Direct procurement buys what goes into the product or service an organisation sells: raw materials, components, ingredients, packaging, goods for resale. Indirect procurement buys everything the organisation needs in order to operate: IT, facilities, MRO, travel, professional services, marketing, office supplies.
Stated that way it sounds like a filing convention, and it gets taught as one. It is not. Almost every operational property of a purchase differs across the line: how the demand arises, how many suppliers are involved, who initiates it, what happens when it goes wrong, and which system carries it. Organisations that treat the two alike end up either strangling small purchases in heavyweight process or running production inputs with no visibility into whether they will arrive.
Which side something falls on
The reliable test is accounting rather than intuition. Direct spend lands in cost of goods sold, because it is consumed in producing what the organisation sells. Indirect spend lands in operating expense. Somebody in finance has already made that call for every category, and it is usually a better answer than a fresh debate about whether packaging counts.
Across sectors it shakes out like this. For a manufacturer, direct is steel, castings, fasteners, electronic components, the corrugated the product ships in. For a food producer, it is ingredients and packaging. For a retailer it is goods for resale, which is direct even though nothing is manufactured. For a software company, the cloud infrastructure the product runs on has a reasonable claim to being direct, and many finance teams now treat it that way.
Indirect is more uniform across sectors, which is why horizontal group purchasing organisations can exist at all. Laptops, cleaning contracts, spare parts for the building rather than for the product, airline tickets, legal advice, agency time, print, furniture. A hotel group and a foundry buy noticeably similar indirect categories and radically different direct ones.
How demand arises
This is the difference that generates all the others.
Direct demand is calculated. A demand forecast produces a production plan, the production plan is exploded against a bill of materials, and the result is a dated requirement for a specific quantity of a specific part. Nobody decides to buy the part; the arithmetic does, and the buyer’s job is to have the contract, the price, the lead time and the capacity in place before the requirement appears. A direct buyer knows in March roughly what they will need in June.
Indirect demand is requested. Somebody in the business needs something, raises a requisition, and the requisition is the first the procurement function hears about it. There is no forecast, no bill of materials, and frequently no repeat. The demand is generated by a person rather than by a plan, which means it arrives unbatched, unpredictably, and in whatever description the requester happened to type.
Supplier base and relationship
Direct procurement runs a small, deliberately managed supplier base. Qualification is heavy: audits, sample approval, quality agreements, capacity reviews, sometimes tooling the buyer owns and the supplier houses. Contracts are long and switching is slow, which is why dual sourcing is a deliberate risk decision rather than a default. Price is negotiated annually against volume, often with index clauses tied to a raw material.
Indirect runs the opposite shape: an order of magnitude more suppliers, most of them shallow relationships, many used once. Requesters sit in every department, and much of the supplier base was created by somebody who needed one thing quickly. That long tail is where maverick spend lives, and nobody owns it the way a commodity manager owns a casting.
What failure looks like
Direct failure is loud. A missed component delivery stops a line, idles a shift, breaks a customer commitment, and gets escalated within hours. The cost is immediate, quantifiable and visible well outside procurement, which is why direct is measured on continuity of supply at least as much as on price.
Indirect failure is quiet. A single overpriced purchase harms nobody perceptibly. A requester who buys off contract because it was faster gets their item and their approval, and the damage shows up only in aggregate, months later, in a spend report nobody reads line by line. Indirect cannot be managed transaction by transaction. It is managed by making the compliant path the easy one.
That asymmetry explains nearly every governance decision on either side of the line.
Why the systems diverge
Given all of the above, the systems that carry each kind of spend were never going to converge.
Direct procurement is driven by planning. MRP inside the ERP generates net requirements, releases them against blanket orders or scheduling agreements, and communicates them to suppliers as forecasts and delivery schedules, frequently over EDI. Consignment stock, kanban pulls and vendor-managed inventory all sit in this world. There is no browsing in the process. Nobody shops for the part in the bill of materials, because the part is not a choice.
Indirect procurement is driven by catalogs and requisitions, and for a structural reason: the requester has to find the thing before they can ask for it. That requires something searchable, which means a procurement catalog, whether hosted inside the buying system or reached live on the supplier’s site through PunchOut. Contract prices have to be attached to items, guided buying has to steer the requester to the right supplier, and approval rules have to run before the order is released.
