Sourcing, Supply Chain & Vendor Management

The make-versus-buy decision

Before choosing a supplier, decide whether you need one at all. Making something in-house gives you control over quality, timing and intellectual property, and it can be cheaper once volume is high enough to spread fixed cost across many units. Buying it in gives you access to specialist capability and equipment you would otherwise have to build, converts a fixed cost into a variable one, and lets you redirect capital and management attention to whatever you actually do best.

The decision turns on a small number of questions asked honestly. Is this something that differentiates you from competitors, or is it a commodity input that any competent supplier can provide equally well? Do you have, or can you justify building, the capability and capacity to do it properly, or would you be a mediocre and distracted version of a specialist? What does it cost fully loaded — not just the visible unit cost of doing it yourself, but the capital tied up, the management time, the quality risk of a function you have no deep expertise in? And how much does control matter here — some things genuinely need to sit inside the business because the risk of losing that control is unacceptable, regardless of the arithmetic.

The trap on one side is romantic self-reliance: insisting on making things in-house because it feels more serious or more controlled, when a competent supplier would do it better and cheaper and free the business to focus. The trap on the other side is hollowing out the business by buying in everything that looks inconvenient, including the capabilities that actually make the business worth running. Neither trap announces itself; both look like sensible decisions at the time they are made.

Category management and spend analysis

Most organisations, once they actually add up what they spend with outside suppliers, find the total is both larger than expected and scattered across more suppliers than anyone intended. Category management is the discipline of grouping that spend into sensible categories — raw materials, packaging, logistics, marketing services, IT, facilities, professional services, and so on — and managing each category as a coherent whole rather than as a pile of unrelated individual purchases.

Spend analysis is the underlying exercise: pulling every invoice and purchase order into one view, however unglamorous that sounds, and asking where the money actually goes. It routinely surfaces the same findings — maverick spend outside agreed contracts, several suppliers providing near-identical things at different prices, categories nobody is actively managing because they arrived by accident rather than decision. None of this requires sophisticated tooling to start; a spreadsheet pulled from the accounting system is enough to begin. What it requires is someone willing to look, because the natural state of spend in a growing business is drift, not order.

Once spend is visible by category, each category can be assigned an owner, a strategy, and a review cadence appropriate to its size and risk — a small, low-risk category might be reviewed once a year in passing; a large or critical one deserves continuous attention. The point of the exercise is not the spreadsheet itself but the decisions it enables.

Direct versus indirect spend

Direct spend goes into what you sell — raw materials, components, contract manufacturing, the inputs that appear, transformed, in the product a customer buys. Indirect spend, sometimes called non-production spend, keeps the business running without appearing in the product itself — office supplies, software, travel, professional services, facilities, marketing production costs.

The distinction matters because the two are usually managed very differently, and both suffer from being managed the wrong way. Direct spend tends to get serious attention because it is visible in cost of goods sold and because a failure there stops production; it typically has dedicated buyers, formal contracts and close supplier relationships. Indirect spend is frequently under-managed precisely because no single failure is dramatic — it leaks out through many small, uncoordinated purchases, each individually defensible, that together represent a real and often surprising sum. A mature sourcing function treats indirect spend with the same rigour it applies to direct spend, once the category analysis above has shown how much of it there actually is.

Single, dual, multiple and parallel sourcing

The table above summarises the trade-off; it is worth walking through the reasoning, because this is one of the genuine disagreements in the field and the right answer depends on the item, not on a general theory of supply chain design.

Single sourcing — using one supplier for a given item — earns simplicity and, often, the best pricing available, because the supplier can plan around guaranteed volume and the buyer avoids splitting demand across smaller, less attractive orders. It also earns concentration risk: a fire, a strike, a bankruptcy, or a simple falling-out at that one supplier stops your supply of that item entirely, with no fallback in place. Single sourcing is defensible where the item is low risk, low value, or genuinely available from many equally good alternative suppliers who could be engaged quickly if needed — the risk of concentration is real but cheap to accept because switching is fast.

Dual and multiple sourcing trade some of that pricing and simplicity for resilience and competitive tension. The honest cost is real: splitting volume across two or three suppliers usually means none of them offers quite the pricing a single supplier taking the whole volume would offer, and running several relationships instead of one costs management time. The case for it strengthens as the item becomes more critical, as switching costs rise, and as the supplier base for that item is genuinely fragile or concentrated in one region.

Parallel sourcing, where two suppliers make literally the same part for the same production line, is a specific and expensive version of dual sourcing usually reserved for critical, high-volume components where the cost of duplicated qualification is worth the direct, ongoing performance comparison and the built-in backup capacity. It is not something a small operation typically needs, but it is worth knowing the term because larger customers or partners will use it and expect the buyer to understand what it implies for tooling and qualification cost.

There is no correct universal answer here. A category strategy should state, for each significant category, which approach applies and why — and should revisit that choice when volumes, risk, or the supplier landscape change materially.

Near-shoring, off-shoring, re-shoring and friend-shoring

These four terms describe geography as a sourcing lever, and each is sometimes presented as a settled answer when in fact each is a trade-off that depends on what the buyer actually values.

Off-shoring — sourcing from distant, typically lower-cost countries — earns lower unit cost and, in many categories, access to manufacturing capability and scale that simply does not exist closer to home. It costs longer and more variable lead times, higher inventory needed to buffer that variability, currency exposure, harder quality oversight because visits are expensive and infrequent, and exposure to whatever happens geopolitically along a long supply line.

Near-shoring — sourcing from a nearby country rather than the cheapest available one — trades some of that cost advantage for shorter and more predictable lead times, easier communication in similar time zones, and the practical ability to visit a factory without a long-haul flight. It rarely matches off-shore pricing on a like-for-like basis, and the supplier pool in a nearby country may simply be thinner or less specialised than the global market.

Re-shoring — bringing sourcing back into the buyer's own country — is usually driven by a specific failure of the off-shore model: a disruption that exposed how exposed the business was, a strategic decision that control of a critical input matters more than its unit cost, or a regulatory requirement for domestic content. The honest constraint is that domestic capability at the needed scale, quality and price sometimes has to be built rather than found, which takes time and capital and is not always successful.

Friend-shoring — sourcing from countries considered politically aligned with the buyer's own — is a response specifically to geopolitical risk rather than to cost or lead time. It reduces exposure to the specific risk of doing business with, or through, a country whose relationship with the buyer's own government might deteriorate, but it does not eliminate concentration risk generally, and it can still leave the buyer exposed to the alliance itself shifting. None of these four is a slogan worth adopting wholesale; each is a specific trade against a specific risk, and a mature sourcing strategy usually blends them by category rather than picking one doctrine for everything.

