Confirmations: distinguish nonresponses from confirmations
A confirmation request is an evidence procedure, not an automatic validation of the amount sent. A useful worksheet records what was selected, whether a reliable reply arrived, what differed, and which assertion any alternative evidence supports. Silence, an exception and a missing document are three different states.
This guide concerns an issuer audit conducted under PCAOB standards, using AS 2310 as read on October 25, 2026. Its current version is effective for audits of fiscal years ending on or after June 15, 2025. Assume a December 2025 year-end audit. The fictional working papers below illustrate the mechanism; ordinary investors are not presumed to have the auditor’s confirmation files.
Define the assertion and retain control
AS 2310 .08–.19 address request design, knowledgeable confirming parties, auditor control and response reliability. Identify the fact to test before recording an amount as supported: existence of a receivable, a payment term, or another relevant assertion.
Use an invoice identifier, customer legal name, balance date, currency and request destination. The auditor-controlled process should connect the selected item to a request and a response obtained directly from the intended source. A management-forwarded PDF is not equivalent merely because it contains a customer logo.
Under .22, a reply returned to someone other than the auditor calls for contacting the confirming party and requesting direct resending. If no direct response follows, the situation is a nonresponse. That is different from automatically declaring the document fraudulent. Reliability must be evaluated and additional evidence obtained.
Work a five-state balance ledger
All values are hypothetical thousands of currency units, not numbers of customers. A selected receivables sample has a recorded carrying amount of 100. Reliable replies agree with balances totaling 50. Replies relating to recorded balances of 10 contain exceptions. Selected balances totaling 40 receive no reply after follow-up.
| Initial state | Recorded sample amount | Evidence status |
|---|---|---|
| Reliable reply without exception | 50 | Supports specified facts subject to the response assessment |
| Reliable reply with exception | 10 | Difference requires investigation |
| No reply | 40 | Not confirmed; alternative procedures considered |
| Total selected balances | 100 | Partition reconciles before further work |
The 10 is the carrying amount of balances with exceptions, not automatically a misstatement of 10. Likewise, the 40 nonresponse amount is not automatically a bad-debt loss. The ledger prevents both shortcuts.
Assume subsequent receipts of 25 are traced to the particular outstanding invoices, supported by reliable payment evidence and appropriate supporting documents. Under the stated facts, those procedures provide evidence for the identified balance assertion on 25. The remaining nonresponse amount is 40 − 25 = 15. It remains unresolved by those procedures.
Do not call the initial 50 plus the subsequent 25 a 75% customer response rate. These are monetary amounts, and 25 was supported by another procedure. A response-rate calculation would need customer or request counts with its own denominator.
Investigate the exception rather than netting it away
Suppose investigation of the 10 exception balance identifies a valid pre-year-end return of 2 that the issuer failed to record. Assume independent records support the remaining 8, the return’s timing and its effect on the receivable. The selected ledger now reconciles:
Supported amount 50 + alternative-supported amount 25 + reconciled exception balance 8 + identified overstatement 2 + unresolved amount 15 = recorded sample 100.
The supported carrying amount is 83 for the specified exercise, the identified overstatement is 2, and 15 is unresolved. This is not a conclusion that all 83 is collectible or that the remaining 15 is misstated. Nor does it measure the entire account outside the selected sample.
AS 2310 .20–.23 and Appendix C distinguish evaluating exceptions, following up nonresponses and performing alternative procedures. An exception can indicate a misstatement, a control issue, or both; the evidence determines its meaning.
A timing explanation also needs substantiation. If the customer recorded a payment before year-end and the issuer recorded it later, inspect bank and settlement evidence, not just the word “timing” on a reconciliation. A plausible explanation is not a completed procedure.
Map receipts to the right debt
Subsequent cash can be persuasive only when it addresses the item being tested. A payment of 25 made by the same customer could instead settle a January invoice, represent an advance, include a deduction, or originate from an unrelated arrangement.
