Exceptions across the source-to-pay process: the full map

Exceptions in the source-to-pay process are the transactions that fall out of straight-through handling and need a person to step in: a requisition that stalls before becoming a PO, a change order the supplier never confirmed, an invoice that fails 3-way match, a credit that expires unclaimed. Every stage of the source-to-pay lifecycle produces its own exception types, and the heaviest concentrations sit at the handoffs between stages. Below is the full map, with links to the deeper articles.

Exceptions concentrate at the handoffs

Five handoffs generate most of the exception volume in a typical S2P process: requisition to PO, PO to supplier confirmation, receipt to invoice, invoice to payment, and contract to compliance check. Each one crosses a system, team, or company boundary, and the riskiest cross all three. A process can run cleanly inside every stage and still throw exceptions all day, because the seam between stages has no owner. The requisition was fine and so was the PO; what broke was the conversion between them, and nobody's job description covers that.

Most procurement organizations staff and tool each stage separately. Sourcing owns RFQs, buyers own POs, AP owns invoices, and tax owns withholding. That structure is why the handoffs go dark: every metric and every system ends at a stage boundary.

Upstream exceptions decide the downstream mess

Upstream exceptions rarely look urgent, so they get less attention than an aging invoice queue. Most downstream pain starts here.

• Requisition-to-PO conversion: reqs arrive with missing account strings, wrong units of measure, or a supplier record that fails validation, then sit in a queue until a buyer rebuilds them by hand.
• Change orders and PO amendments: terms move after the PO is issued, the supplier confirms by email, and the ERP never hears about it. The mismatch surfaces weeks later as a match failure.
• RFQ and quote handling: quotes land in inboxes in ten formats, comparison happens in a spreadsheet, and the awarded terms sometimes never make it onto the PO.
• Supplier master data: a duplicate vendor record or a stale remit-to detail looks minor until it blocks every transaction that touches it.
• Maverick spend and contract leakage: purchases that skip the PO entirely, or ride a negotiated contract at the wrong price. The PO-first culture problem covers why policy memos never fix this.

Downstream exceptions are upstream failures arriving late

The procure-to-pay half of the lifecycle is where exceptions get visible and expensive, because a payment deadline is now attached.

• Invoice exceptions and 3-way match failures: price, quantity, or receipt mismatches between the invoice, the PO, and the goods receipt. Usually the biggest queue in AP.
• Duplicate invoices: the same invoice arrives through a portal and again by email with slightly different numbering, and slips past exact-match checks.
• GL coding errors: non-PO invoices coded to the wrong account or cost center, usually found at month-end close, if they are found at all.
• Credits, rebates, and deductions: earned credits that never get applied and rebates that go unclaimed because the tracking lives in a spreadsheet nobody reconciles.
• Tax holds and withholding: payments frozen over a missing certificate or a jurisdiction mismatch, usually discovered when the supplier calls to ask where the money is.

Fixing one stage at a time leaves the seams unowned

RPA is rule-based and single-system, and the handoffs are exactly where the rules run out and the data lives in two places. ERP suites only see their own records, while the riskiest handoffs cross into supplier portals and email threads the ERP never touches. BPOs add labor to individual queues, which speeds up stages and leaves the seams alone. The queues refill either way; why exceptions never go to zero covers the mechanics.

Fragment works the seam itself. Its agents reason across the ERP, MES, supplier portals, email, and spreadsheets the way an analyst would, so a change order confirmed in an inbox and a match failure in the ERP read as one story instead of two tickets. Nothing gets ripped out or replaced, no data standardization is required, a human stays in the loop by default, and production takes weeks. The capacity that frees up goes back into supplier management.

If you are scoping your own exposure, the workflows Fragment covers run the full lifecycle: PO changes, supplier communication, master data, tax holds, credits, and invoices. Or book a demo and bring your ugliest exception queue.