Total cost of ownership versus unit price

The unit price on a quotation is the easiest number to compare and, on its own, one of the most misleading. Total cost of ownership is the discipline of asking what an item actually costs across its whole life with you, not just what it costs to buy.

Consider, illustratively, two suppliers of the same component. Supplier A quotes a lower unit price but requires a larger minimum order quantity, ships from further away with a longer and more variable lead time, and has a slightly higher historical defect rate that generates returns and rework. Supplier B quotes a higher unit price but ships faster and more reliably, holds a smaller minimum order, and has a track record of near-zero defects. If a buyer compares only the unit price, A wins. If the buyer adds the cost of carrying more inventory to buffer A's longer and less predictable lead time, the cost of capital tied up in the larger minimum order, and the cost of handling and reworking the additional defects, the comparison can flip entirely. None of the additional costs are exotic; they are simply invisible on the face of the quotation, which is exactly why unit-price comparison is such a common and expensive mistake.

Total cost of ownership is not a single formula to apply mechanically; it is a habit of asking, for any significant purchase, what else this decision costs beyond the invoice — freight, duty, inventory carrying cost, quality cost, the administrative cost of managing the relationship, and the cost of switching away later if the choice turns out to be wrong. For low-value, low-risk purchases the exercise is not worth the time it takes; for anything material to the business, it is the difference between a sensible decision and an expensive one dressed up as a bargain.

Specification writing

A specification tells a supplier exactly what is being bought — the dimensions, materials, tolerances, performance requirements, packaging, labelling, and acceptance criteria that define whether what arrives is what was wanted. A vague specification is not a shortcut; it is a decision to let the supplier, or whoever interprets the gap most conveniently for themselves, decide what you actually meant.

The most expensive document in many businesses is a specification that seemed adequate when it was written and turned out to be silent on exactly the point that later caused a dispute — a tolerance that was assumed rather than stated, a material substitution that was technically compliant with the words used but not with the intent, an acceptance test that nobody defined precisely enough to apply consistently. Every one of these gaps becomes, downstream, either a quality problem, a contractual argument, or both, and resolving either after the fact costs far more than writing the specification properly would have cost up front.

A workable specification states what matters and, as importantly, is explicit about what does not matter, so the supplier has room to find efficient ways to meet the requirement rather than guessing which details are load-bearing. It distinguishes must-meet requirements from preferences. And it is written assuming the reader has none of the context in the buyer's head — because whoever executes the order on the supplier's side genuinely does not have that context, however well the relationship manager understands the business.

Market scanning and supplier discovery

Finding suppliers is a research exercise before it is a negotiation. Trade directories, whether general or specific to an industry, list registered suppliers and are a reasonable starting point for breadth, though listing in a directory is not itself evidence of quality. Trade shows and exhibitions let a buyer see multiple suppliers in one place, examine physical samples, and have the kind of conversation that reveals more than a written proposal does — how a representative answers an awkward question is often more informative than the answer itself. Chambers of commerce and export promotion bodies, which many governments and trade associations operate to help their domestic producers find international buyers, can be a useful and relatively low-risk source of introductions, since they typically have some interest in the credibility of the suppliers they promote. Referrals from other buyers, from industry contacts, or from within the same corporate group carry real weight because someone has already tested the relationship, though a referral should still be independently verified rather than trusted purely on the strength of who gave it.

None of these channels replaces the others; a serious sourcing exercise for anything material uses several in combination, because each surfaces a different slice of the available market and each has its own blind spots. A directory search alone will miss suppliers who rely on reputation rather than marketing; a referral-only search will miss capable suppliers outside the buyer's existing network entirely.

Pre-qualification

Before investing serious time in any supplier, a short pre-qualification step filters out the candidates who are obviously unsuitable, saving the detailed evaluation effort for those worth it. This typically covers basic legal existence and standing, relevant certifications or licences, a rough sense of scale relative to the order being considered, and any immediately visible red flags — a supplier who cannot answer basic questions about their own operation, or whose claimed capability plainly does not match their apparent size.

Pre-qualification should be proportionate. For a small, low-risk purchase, a few questions and a look at the company's registration is enough. For a supplier who will become a significant, ongoing part of the supply chain, pre-qualification is the first filter in a longer process, not the whole of it.

RFI, RFQ and RFP — and when each applies

These three instruments are often used loosely and interchangeably, which causes real confusion between buyer and supplier about what is actually being asked for. A request for information is used when the buyer does not yet know enough about the market to specify what they want precisely — it asks suppliers to describe their capabilities, capacity and general approach, and is used to narrow a wide field before anyone commits to a detailed specification. A request for quotation is used when the requirement is well defined and the buyer mainly wants to compare price and terms for a known specification — it assumes the what is settled and asks only how much and how soon. A request for proposal is used when the buyer knows the problem but not the best solution, and wants suppliers to propose an approach as well as a price — appropriate for more complex purchases, services, or anything where supplier expertise should shape the solution rather than simply price a fixed spec.

Using an RFQ when an RFP was needed produces quotes that are technically comparable but miss the point, because suppliers were never asked to bring their expertise to bear on the actual problem. Using an RFP when an RFQ would have done wastes everyone's time on proposal-writing for a requirement that was already fully specified. Matching the instrument to the actual state of the buyer's knowledge is the whole skill.

Evaluation criteria and weighted scoring

Choosing among suppliers on price alone is easy and frequently wrong, because price is only one of the things that determines whether a supplier relationship actually works. A weighted scoring model makes the trade-offs explicit: criteria such as price, quality, delivery reliability, capacity, financial stability, and service are each given a weight reflecting their importance to this particular purchase, suppliers are scored against each criterion, and the weighted scores are totalled to produce a ranking.

The value of the exercise is less the final number than the discipline of deciding, before seeing any bids, what actually matters and by how much — because deciding the weights after seeing the bids invites the buyer to reverse-engineer weights that justify whichever supplier they already preferred. That reversal is one of the more common ways scoring is gamed, deliberately or not. Other common distortions include criteria that are technically scored but effectively cannot distinguish between suppliers because everyone scores similarly on them, wasting weight on a distinction that does not exist; scorers who know which supplier submitted which bid and score accordingly, whether consciously biased or not; and suppliers who learn to write proposals that hit the scoring criteria as stated without actually being strong on the underlying capability the criteria were meant to measure. A scoring model is a tool for structuring judgement, not a substitute for it, and its outputs are worth sense-checking against plain professional instinct before a decision is finalised.