Trace remittance details to invoice numbers, currency, gross balance and any netting. Identify who paid and whether the receipt belongs to the reporting entity. For a batch receipt, document the allocation instead of assigning it to whichever old invoice closes the worksheet most neatly.
If only 20 of the assumed 25 can be linked to the selected year-end invoices, the alternative-supported amount becomes 20 and the unresolved nonresponse amount becomes 20. With the exception investigation unchanged, the ledger becomes supported 78, known overstatement 2 and unresolved 20. The original 100 still reconciles, but the evidence conclusion has changed.
This sensitivity tests evidence linkage, not a new estimate of default probability. It shows why an aggregate bank deposit cannot simply be applied to an aggregate sample.
Boundaries that change the conclusion
A customer may accurately confirm that it owes 50 while lacking the resources to pay. The reply can address existence or contractual information without resolving valuation or expected credit losses. Inspect aging, subsequent collection, disputes and other relevant evidence separately before concluding on recoverability.
A second boundary concerns negative requests, which ask for a response when the recipient disagrees. AS 2310 .12–.13 explain their weaker evidence and that negative requests alone do not provide sufficient appropriate evidence for the relevant risk. No reply to a positive request cannot be relabeled as a successful negative confirmation.
Electronic transmission also does not remove the control question. An intermediary may facilitate direct exchange, but its reliability and the possibility of interception or override require assessment under Appendix B. A secure-looking interface is not the same as evidence that the particular exchange remained controlled.
Execute a document-level evidence table
| Field | What to retain | What an unresolved field means |
|---|---|---|
| Selected item | Invoice, customer, amount and date | Population and assertion are not aligned |
| Channel | Destination and independently assessed source | Response reliability not established |
| Reply | Direct response and terms answered | No agreement may be inferred from silence |
| Exception | Recorded versus external facts | Difference remains uninvestigated |
| Alternative work | Receipt allocation, shipping or signed terms | Procedure may not address the item |
| Conclusion | Assertion, supported amount and remaining issue | No universal “confirmed” label |
Keep the selected carrying amount stable while statuses change. Do not remove nonresponses from the denominator or replace difficult items with easy respondents. Otherwise the final apparently clean list describes a different selection.
Public reports may describe confirmation limitations, auditor procedures or a material dispute. They usually do not disclose invoice-level replies, private bank evidence or the complete audit selection. A public researcher should record what the issuer or auditor actually states and what cannot be verified, rather than fabricating an 83/100 coverage statistic for a real company.
When a direct reply conflicts with a ledger, preserve both source versions and assess source reliability, balance date, currency, settlement timing and side agreements. Management’s explanation and external evidence should remain separately attributable until reconciled.
Update the worksheet if a reliable reply arrives, a receipt allocation changes, a credit note establishes an error, or contradictory information changes the risk assessment. Resolve the specific assertion and amount; do not overwrite the original nonresponse history as if the confirmation had always succeeded.
Keep the sample conclusion inside its population
Suppose the complete recorded receivables account is 1,000 while the selected balances in the main case are 100. The supported sample amount of 83 does not establish that the other 917 is correct or incorrect. Nor does 83/100 become a measured assurance level for the whole account. The selection design and other work would be needed for any broader evaluation.
Maintain separate account-total, selected-total and evidence-status fields. If there are ten requests with different balances, a high monetary response share can coexist with a low count response share. Neither statistic substitutes for the assertion-specific assessment of reply reliability.
Also record which questions were answered. A reply agreeing to a balance might omit the requested information about side agreements or collateral. It can be complete for one question and incomplete for another. Return the unanswered question to the evidence ledger instead of marking every field successful because one number agrees.
FAQ
Is an unanswered positive request evidence of agreement?
No. Follow-up and appropriate alternative procedures are separate evidence steps.
Does a confirmed receivable prove the customer can pay?
No. Existence, terms and recoverability require different evidence.