Procure-to-pay looks tidy on a slide. Requisition, approval, purchase order, receipt, invoice, match, payment. Seven boxes, one arrow each. The money does not leak in the boxes. It leaks in the gaps between them — and the gaps are where nobody's job description ends up. The most persistent evidence of this is duplicate payments. Recovery audit firms consistently report that duplicate payments are the most frequent category of recoverable loss they find, and that finding has not changed across two decades of steadily better systems. An entire industry exists to recover money that well-run organizations paid twice, to suppliers they chose, through processes they designed.
Where the Money Actually Goes
Maverick spend. Purchases made outside the process — no requisition, no contract, no negotiated rate. The organization has a framework agreement with a preferred supplier at a discount, and a manager buys the same item retail because it takes four days to get a PO. The loss is the price difference multiplied by every instance nobody counted. Approval theatre. Every requisition routes through four approvers. None of them rejects anything, because none has the information required to disagree. What this adds is delay, which in turn drives more maverick spend, which is the loop most organizations are stuck in. Receipting gaps. Goods arrive and nobody records receipt. Now three-way matching cannot run, the invoice goes to manual review, and the control that was supposed to prevent overpayment has been quietly disabled by a missing keystroke. Invoice exceptions. Price variance, quantity variance, missing PO reference, wrong entity, wrong currency. Each exception is handled by a human, each costs substantially more than a clean invoice, and the exception rate — not the transaction count — determines the real cost of running accounts payable. Duplicate payments. The mechanics are well documented and boringly consistent: the same invoice submitted under slightly different vendor records, a partial payment followed by payment of the full amount, invoices paid against both a blanket PO and a line PO, and vendor-name variations that defeat the duplicate check. Every one of these is a data problem dressed as a payment problem. Payment timing. Paying early costs working capital. Paying late costs early-settlement discounts and, increasingly, supplier goodwill. Most organizations do both simultaneously, in different parts of the ledger, and report an average that conceals it.
Why the Gaps Persist
Procure-to-pay crosses at least three functions. Procurement owns sourcing and supplier terms. Operations owns receipt. Finance owns the invoice and the payment. Each function optimises its own metrics, and every one of the leaks above happens at a boundary where no single metric is accountable. This is why buying a system rarely fixes it. The technology assumes a process discipline the organization does not have; the gaps simply reappear as workarounds inside the new tool. A procure-to-pay platform with a 40 percent exception rate is an expensive way to route the same problems to the same people.
| Handoff | Review point |
|---|---|
| Buying to procurement | Track off-contract purchases and speed of the authorised path. |
| Receiving to accounts payable | Record the actual receipt needed for matching. |
| Master data to payment | Review supplier duplicates and bank-detail change controls. |
| Payment to treasury | Separate early, on-time and late payment outcomes. |
Qualitative summary of this article's source text, not a measured outcome or performance estimate.
Fixing the Flow, Not the Boxes
- Measure the exception rate first. What percentage of invoices match cleanly and pay without human intervention? Below 70 percent means the upstream data — POs, receipts, vendor master — is the problem, not AP capacity.
- Clean the vendor master before anything else. Duplicate vendor records are the direct cause of the most common duplicate-payment pattern. Deduplicate, standardise naming, enforce bank-detail change controls, and deactivate dormant records.
- Make the compliant path the fast path. Maverick spend is a symptom of friction. If a requisition clears in an hour, people use it; if it takes a week, they do not, whatever the policy says.
- Cut approval layers and raise thresholds. Approvals that never reject anything add cost and delay without adding control. Replace low-value pre-approval with post-transaction sampling.
- Enforce receipting as a hard control. No receipt, no match, no payment — and make receipting take seconds on a mobile device rather than minutes at a desk.
- Run duplicate detection continuously, not annually. Recovery audits find real money, but they find it after it left. Pre-payment detection on fuzzy vendor names, amounts and invoice numbers is where the value is.
- Report payment timing by distribution, not average. Show the share paid early, on time and late. Averages hide the two opposite failures that cancel each other out on paper.
What Changes With Automation — and What Does Not
Machine reading of invoices is now genuinely good. Extraction accuracy on unstructured documents has moved from frustrating to reliable, and matching engines can handle variance tolerances, partial receipts and multi-line allocations that defeated older rules. What automation does not fix is the vendor master, the missing receipt, or the fact that three functions each own a third of the process. A model that reads an invoice perfectly still cannot match it to a purchase order that does not exist, and it will pay a duplicate with great efficiency if the vendor record is duplicated too. The organizations that get the most out of procure-to-pay automation are the ones that did the unglamorous work first: cleaned the master data, removed the approval layers nobody used, and made receipting non-optional. The ones that automated on top of the existing mess bought faster leakage and better dashboards to observe it.
Common Questions
What is procure-to-pay?
The end-to-end cycle from requisition and approval through purchase order, goods receipt, invoice matching and payment — spanning procurement, operations and finance rather than sitting within any one of them.
Why do duplicate payments keep happening?
Usually because of vendor master data: the same supplier exists under multiple records, or invoices are entered with different references. Recovery auditors consistently report duplicates as the most common recoverable loss.
What is a good invoice exception rate?
Fewer than 20 to 30 percent of invoices requiring manual intervention is a reasonable target for a mature process. Higher rates almost always indicate upstream problems with purchase orders, receipts or supplier data.
Does procure-to-pay software eliminate maverick spend?
No. Maverick spend reflects process friction. Software helps only if the compliant route becomes faster than working around it; otherwise the workarounds continue outside the new system.
Procure-to-Pay Assessment — Outpace traces your real spend flow end to end, quantifies leakage at each handoff, and fixes the master data and approval design before automation locks the losses in.