Samples, factory audits and third-party inspection

A sample shows what a supplier can produce under the most favourable conditions, with their best people paying close attention, on a small quantity they know is being judged. It is a necessary check and a poor predictor of what a full production run, made under normal conditions by ordinary shift workers under ordinary time pressure, will actually look like. Treating a strong sample as proof that production quality will match it is one of the most common and avoidable disappointments in sourcing.

A factory audit closes some of that gap by examining the actual production environment — the equipment, the process controls, the workforce conditions, the quality system — rather than relying on a curated sample. Audits range from a brief visual walkthrough to a formal, standards-based assessment against a recognised framework, and the right level of rigour depends on how critical and how risky the relationship is. Third-party inspection, carried out by an independent inspection firm rather than the buyer's own staff, is particularly useful where the buyer cannot easily travel to the supplier's location, or where independence from the buyer-supplier relationship adds credibility to the finding. None of these tools removes risk entirely; they reduce it, at a cost in time and money that should be proportionate to what is at stake in the relationship.

Capability versus capacity

Capability is whether a supplier can do the thing at all — the equipment, the expertise, the quality system needed to meet the specification. Capacity is whether they can do enough of it, fast enough, alongside everything else they are already committed to. A supplier can be genuinely capable and still be the wrong choice because their capacity is already fully allocated to other customers, leaving your orders competing for attention and slipping in priority whenever theirs gets busy.

Capacity is worth probing directly rather than assumed: what proportion of their capacity is currently committed, what happens to your orders if a larger customer needs the same production slot, and what their plans are for expanding capacity if your volume grows as expected. A supplier who is honest that they are near capacity and cautious about new commitments is often a better bet than one who enthusiastically promises capacity they do not currently have and would have to build to deliver.

Financial stability checks

A supplier's technical capability is worth little if the business itself is not going to survive long enough to deliver. Checking a prospective supplier's financial stability — through credit reports, published accounts where available, trade references, and simple questions about ownership, debt and recent history — is a standard part of due diligence for any supplier that will be significant to the business, and it is routinely skipped by buyers in a hurry.

The specific numbers to look for and how to interpret them properly is accountancy and credit-analysis work; where a supplier relationship is large enough to matter materially, a proper financial review by someone qualified to do one, rather than an amateur read of a balance sheet, is the appropriate step, and is worth the cost relative to the exposure being taken on.

The first-order problem

The first order with a new supplier carries a specific risk that later orders do not: the buyer has no track record with this supplier yet, and the supplier has every incentive to perform well on the order that is being watched most closely, whether or not that performance is representative of what will happen once the relationship settles into routine. This is sometimes called the honeymoon effect, and it means a single successful first order is weak evidence that the relationship will hold up.

De-risking the first order means keeping its size proportionate to what the buyer can afford to lose if it goes wrong, inspecting the output more closely than will be sustainable long term, building in milestones or partial payments tied to verified progress rather than paying everything up front, and treating the second and third orders, not the first, as the real test of whether the supplier performs consistently under normal conditions rather than under maximum attention.

Supplier onboarding

Once a supplier is selected, onboarding turns a decision into a working relationship: setting up the supplier in whatever systems track purchase orders and payments, confirming banking and invoicing details through a verified channel rather than trusting whatever arrives by email, agreeing communication points and escalation contacts on both sides, and walking through the specification, quality expectations and delivery process together before the first real order rather than assuming everything in the tender documents was fully absorbed.

Onboarding is also the moment to confirm, in writing, anything that was agreed verbally during selection and negotiation — because memories of a verbal understanding diverge remarkably quickly once both sides move on to other things, and the gap tends to surface at the worst possible moment, which is usually when something has gone wrong and both sides remember the earlier conversation differently.

Preparation as most of the work

The negotiation itself — the meeting or exchange where numbers are actually discussed — is a small fraction of the effort that determines the outcome. Most of the work happens beforehand: understanding what the item actually costs the supplier to produce, so an unreasonable price can be recognised as such; understanding what alternatives genuinely exist, so a walk-away point is real rather than bluffed; deciding in advance which terms matter most and which can be traded away; and understanding what the supplier needs from the deal, because a negotiation that only accounts for the buyer's interests tends to produce an agreement the supplier resents and under-delivers against over time.

A buyer who walks into a negotiation having done this preparation is negotiating from a position that does not depend on performing confidence; a buyer who has not done it is relying on the supplier not noticing, which is a poor long-term strategy against any supplier worth doing repeat business with.

Interests versus positions

A position is what someone says they want — a specific price, a specific delivery date. An interest is why they want it — the underlying need the position is meant to satisfy. Two sides arguing over positions often reach a stalemate that dissolves the moment either side asks what the other actually needs and discovers the positions were not as opposed as they looked.

A supplier insisting on a higher price, for instance, might actually need cash flow certainty rather than more total revenue — in which case a faster payment schedule at a slightly lower price satisfies their real interest better than holding firm on price would have, and costs the buyer less than meeting the stated position. Identifying the interest behind a position takes active, sometimes uncomfortable, questioning rather than assuming the stated position is the whole story, but it routinely opens up trades that a position-only negotiation never finds.

BATNA and how it is actually improved

BATNA — the best alternative to a negotiated agreement — is what a buyer does if this particular negotiation fails. It is the single most important piece of leverage in any negotiation, and it is not improved by asserting confidence at the table; it is improved beforehand, by actually doing the work of developing a real alternative. Qualifying a second supplier before a negotiation with the first, even one not currently in use, changes the buyer's real alternative from nothing to something, and that change is what shifts the negotiation, not the tone of voice used to imply it.

A buyer who claims to have alternatives they have not actually developed is bluffing, and experienced suppliers on the other side of the table are generally able to tell the difference between a bluff and a genuine alternative, because the questions a buyer with a real alternative asks are different from the questions a buyer without one asks. Improving BATNA is unglamorous, precedes the negotiation by weeks or months, and is worth more than any tactic used in the room.