It follows, and it is worth saying plainly, that catalog and PunchOut infrastructure is fundamentally an indirect-spend concern. It is not a lesser version of direct procurement tooling. It solves a different problem: presenting a contracted assortment to non-specialist requesters at the moment they need it. Whether to hold that assortment locally or reach for it live is its own decision, covered in hosted catalog vs PunchOut.
Services, the awkward third category
Services fit neither definition cleanly and are usually filed under indirect for lack of anywhere better. Consulting, legal, agency work, cleaning, maintenance, security, temporary labour, software subscriptions.
What makes them awkward is that the mechanics of goods buying do not apply. There is often no unit, no catalog line, no shipment and no clean receipt. The commercial instrument is a statement of work, a rate card or a subscription term rather than a part number and a price. Three-way matching struggles here, because the middle document is missing: what exactly was received, and who confirms it, when the deliverable is forty hours of somebody’s time. Organisations that take services seriously tend to run them through a separate process with milestone or timesheet approval standing in for goods receipt.
Services also fall between owners. The direct team does not want them, the indirect catalog programme cannot represent them, and the result is a large block of spend governed more loosely than the categories on either side of it.
Why indirect gets most of the process attention
In manufacturing, direct is usually the larger share of spend by value. Yet process improvement work concentrates on indirect, and this is rational rather than a misallocation.
The reason is transaction count. Direct spend is a small number of large, planned, already-systematised movements against existing agreements. Indirect is a very large number of small, unplanned ones, each dragging a requisition, an approval, a purchase order, a receipt and an invoice behind it. Process cost is paid per transaction, not per euro, so the administrative burden sits almost entirely on the indirect side. Add to that the fact that direct spend is already disciplined by the production plan, while indirect is where uncontracted and unmanaged buying actually lives, and the attention goes where the slack is.
Where the boundary blurs
The clean definition breaks down in several sectors, and hospitality is the clearest case.
Food and beverage in a hotel or restaurant group is direct by any reasonable reading: it is consumed in producing what the guest pays for, it belongs in cost of sales, and a shortage stops service the way a missing component stops a line. Recipes function as bills of materials. But it is bought through entirely indirect-style processes: catalogs of contracted items, requisitions raised per outlet by chefs rather than by planners, daily or twice-weekly ordering across many local suppliers, and substitution decided at the point of order. The spend behaves like direct and the transaction behaves like indirect.
Retail goods for resale sit similarly, as does clinical supply in healthcare and most project-based construction spend. The practical resolution is to stop asking which label applies and ask which properties apply. Is the demand planned or requested? Does a shortage stop service? Is the supplier base concentrated or fragmented? Those three answers determine the controls and the system far more usefully than the category name does.
How the split fails in practice
Direct-grade governance applied to indirect spend. Three approvers, a sourcing event and a signed contract for a purchase worth less than the labour spent approving it. This is the most common failure, and it does not produce control, it produces workarounds: expense claims, corporate cards and off-contract buying, all of which are less visible than the requisition would have been.
Indirect-style catalog buying applied to direct spend. A production input bought by requisition against a catalog line, with no forecast shared with the supplier, no capacity commitment and no visibility of supply continuity. The price may be fine. The exposure is that nobody finds out about a problem until the delivery does not arrive.
One approval threshold for everything. Thresholds set purely on value ignore risk. A low-value single-source chemical with a twelve-week lead time deserves more scrutiny than a high-value laptop order four suppliers could fill tomorrow.
Systems selected for one half and rolled out to both. A platform chosen by the direct team for planning strength, then handed to indirect requesters who find it unusable, or a catalog programme judged against total organisational spend when it was only ever addressing the indirect slice.
Savings claimed on incompatible bases. Direct savings measured against last year’s contracted price mean something specific. Indirect savings measured against a list price nobody was ever going to pay mean much less, and reporting them together flatters the indirect number.
Making the indirect side work is mostly a content problem: the right items, at contracted prices, in front of the right requester, in the system they already use. That is what we build. SupplierForge ingests supplier catalog content through CSV, API, SFTP or EDI and publishes it into buying systems with native PunchOut, cXML and OCI support, which is the layer that turns a negotiated contract into the option a requester sees. It applies wherever the buying behaves like indirect, including the hospitality food and beverage spend the textbook definition files on the other side.