Research · Published:
How should AP study partial-receipt evidence before invoice review?
An evidence study of ordered, received, invoiced, and accepted quantities when outsourced AP prepares a three-way-match exception.
Research question and method
Research question: when an invoice covers more or fewer units than a receipt record, what evidence can outsourced accounts payable prepare without deciding whether the invoice should be paid? The study compares ordered quantity, recorded receipt, invoice quantity, returns, corrections, and owner responses across a bounded sample. GAO internal-control guidance informs the treatment of reliable information and documented review. NIST access-control guidance informs who may view or change purchasing and receipt records. The sample should include a full receipt, split delivery, over-receipt, short shipment, return, service line, unit-of-measure conversion, backdated correction, and receipt recorded against the wrong purchase-order line. These cases test different explanations. The research does not assume that every mismatch is an error or that a match proves the goods were accepted.
Evidence population
Start by freezing the comparison point. Record the entity, purchase order and line, ordered quantity and unit, invoice identifier, invoiced quantity and unit, receipt transactions, return transactions, source-system timestamps, extraction time, and the review cutoff. Facts must stay attached to their sources. "The system shows six received units as of 10:00 UTC" is a fact when the timestamped record supports it. "Four units are missing" is analysis based on a stated comparison. "Pay six now" is a decision. The packet should never collapse those categories. An outsourced specialist may collect the values and show the arithmetic, but the buyer, receiving owner, and finance approver retain responsibility for acceptance, tolerance, accrual, exception override, and payment decisions.
Case analysis
Unit-of-measure cases deserve their own test because equal-looking numbers can describe different quantities. A purchase order may use cases, the receipt may record individual units, and the invoice may use packs. The preparer should not invent a conversion. Locate an approved item or purchasing record, name the conversion source, show the calculation, and stop if sources conflict. A later master-data correction should not erase the value used at the original review. Preserve the initial comparison and append the corrected calculation. The same rule applies to service receipts. A recorded service entry may reflect administrative confirmation rather than proof of contractual completion. The study can show that the record exists and who created it; it cannot establish that the service was delivered or that recognition is correct.
Reproduction test
The reproduction exercise asks two reviewers to build the quantity bridge independently. Each bridge begins with ordered quantity, adds approved changes, separates each receipt and return, and ends with the invoiced amount at the chosen cutoff. Reviewers must cite the exact line or event for every number. Compare their treatment of reversed receipts, duplicate events, late entries, and inaccessible warehouse records. If their totals differ, retain both calculations and locate the rule or source behind the difference. Do not average them or select the cleaner result. The disagreement is evidence about workflow clarity. Reperform the exercise after a correction to confirm that the record preserves what the first reviewer could see before the correction arrived.
Operating boundary
A queue for partial-receipt exceptions needs descriptive states. Examples include awaiting receiving evidence, unit conversion unresolved, return pending, purchasing change pending, invoice correction requested, owner override recorded, and ready for authorized review. None means approved for payment. The status should also name the current owner and the last evidence event. Aging begins from a defined event, such as first complete comparison, rather than whichever timestamp makes the queue look faster. This research does not supply a target age or tolerance percentage. Those thresholds depend on the company's policy, goods, system, contract, and risk. A bounded outsourced lane can maintain the queue and follow up on named evidence while keeping judgment with the responsible employee.
Interpretation
Controls should prevent the preparer from repairing a mismatch by altering source records. If support can edit receiving data, change purchase-order quantities, and clear the resulting exception, the evidence no longer separates preparation from authorization. Least privilege suggests granting the minimum functions needed for the assigned lane. Where a correction is legitimate, the source-system history should name its actor, time, prior value, reason, and approving owner under company policy. The AP packet can then link to that history. An email saying "received" may prompt follow-up, but it should not silently replace the designated receiving record. Urgency does not change the evidence standard. It changes the escalation timing.
Limitations
Limitations are substantial. Public control guidance does not determine ownership transfer, acceptance terms, materiality, accounting recognition, payment tolerance, or legal rights under a purchase agreement. Systems model receipts and reversals differently. A sample chosen from visible exceptions may miss silent errors in records that appear to match. Timestamp quality depends on configuration and user behavior. The proposed quantity bridge is an analytic aid, not an accounting conclusion. It must identify exclusions, inaccessible records, manual adjustments, and assumptions. Management should approve the population and interpretation rules before using findings to change access, staffing, or workflow. No result from the sample should be presented as a market benchmark or a performance promise.
Evidence-led conclusion
Evidence-led conclusion: a partial-receipt exception becomes reviewable when ordered, changed, received, returned, and invoiced quantities can be reconstructed at a stated cutoff from named sources. The most useful outsourced AP contribution is a transparent bridge plus an exception history, not a judgment that the mismatch is acceptable. Separate arithmetic from business acceptance and preserve every correction. If two reviewers cannot reproduce the same bridge, the packet should remain open with the conflicting source or rule identified. If they agree, an authorized owner still decides the commercial and accounting outcome. This approach keeps routine preparation moving while protecting the meaning of receipt and approval records.
Sources
These primary sources support the control principles and evidence boundaries in this report.
FAQs
Are the planning numbers benchmarks?
No. They describe a testable workflow shape and are not promises, market averages, or production targets.
What should an outsourced AP assistant own?
Repeatable preparation, documentation, status tracking, and follow-up within least-privilege access. Named finance owners retain approval and payment decisions.
When should an item be escalated?
When evidence is missing, a request changes payment details, a duplicate or fraud signal appears, or the item falls outside the written rule.