Price versus terms — where the real money sits

Unit price is the term negotiators fixate on because it is the most visible and the easiest to compare, but for many purchases it is not where the largest value sits. Payment terms — how many days after invoice the buyer actually pays — affect cash flow directly and, for a buyer managing working capital tightly, can matter more than a percentage point or two on price. Volume commitments, where the buyer agrees to a minimum quantity over a period in exchange for better pricing or priority capacity, trade certainty for the supplier against better terms for the buyer, and are only sensible where the buyer is confident the volume will actually be needed. Tooling costs, for anything requiring supplier-specific moulds, dies or fixtures, are a one-off the buyer either pays upfront, amortises into unit price over an agreed volume, or negotiates the supplier to absorb in exchange for exclusivity — each allocation changes who bears the risk if volumes fall short. Minimum order quantities set the smallest batch a supplier will produce, and negotiating them down matters enormously to a buyer without the capital or the demand to justify large batches, even at a worse per-unit price. Lead times, and specifically the reliability of a stated lead time rather than its length alone, drive how much inventory the buyer needs to hold as a buffer, which is itself a real cost.

A negotiator who trades away all of these in exchange for a marginally better unit price has usually made a bad trade, because the total cost impact of terms frequently exceeds the unit price difference being fought over. Deciding in advance, per the preparation stage, which of these terms matter most for this particular purchase avoids that trap.

The negotiation itself, and its failure modes

An aggressive approach — treating the negotiation as a contest to be won, extracting every possible concession, signalling that the relationship matters less than the deal — can produce a favourable one-off outcome and a supplier who complies with the letter of the agreement while looking for opportunities to recover what they feel they lost, whether through subtly reduced quality, deprioritised attention when capacity is tight, or simple reluctance to help when something goes wrong outside the contract's strict terms. The short-term win is often a long-term cost that never appears on the invoice.

A passive approach — avoiding conflict, accepting the supplier's first position to keep the relationship comfortable, declining to push on terms that genuinely matter — leaves value on the table that a competent supplier will not volunteer to hand back, and over successive negotiations compounds into materially worse terms than a more assertive but still respectful buyer would achieve. Neither extreme serves the buyer well; the useful middle ground is firm on interests, flexible on how those interests are met, and honest enough that the supplier can trust what is said across future negotiations, because a reputation for straight dealing is itself a negotiating asset that a one-time maximiser never accumulates.

What a supply contract must contain

A supply contract that only covers price and quantity leaves almost everything that actually goes wrong in a supply relationship unaddressed. A properly built contract covers the specification, incorporated by reference to a controlled document so that changes to the spec are tracked and agreed rather than drifting informally; the quality standard the goods or services must meet and how conformance is verified; delivery terms, stating clearly who bears risk and cost at each stage of transport; the price mechanism, including whether and how price can change over the life of the agreement in response to input cost movements, commonly called indexation; liability, setting out who is responsible for what kind of loss and up to what limit; warranties, covering what the supplier guarantees about the goods and for how long; intellectual property, covering ownership of any designs, tooling or improvements developed during the relationship; confidentiality, protecting information shared in both directions; force majeure, defining what events excuse non-performance and what happens when they occur; termination, setting out how either party can end the agreement and what obligations survive that ending; dispute resolution, specifying how disagreements are resolved before they become expensive litigation; and governing law, stating which country's or state's law applies and where disputes are heard.

Every one of these clauses exists because someone, somewhere, was badly hurt by its absence. A contract missing a liability cap can expose a small buyer to catastrophic damages from a large customer's claim passed through the supply chain; a contract silent on IP ownership can leave a buyer unable to use a design they paid to develop; a contract with no clear force majeure clause can leave both sides arguing, expensively, about who bears the cost of an event neither controlled.

Purchase orders and framework agreements

A purchase order is the instrument for a specific transaction — this quantity, this price, this delivery date, issued under the terms of a broader agreement or, for a one-off purchase, on its own terms and conditions. A framework agreement sits above individual purchase orders and sets the terms that will govern all of them over a period — pricing mechanisms, quality standards, liability, and the other contractual content above — so that each individual order does not need to renegotiate the whole relationship from scratch.

For a buyer placing repeated orders with the same supplier, a framework agreement is worth the upfront effort of negotiating it properly, because it turns every subsequent order into a simple, fast transaction rather than a fresh negotiation, and it closes the gap where terms might otherwise be inconsistent or absent from order to order.

When a lawyer is needed

This page can describe what a supply contract needs to contain; it cannot draft one, and it should not be relied on as a substitute for a qualified commercial lawyer. A lawyer is needed before signing any contract involving meaningful value, any contract with a supplier in a different legal jurisdiction, any contract involving intellectual property the business depends on, and any contract whose liability, indemnity or termination terms are not fully understood by whoever is about to sign it. The cost of proper legal review is small next to the cost of discovering, after a dispute has already begun, that a clause everyone assumed protected them actually did not.

Quality management: inspection versus process control

Inspection catches defects after they have already been made, by checking finished goods against the specification before they are accepted. Process control prevents defects from being made in the first place, by controlling the conditions of production — materials, equipment calibration, worker training, in-process checks — so that the output is consistent by design rather than sorted after the fact.

Inspection alone is expensive and incomplete: it costs money on every batch, it only ever samples rather than checking everything unless the item is critical enough to justify full inspection, and it does nothing to reduce the underlying defect rate, meaning the same proportion of bad units keeps being produced and keeps needing to be caught. Process control is more expensive to establish and requires more supplier maturity to sustain, but it reduces the defect rate at the source, which compounds in the buyer's favour over the life of the relationship. A mature approach uses inspection as a check on process control rather than as a substitute for it, and pushes suppliers, particularly significant ones, toward stronger process control over time rather than accepting permanent reliance on catching problems at the door.

Acceptance criteria and sampling in principle

Acceptance criteria state, unambiguously, what makes a delivered batch acceptable — which characteristics are checked, what tolerance is allowed, and what happens if the batch fails. Sampling is the practice of checking a subset of a batch rather than every unit, on the reasoning that inspecting everything is uneconomic for most goods and a well-designed sample gives a statistically defensible read on the whole batch's quality.

The specific statistical sampling plans used in formal quality systems are a technical discipline in their own right and are not reproduced here; what matters for a buyer is understanding, in principle, that a sampling plan is a deliberate trade-off between inspection cost and the risk of accepting a bad batch or rejecting a good one, and that this trade-off should be set deliberately for each category rather than defaulted to whatever the supplier happens to propose, since the supplier's incentive in setting the sample size is not automatically aligned with the buyer's.

Non-conformance, corrective action and supplier development

When a delivery fails to meet the agreed specification — a non-conformance — the useful response is not simply rejecting the batch and moving on, though that may be necessary for the immediate order. The useful response also identifies why the failure happened and what will stop it happening again, formalised as a corrective action: what the supplier will change in their process, materials or controls to address the root cause rather than just the symptom that was caught this time.

