Procurement: Improving Supplier Coordination Without Replacing Business Judgment

Procurement throughput improves when requests, quotes and supplier records arrive structured, because the negotiation is only as good as the completeness of what arrived before it started.

The buying decision is made twice, and the first one is unmanaged

A requester needs a component that is not in stock. They find a supplier they have used before, ask for a price by message, get one, and send it onward. A buyer then discovers the price excludes tooling, the lead time is unstated, the supplier is not approved for that material, and the specification the quote was written against differs from the one on the drawing.

Nobody did anything wrong. The point is that a commitment was made before anyone with the authority to make it was involved, and the work of procurement is now to reconstruct a decision that has already shaped itself. That is where the time goes, and it is invisible in any process diagram that starts at the approved request.

A request is four things, and three of them are usually missing

A purchase requisition that a system can act on carries more than an item and a quantity. It carries the specification, the standard the item must meet, the date it is genuinely needed and why, and the consequence of missing it. In most organisations three of those four are supplied verbally, later, by the person who remembers.

  • The specification, which exists in a drawing or a test method that the quote is never sent with.
  • The requirement date, which is often the date somebody needs it rather than the date the business can justify.
  • The consequence of delay, which determines priority and is almost never recorded.
  • The budget owner, which decides who can authorise the spend and is routed to a mailbox rather than a role.

A structured request does not prevent judgement. It relocates it to where it belongs: to a person who can decide whether a specification is acceptable, a date is real, and a supplier is capable. The people doing the assembling are not qualified to make those calls and are doing them badly, without any authority to do so.

Quotes are not comparable until the terms are comparable

Three quotes for the same item routinely differ by twenty percent, and the gap is frequently about terms rather than price: unit of measure, minimum order quantity, tooling amortised into the first run, delivery terms, payment terms, currency of the quotation, and validity period. A comparison of the headline numbers alone rewards whichever supplier quotes the narrowest scope, which is rarely the one who will perform.

This is a specification problem in the strict sense. The comparison cannot be done until the business has decided which fields are part of the offer, which of them it is prepared to trade, and how it will treat a quotation that omits one. Those decisions belong to the buyer and should be written once rather than re-derived per request.

FieldWhy it changes the comparisonDecision owner
Scope of supplyTwo quotations may price different thingsCategory buyer
Unit and price basisCompares per piece against per batchCategory buyer
Lead time and its basisOne is from order, one from readinessPlanning, with the buyer
Tooling and amortisationMoves cost across the first orderFinance and the buyer
Payment termsDetermines the true cost of the purchaseFinance policy

A quote arriving in prose is the normal condition, not an exception. Reading a supplier quotation or a specification sheet, pulling out the fields the business actually compares, and returning the result for a person to confirm is precisely where unstructured extraction belongs.

Supplier records drift because nobody owns the definition

The supplier file is the least contested asset in procurement and the least reliable. Contact details sit with one person, performance notes in another, approved-material status in a quality system, and payment terms in the ledger. A supplier can be simultaneously approved and flagged, and the conflict surfaces at the worst moment because there is no single record to check.

The fix is ownership rather than tooling. Somebody must be accountable for the supplier record as a whole, must know what fields it must hold, and must be able to say which system wins when two disagree. Until that exists, better extraction produces a faster way to read a record that is itself uncertain.

Escalation fails when the trigger is not written down

Supplier problems — a missed delivery date, a certificate that does not arrive, a quality rejection, an invoice that breaks agreed terms — are handled informally, and informally they are handled by whoever has the relationship. The result is that a supplier learns what is tolerated from the person least able to withhold tolerance, and the organisation discovers its policy from its own behaviour.

Naming the trigger changes this. A late confirmation against a stated date, a second non-conformance on the same line, a certificate outstanding beyond an agreed interval: each has a consequence, an owner and a clock. The person holding the relationship then has authority to exercise it rather than having to invent it. That is what separates coordination from negotiation, and it is why the mechanism described in designing AI workflows with approval and exception handling matters more here than the automation does.

What should not be delegated, and to whom

Extracting terms, checking them against policy, assembling the comparison and flagging the deviations is delegable. Deciding whether a supplier's performance justifies continuing to trade with them, whether a specification is acceptable as offered, whether to accept a one-off concession or hold the position, and whether a technical deviation is tolerable are not. Those are judgements about the relationship and about the risk the business is carrying, and they belong to a named buyer who can be held to them.

What to start with on Monday

  • Pick one category with high request volume and observe how ten recent requests actually arrived.
  • Write down the fields a request must carry, and which of them are currently reconstructed from conversation.
  • Define which fields are part of a comparable quotation and which are negotiable, with the owner of that decision.
  • Name the owner of the supplier record and the source that wins when two systems disagree.
  • Write the escalation triggers with their consequences and time limits.
  • Record the baseline: cycle time from request to order, share of requests returned for missing information, and invoice queries raised.

The measures a procurement leader can defend are the proportion of requests that reach a buyer complete, the time from request to approved order, the share of comparisons made on like terms, and the number of post-delivery exceptions per supplier. Those figures usually show that the bottleneck was specification rather than negotiation. The organisational work of moving judgement to the right role is discussed in business and organizational transformation, and the adjacent collection article on sales administration covers the mirror-image problem in another function.