by classification together with customs value and origin, origin being the "economic nationality" of the goods rather than the country they shipped from. Where the answer must be certain in advance, a Binding Tariff Information decision is a legal decision from an EU customs authority on a product’s classification, generally valid for three years and binding on all EU customs administrations and on the holder. Lawful design-time levers exist and are worth knowing before a specification is frozen: tariff engineering strategies covers them.
Skipped, and what it costs later. Leave the delivery term, the freight quote and the duty position to the end and landed cost is discovered on an invoice. A discovery is not a decision: by then the material, factory, quantity and port are fixed, and the only variable left is your margin.
Why the eight stages are one chain
The expensive thing about a fragmented process is not the handovers. It is that nobody holds the constraint graph. Each decision gets taken by a competent person with good information about their own stage and no view of what their choice just eliminated three stages downstream.
brief → construction → material → eligible factories → achievable quality → weight, volume, HS code → freight and duty → landed cost
Left to right is what a decision constrains. Right to left is what a landed-cost target is actually asking you to change.
| The decision | What it silently constrains | Where the cost shows up |
|---|---|---|
| A fabric with a high minimum order from a single mill | Which factories can buy it at your quantity, and material lead time inside the plan | Unit price at quote, then the ship date |
| A factory chosen on unit price in a distant inland region | Inland haulage to port, transit time, and how practical a re-inspection or rework is | Freight, working capital, rework |
| A construction change agreed after sample approval | The tolerance inspection runs against, and the sealed sample’s authority | Inspection disputes, and defects that are nobody’s fault |
| A component that changes the product’s composition | Its tariff classification, and any testing the new composition requires | Duty and testing on every future shipment |
Every row has the same shape: the decision is cheap and local, the consequence expensive and somewhere else. All four are knowable when the decision is made, and none is visible in the document where it gets written down.
What it costs to fix a decision late
The cost of a change is not proportional to the size of the change. It is proportional to what has already been committed when it is discovered. The same corrected detail — a measurement, a material, a label — costs four different things depending on when it surfaces.
| Found at | Already committed | What the fix consumes |
|---|---|---|
| The brief or tech pack | Time only | A document revision, and a re-quote if the price basis moved |
| Sampling | One sample and a round of factory time | Another round and its lead time — the date moves before the money does |
| Production | Material bought and cut, capacity booked, deposit paid | Material and labour paid twice, or an accepted compromise on the product |
| After shipment | Unit cost, freight, duty, clearance — and the selling window | Discount, return, in-market rework or written-off stock, none of which refunds freight or duty |
Deliberately no multipliers. How much worse each row is depends on the product, the material and the order size, so any single figure would be invented. The ordering does not depend on any of that.
A readiness check before the first order
- Before the PO goes out
- The brief names a target landed cost, and states what you are willing to trade.
- The tech pack carries measurements with tolerances, and the tolerances are yours rather than the factory’s default.
- Every material has a named supplier, a price with its full basis and validity, and any test result the destination market requires.
- Quotes were normalised to the same quantity, construction and delivery term before comparison.
- The factory that quoted is the factory that will produce, and you know which line.
- Confidentiality, non-use and manufacturing terms were agreed before the tech pack was sent.
- The approved sample is sealed, dated, identified, and named in the inspection protocol.
- The inspection standard, defect classification and acceptance limits are agreed in writing.
- The production plan has dates for material in-house, start of production and ex-factory, and somebody checks them.
- Freight and duty come from a current quote for this shipment, and you know which figures are still estimates.
How Library of Trade runs it
We run these eight stages as one managed process rather than eight conversations. A brief comes in; AI does the work scale rewards — structuring it, searching materials and suppliers, checking capacity, normalising quotes, comparing cost and freight options, and surfacing where a decision at one stage has moved a number at another. Then the judgment happens: our expert team evaluates the recommendation, negotiates and decides, and supplier partners execute. Nothing enters production because a model produced it.
Two things we deliberately are not. We are not a marketplace and not a directory: you contract directly with your suppliers and keep your designs, specifications and relationships. And we do not replace your judgment — the comparison arrives structured so the trade-off is visible, which is the part a fragmented process cannot show you. How Library of Trade works sets out the mechanics.
Topics covered
- The eight sourcing stages
- Briefs and specifications
- Material and factory decisions
- Sampling and approval
- Quality and inspection
- Freight, duty and landed cost
- The cost of deciding late