Supplier development takes this further, treating a struggling but strategically important supplier as worth actively helping to improve — sharing expertise, funding process improvements, or simply providing clearer feedback and more structured engagement — rather than simply switching away at the first sign of trouble. This is worth doing selectively, for suppliers whose relationship is worth preserving and whose underlying capability is sound even where current performance is not; it is not worth the investment for a low-value, easily replaced supplier where switching is simply cheaper than developing.

Supply risk categories

Supply risk arrives in several distinct shapes, and confusing one for another leads to mitigations aimed at the wrong problem. Single-source risk is the exposure created by having no alternative if one supplier fails, discussed above under sourcing strategy. Geographic concentration risk arises when many suppliers, even if nominally different companies, sit in the same region and are therefore exposed to the same local disruption — a natural disaster, a regional conflict, a change in local regulation — regardless of how diversified the supplier list looks on paper. Financial distress risk is the risk that a supplier's own business fails, for reasons that may have nothing to do with the buyer's relationship with them. Geopolitical risk covers trade restrictions, sanctions, tariff changes and diplomatic breakdowns that can disrupt a supply route with little warning. Climate risk covers physical disruption from extreme weather and the longer-term effect of a changing climate on the viability of production in a given region. Cyber risk covers the exposure created when a supplier's systems are compromised, which can disrupt their ability to fulfil orders or expose the buyer's own data if systems are connected. Logistics disruption risk covers port congestion, carrier capacity shortages, and the many points along a physical supply route where a delay or blockage can stop goods moving regardless of how well the supplier itself is performing.

Each of these calls for a different mitigation, and mitigations aimed at one do little against the others — qualifying a second supplier addresses single-source risk but does nothing for geographic concentration if both suppliers sit in the same region; financial monitoring addresses distress risk but does nothing for a geopolitical trade restriction. A serious risk assessment treats these as separate categories and asks, for each significant category of spend, which of these risks actually applies and how exposed the current supply base is to it.

Business continuity and dual qualification

Business continuity planning, applied to sourcing, means having a genuine answer to the question of what happens if a critical supplier disappears without warning — not a hoped-for answer, but one that has actually been tested. Dual qualification, having already qualified a second source for a critical item even if that second source is not currently used for live production, is the most reliable version of this answer, because qualification itself takes time and cannot be done quickly once a crisis has already started.

The discipline that makes dual qualification actually useful, rather than theoretical, is periodically placing a small real order with the backup supplier, so that their process, quality and lead time are current rather than assumed from a qualification exercise carried out years earlier under different conditions.

Ethical and sustainable sourcing

Labour standards in the supply chain — working conditions, wages, working hours, freedom of association — are a direct responsibility of the buyer in the eyes of customers, regulators and increasingly the law, regardless of whether the buyer directly employs the workers concerned. Many jurisdictions now impose modern slavery reporting or due diligence obligations on buyers of a certain size, requiring them to understand and disclose the risk of forced labour in their supply chains; the specific obligations vary by jurisdiction and are a legal compliance question best answered with qualified legal advice rather than general guidance.

Environmental claims made about sourced goods — recycled content, sustainable materials, carbon impact — carry their own regulatory and reputational risk if they cannot be substantiated, and a buyer repeating a supplier's environmental claim without verifying it takes on the risk of that claim being false. Chain of custody is the documented trail showing where a material actually came from and how it moved through the supply chain to reach the buyer; it is the mechanism that makes any sourcing or sustainability claim verifiable rather than merely asserted. Certification, where an independent body attests that a supplier or product meets a defined standard, is a useful signal but is not the same as assurance, which is the buyer's own ongoing verification that the standard is actually being met in practice; certifications can lapse, be granted on the basis of a single audit that does not reflect ongoing operations, or in rare cases be obtained fraudulently, and a buyer relying purely on a certificate without any of their own verification has not actually eliminated the underlying risk, only the appearance of it.

Conflict minerals and restricted substances, in outline

Certain minerals, sourced from specific conflict-affected regions, are subject to due diligence and disclosure requirements in a number of jurisdictions, aimed at preventing the trade from funding armed conflict. Certain substances are restricted or banned in specific products or markets on health, safety or environmental grounds, with the specific list and threshold varying by product category and destination market. Both of these are specialist regulatory compliance areas with real legal consequences for getting wrong, and a buyer whose products or components could plausibly be affected should get a proper compliance assessment from someone qualified in the relevant regulations rather than relying on general awareness that the issue exists.

Sanctions and denied-party screening

Sanctions regimes restrict or prohibit trade with specific countries, entities and individuals, and denied-party screening is the practice of checking counterparties — suppliers, their owners, and sometimes their own suppliers — against the relevant lists before doing business with them. The consequences of a sanctions breach can be severe for the buyer, regardless of whether the breach was intentional, and the applicable lists and rules vary by the buyer's own jurisdiction and the jurisdictions the transaction touches. This is another area where the correct action is engaging a qualified compliance professional or customs broker rather than relying on a general description of the concept; what matters here is knowing that the obligation exists and applies before, not after, a supplier relationship is established.

Demand planning and forecasting

Demand planning is the exercise of estimating what will be needed, when, so that sourcing and production can be arranged in time. Every forecast is wrong to some degree; the useful goal is not eliminating forecast error, which is not achievable, but understanding its likely size and building the rest of the supply chain to absorb it gracefully rather than being surprised by it every time.

Forecast error tends to be larger the further out the forecast reaches, larger for new or volatile items than for stable, established ones, and larger when demand depends on external factors the business does not control, such as competitor actions or broader market conditions. A demand plan that states a single number without any sense of the likely range around it invites the rest of the organisation to treat that number as more certain than it is, which is where much of the downstream cost of forecast error actually comes from — not the error itself, but the false confidence placed in the original number.

The bullwhip effect, explained mechanically

The bullwhip effect is the tendency for small fluctuations in actual customer demand to become progressively larger swings as they are passed back up the supply chain from retailer to distributor to manufacturer to raw material supplier. It is a mechanical consequence of how ordering decisions are made at each stage, not a mysterious market phenomenon, and it is worth understanding step by step because understanding the mechanism is what allows it to be dampened.

Consider, illustratively, a retailer who sees a small, genuine increase in customer demand for an item. Wanting a buffer against running out, and perhaps expecting the increase to continue, the retailer orders somewhat more from their distributor than the increase alone would justify. The distributor, seeing this larger order and applying the same buffering logic, orders more still from the manufacturer, on top of also trying to build a bit of extra safety stock of their own. The manufacturer, seeing an even larger jump in orders from the distributor, and unsure how much of it is real underlying demand versus buffering by others, ramps production up more than the original demand increase warranted and orders more raw material to match. Each stage's rational, self-protective response to the order it sees, rather than to the actual end customer demand it cannot see directly, amplifies the swing at every step. When the original demand increase levels off or reverses, the same mechanism runs in reverse, and the amplified excess inventory sits unwanted throughout the chain.

The mitigations that work follow directly from the mechanism: sharing actual end-customer demand data further up the chain rather than only sharing order quantities, so that each stage can react to real demand rather than to the amplified signal of the stage below it; shortening lead times, which reduces how much buffering each stage feels it needs; stabilising pricing and promotions, since sudden price changes and promotional spikes are a major source of the artificial demand swings that trigger the effect in the first place; and smaller, more frequent ordering rather than large infrequent batches, which reduces the amplitude of each individual order signal passed upstream.

Inventory policy: cycle stock, safety stock and reorder points

Cycle stock is the inventory held to cover expected demand between one order and the next, sized around the ordering pattern and the demand rate. Safety stock is additional inventory held beyond that, specifically to absorb the unpredictable parts of the equation — demand coming in higher than forecast, or a delivery arriving later than expected — without running out before the next delivery arrives.

The reorder point is the stock level at which a new order is triggered, calculated to account for how much will be consumed during the lead time it takes the new order to arrive, plus the safety stock buffer. Getting these three right is a balance: too little cycle and safety stock risks stockouts, lost sales and, for anything feeding production, halted lines; too much ties up capital, incurs storage cost, and for perishable or fashion-sensitive goods, risks the stock becoming obsolete or unsellable before it can be used. Neither extreme is safe, and the right balance depends on the specific cost of a stockout versus the specific cost of holding excess stock for the item in question, which differs by item and is worth calculating deliberately rather than applying a single policy across everything the business holds.

Economic order quantity as a teaching model

The economic order quantity model is a classic, simplified way of thinking about how much to order at once, balancing the cost of placing an order — which is roughly fixed regardless of how much is ordered — against the cost of holding inventory, which rises with the quantity held. It is a useful teaching model precisely because its assumptions are simple enough to reason through, and understanding those assumptions is more valuable than the formula itself, because the assumptions are also where the model breaks down in the real world.

Illustratively: suppose a business uses 1,200 units of an item per year, it costs 25 in administrative and handling cost each time an order is placed regardless of size, and it costs 2 per unit per year to hold one unit in stock, covering capital, storage and handling. The economic order quantity model finds the order size that minimises the combined cost of ordering and holding, and for these stated figures, working through the standard calculation, the answer comes out at roughly 173 units per order, ordered a little under seven times a year. The point of walking through the arithmetic is not the specific number, which depends entirely on the input assumptions used, but the shape of the trade-off: ordering more at once reduces how often the fixed ordering cost is paid but increases the average inventory sitting in the warehouse; ordering less does the reverse.

The model's assumptions, stated explicitly, are that demand is constant and known, lead time is fixed and known, price does not change with order size, and no stockouts are allowed. Real purchasing rarely matches all four assumptions at once — demand fluctuates, lead times vary, suppliers offer volume discounts that change the true cost of larger orders, and a business may tolerate occasional stockouts rather than pay to eliminate them entirely. The model is worth knowing as a way of thinking about the trade-off, not as a formula to apply mechanically to a real, messier situation without adjustment.

Lead time and variability

Lead time is how long it takes from placing an order to having usable stock in hand, and it matters as much for its variability as for its average length. A supplier with a longer but highly consistent lead time is often easier to plan around than one with a shorter but wildly unpredictable one, because the buffer needed to protect against variability grows with the uncertainty, not just with the average wait. Reducing variability, through closer supplier relationships, better visibility into the supplier's own production schedule, or simply choosing more reliable logistics routes, is frequently a more valuable improvement than reducing the average lead time itself, because it lets the buyer hold less safety stock for the same level of protection against running out.

MRP and production planning, in outline

Material requirements planning is the logic, usually run through dedicated software, that translates a production schedule into a detailed plan of what raw materials and components need to be ordered, in what quantities, and by when, working backward from finished-goods due dates through the bill of materials and each component's own lead time. For a business making anything from multiple components sourced from multiple suppliers, some version of this logic, even a simplified spreadsheet version well short of a full MRP system, is what prevents a shortage of one cheap, easily overlooked component from stopping an entire production run of an otherwise fully stocked product.

Full production planning, integrating capacity constraints, workforce scheduling and multiple competing products, is a specialist operations discipline in its own right; what matters for sourcing is understanding that the ordering decisions covered in this page feed directly into that planning process, and that a change in lead time or minimum order quantity negotiated with a supplier has a direct, calculable effect on how far in advance production needs to be planned.

Just-in-time and its resilience trade-off

Just-in-time sourcing aims to have materials arrive as close as possible to the moment they are actually needed, minimising the inventory held at any point in the chain. Done well, it reduces the capital tied up in stock, reduces warehousing cost, and forces a discipline of quality and reliability throughout the chain, because there is no buffer stock available to hide a problem behind.

The trade-off is resilience. A chain run on minimal buffers has very little slack to absorb a disruption — a late delivery, a quality failure, a transport delay — before it stops production entirely, because the buffer that would normally absorb such an event has been deliberately removed. Whether that trade-off is worth making depends on how reliable the supply chain genuinely is and how costly a disruption would be if it occurred; a business with highly reliable, well-understood suppliers and low disruption risk can run lean with real benefit, while a business exposed to significant supply risk, of the kinds discussed above, may find that the working capital saved by minimal inventory is a poor trade against the cost of a production stoppage. Neither position is correct in general; both are correct for the specific conditions that justify them, and this is one of the genuine, ongoing disagreements between operations practitioners who have each seen the trade-off play out differently in their own circumstances.

Warehousing and inbound logistics

Inbound logistics covers moving purchased goods from the supplier's location into the buyer's own storage or production, and warehousing covers what happens to them once they arrive and before they are needed. Decisions here include whether to hold buffer stock close to the point of use or centrally, how goods are received and checked against the purchase order and specification on arrival, and how storage conditions are managed for anything sensitive to temperature, humidity or shelf life.

Inbound logistics quality is easy to overlook because it sits between two more visible activities — the supplier's production and the buyer's own use of the goods — but damage, loss or delay in this middle stage is a common and avoidable source of cost, and a receiving process that actually checks what arrives against what was ordered, rather than simply accepting deliveries into stock, is one of the cheaper controls available in the whole chain.

Freight modes and their trade-offs

Sea freight is generally the least expensive way to move large volumes over long distances, and generally the slowest, with the longest and often most variable lead times of the common modes. Air freight is fast and reliable in transit time, at a cost premium that is usually justified only for high-value, time-sensitive or perishable goods where the cost of the delay avoided outweighs the freight premium paid. Road freight suits shorter distances and offers flexibility in routing and scheduling that larger vessels or aircraft cannot match, at a cost and speed that sit between the other two for medium distances. Rail freight, where the infrastructure exists, can offer a useful middle ground of cost and speed for certain overland routes, along with generally lower emissions per unit moved than road or air.

Choosing a mode is a trade-off among cost, speed and reliability specific to the item being moved and the urgency of the need, and larger buyers frequently use a mix — sea freight for planned, predictable volume, with air freight reserved as an expensive but fast fallback when something has gone wrong with the planned supply and a stockout is the more expensive alternative. Specific freight rates change constantly with fuel costs, capacity and season, and any number quoted today is unlikely to hold; rate levels are a live commercial question to check with a freight forwarder or logistics provider at the time of shipping, not something to treat as fixed general knowledge.

Customs and documentation on the inbound side

Goods crossing an international border require accurate documentation — commercial invoices, packing lists, certificates of origin, and whatever else the specific goods and destination require — and are subject to duties, taxes and regulatory checks that vary by product classification and country. Getting the classification and documentation wrong can mean delays, unexpected costs, or in serious cases, seizure of goods, and the specific rules are detailed, frequently updated, and vary enough by jurisdiction and product that this page cannot responsibly summarise them as general guidance. A customs broker or freight forwarder with expertise in the specific trade lane and product category is the right source for this, and for any material volume of cross-border sourcing, engaging one properly is a sound cost rather than an avoidable overhead. The wider question of selling and shipping across borders to customers, as opposed to bringing goods in, is covered on the e-commerce and cross-border trade page.

Supplier performance measurement

On-time-in-full measures the proportion of orders delivered both on the agreed date and at the full agreed quantity, and is one of the more informative single metrics because it fails a delivery that is technically on time but short of quantity, or full but late, both of which cause real operational problems that a metric measuring only one dimension would miss. Quality rate measures the proportion of delivered goods that meet the specification without rework or rejection. Responsiveness measures how quickly and effectively a supplier answers questions, resolves problems and adapts to changed requirements, and while it is more subjective than the other two, it is often the metric that best predicts how a relationship will hold up under stress, when the formal numbers have not yet caught up with an emerging problem.

Scorecards built from these and similar metrics are useful when they drive an actual conversation with the supplier about what is behind a bad reading and what will change, and become theatre when they are compiled, filed and never discussed — or worse, when suppliers learn to manage the metric rather than the underlying performance, for instance by splitting a late, full delivery into an on-time partial delivery followed by a late completion, which can improve the on-time-in-full number on paper while making the buyer's actual planning problem worse, not better.

The supplier relationship spectrum

Supplier relationships range from purely transactional, where either party could be replaced with minimal disruption and the interaction is essentially a series of arm's-length purchases, through to strategic partnership, where both sides invest in the relationship beyond what any single transaction requires because the ongoing relationship itself creates value neither could achieve alone — joint product development, shared investment in capacity, deep information sharing that would be commercially reckless with a purely transactional supplier.

It is worth being honest that genuine strategic partnerships are rare, and that most relationships sit somewhere well short of that end of the spectrum for entirely sensible reasons: partnership-level investment of time and trust is expensive for both sides, and is only worth making where the item or service is genuinely critical and the relationship is genuinely differentiated rather than easily replaced. Treating every supplier as a strategic partner dilutes the attention available for the few relationships that actually warrant it, and treating a genuinely strategic supplier as purely transactional forgoes value that a closer relationship would have created. The useful discipline is placing suppliers honestly on this spectrum, based on actual criticality and switching cost rather than on how the relationship feels, and investing relationship effort accordingly.

Segmentation of the supply base

Segmentation groups suppliers, usually along two axes — the value or risk of the spend, and the complexity or scarcity of the supply market — into categories that call for different management approaches. High-value, high-complexity suppliers, sometimes called strategic, warrant the closest attention and the relationship investment discussed above. High-value, low-complexity suppliers, where the spend is large but alternatives are readily available, are the category where competitive pressure and price negotiation earn the most return. Low-value, high-complexity suppliers can be disproportionately risky relative to their spend, because scarcity means a disruption is hard to replace even though the total spend involved is modest, and this category is easy to under-manage precisely because the spend total looks unimportant. Low-value, low-complexity suppliers are the routine, easily replaced purchases that warrant the least individual attention and are best managed through simple, efficient processes rather than case-by-case scrutiny.

The value of this kind of segmentation is stopping a sourcing function from spreading equal attention across every supplier regardless of how much that attention is actually worth, which is the default failure mode of an unstructured supply base as it grows.

Joint improvement and cost-down programmes

For strategic and significant suppliers, some of the most durable value comes not from renegotiating price periodically but from working with the supplier to genuinely reduce the cost or improve the performance of what is being produced — jointly identifying waste in the process, redesigning a component for easier manufacture, or improving the flow of information between the two organisations so that both spend less time and money on coordination.

These programmes work when both sides see a genuine share of the benefit, whether through a formal gain-sharing arrangement or simply through a track record of the relationship being renewed and grown as a result. They fail, and become a source of resentment, when a buyer treats a cost-down initiative purely as a mechanism for extracting savings without the supplier seeing any benefit from the improvements they helped create, since a supplier who correctly perceives this pattern will, reasonably, stop volunteering improvement ideas.

Open-book and should-cost approaches

An open-book relationship is one where the supplier shares detailed cost information — materials, labour, overhead, margin — with the buyer, allowing negotiation to be grounded in the actual cost structure rather than in opaque quoted prices. It requires a level of trust that only a genuinely close, long-term relationship can sustain, since the supplier is disclosing information that would weaken their position with a less trustworthy buyer.

A should-cost model is the buyer's own independent estimate of what an item ought to cost to produce, built from an understanding of materials, process and reasonable margin, used as a benchmark against a supplier's quoted price whether or not that supplier is willing to open their books. It is more work to build than simply comparing quotes, and it is the tool that lets a buyer recognise an unreasonable price even from a supplier unwilling to explain their own cost structure, which is most of them.

The systems: ERP, procurement suites, portals and repositories

An enterprise resource planning system, where one is in use, typically holds the core transactional record — purchase orders, inventory levels, invoicing — and integrates sourcing activity with the rest of the business's finance and operations data. Dedicated procurement suites add functionality specific to sourcing — running RFx processes, managing supplier scorecards, tracking contract terms — that a general ERP system often handles only thinly. Supplier portals give suppliers direct, self-service visibility into orders, forecasts and performance data, reducing the manual back-and-forth that otherwise consumes buyer time. Contract repositories hold the actual signed agreements in a searchable, accessible place, which sounds trivial and is one of the more common practical failures in smaller organisations, where a critical contract term is known to exist somewhere but cannot actually be found when a dispute arises and it is needed quickly.

A small operation does not need all of this at once, and buying an enterprise procurement suite before the business has the volume or complexity to justify it is a common and expensive mistake. What a small operation genuinely needs, from the start, is a single, reliable place where every supplier's agreed terms, contact details and order history actually live and can be found quickly, and a simple, consistently followed process for issuing and tracking purchase orders. A well-organised spreadsheet, used consistently, outperforms a sophisticated system used haphazardly, and the right moment to invest in more capable systems is when the manual approach starts visibly failing under its own volume, not before.

Data quality as the precondition

Every technique described in this page — spend analysis, category management, supplier scorecards, should-cost modelling, forecast-driven inventory policy — depends on the underlying data being accurate, complete and consistently recorded. A spend analysis built on inconsistently categorised invoices produces a misleading picture no matter how sophisticated the analysis technique applied to it; a supplier scorecard built on incomplete delivery records understates or overstates performance in whichever direction the gaps happen to fall.

Data quality is unglamorous work — consistent supplier naming, consistent categorisation, complete and timely recording of deliveries and defects — and it is the actual precondition for everything more advanced that a sourcing function might eventually want to do, including the use of any analytical or AI tool discussed next, none of which can produce a reliable output from unreliable input.

Where AI genuinely helps in sourcing, and where it does not

Described soberly rather than aspirationally: tools that use AI techniques currently show genuine, practical use in sourcing in a handful of areas. Spend classification — automatically categorising large volumes of invoice line items into consistent spend categories — is faster and often more consistent than manual categorisation, and this is exactly the kind of large, repetitive pattern-matching task the underlying techniques handle well. Contract review assistance — scanning a contract to flag missing clauses, unusual terms, or deviations from a standard template — can speed up an initial review, though it does not replace a qualified lawyer's judgement on anything that actually matters in a contract, and treating an automated flag as the final word rather than a first pass is a mistake. Demand forecasting support, where historical pattern-finding can improve on simple manual extrapolation for stable, well-established product lines, is a genuine improvement over guesswork, though it is materially weaker for new products or volatile demand with limited history to learn from, which is precisely where good forecasting matters most.

Where these tools do not yet reliably help is judgement under genuine uncertainty — assessing whether a new supplier in an unfamiliar market can actually be trusted, reading the subtext of a supplier's tone in a negotiation, or weighing a values-based trade-off between cost and ethical sourcing standards that has no clean historical pattern to learn from. These remain, and are likely to remain, human judgement calls informed by tools rather than made by them, and a buyer who defers judgement to a tool's output in these areas is trusting a pattern-matcher with a decision it was never actually built to make.

Governance, delegated authority and approval thresholds

As a sourcing function grows beyond one person making every decision personally, governance becomes necessary: who is authorised to commit the business to what level of spend, what level of spend requires a second signature or a more senior approval, and what categories of decision — a new single-source supplier for a critical item, a contract with an unusual liability clause — require escalation regardless of the value involved.

Delegated authority limits exist precisely so that a business does not depend entirely on one person's judgement for every purchase, and so that the size of a commitment a single individual can make without oversight is bounded to something the business can absorb if that judgement turns out to be wrong. Setting these thresholds too low creates bureaucratic friction that slows down routine, low-risk purchases for no real benefit; setting them too high removes the check that governance exists to provide. The right thresholds are specific to the size and risk appetite of the particular business and are worth revisiting as the business grows, since a threshold that made sense when the business was smaller can quietly become inadequate as spend volumes rise.

The annual review

A periodic, deliberate review of the sourcing function as a whole — typically annual, though the right cadence depends on how fast the business and its supply base are changing — is where the individual disciplines covered throughout this page are checked against each other rather than left to run independently. Are the category strategies set out in part one still the right ones given how spend and risk have shifted over the year? Has the risk assessment from part four been updated for suppliers added or changed since the last review? Are the performance metrics from part five actually being discussed with suppliers, or filed and ignored? Has segmentation been revisited as some suppliers have grown in importance and others have faded?

Without a scheduled review, all of this drifts quietly out of date, not through any single failure but through the ordinary accumulation of small, unexamined changes over a year of normal operation. The review does not need to be elaborate to be useful; it needs to actually happen, on a fixed schedule, rather than being deferred indefinitely by more urgent day-to-day demands.

Standing up a sourcing function from nothing

A business with no formal sourcing function, run by one buyer with no systems beyond email and a spreadsheet, does not need to build everything in this page at once, and trying to would be a mistake — most of the value is in a small number of basic disciplines, applied consistently, long before any of the more advanced material becomes relevant.

The checklist worth working through, roughly in order: know what is actually being spent, and with whom, by pulling every invoice into one place rather than trusting memory or scattered records. Write down, for anything significant, what is actually being bought, in enough detail that a stranger reading the specification would order the right thing. Never place a first order with a new supplier of any significant value without checking that they legally exist, have some financial standing, and can be reached by more than one contact person. Put the agreed terms in writing, even briefly, every single time, rather than relying on what was said in a phone call or an email thread that will not be findable in a year. Know the lead time and the minimum order quantity for every significant item, and hold enough safety stock to survive the ordinary variation in both. Identify, honestly, which single suppliers represent single points of failure the business could not currently survive losing, and start, even slowly, qualifying an alternative for the worst of them. And revisit all of it at least once a year, on a date fixed in advance rather than left to whenever there is time, because the single most common failure in sourcing is not any individual bad decision described in this page — it is a sound set of decisions made once and never looked at again while the business, its suppliers and the world around both of them continued to change.