How do I estimate month-end AP accruals without guessing?

Accounts payable is a recorded liability supported by a received invoice; an accrued expense is a liability you've incurred but haven't received (or processed) the invoice for. Both are current liabilities - AP is invoice-backed and precise, accruals are estimate-backed and reverse when the real invoice arrives.

PO-based accruals fit goods and contracted services where receiving is tracked - the receipt is your evidence. Invoice-in-hand accruals cover everything captured but stuck in workflow at cutoff. Run-rate accruals cover predictable recurring spend with lagging invoices. Most companies need all three; the mix shifts toward PO-based as procurement discipline matures.

Debit the expense (or asset) account, credit accrued liabilities (a separate account from trade AP), typically flagged to auto-reverse on day 1 of the next period. Keep accruals out of the AP trade account so the sub-ledger reconciliation stays clean.

Run the received-not-invoiced report (receipts without matched invoices), value each line at PO price times received quantity, exclude lines already invoiced or canceled, and book the total by GL account from the PO coding. This is the most defensible accrual basis because every line traces to a receipt.

For each recurring vendor, accrue the contracted amount or trailing 3-month average for any period without a posted invoice. Maintain the list as a standing schedule - the review each month is "did the invoice arrive?", not "what should this be?"

Yes - they're your highest-quality accrual source because the amount is known, not estimated. Pull everything captured but unposted at cutoff, accrue at invoice amount with the coded GL distribution, and reverse next period. Systems with real-time invoice status make this a report, not a hunt.

Triangulate from commitments: open POs and receipts first, then contracts and statements of work for service spend, then vendor statements for high-volume accounts. The discipline is keeping commitment data (POs, contracts) in systems rather than inboxes so "what's coming" is queryable.

No. Set a per-item floor (commonly $1K-$5K mid-market, scaled to your materiality) and accrue the small stuff as a single historical-average plug if it's collectively meaningful. Document the threshold; auditors accept consistency plus a reasonableness check far more readily than heroic precision.

Default to system data (POs, receipts, in-flight invoices, run-rates) and only ask humans about what systems can't see - unbilled project work, verbal commitments. Send those few budget owners a pre-populated template on day 1 with a 24-hour SLA. If you're chasing 30 people, your commitment data lives in too many heads.

They move the upstream commitment into a system: purchase requests and POs create a record before the invoice exists, receiving confirms delivery, and centralized intake captures invoices the day they arrive. Accrual quality is a by-product of procure-to-pay discipline - you can't estimate what you never recorded.

Accrue from shipment data, not invoices: apply contracted or historical rates to the period's receipts/shipments, true up when carrier and broker invoices arrive (they lag 30-60 days). Importers should hold a standing duty accrual driven by customs entries.

Accrue from the engagement side: approved timesheets, contract rates, or straight-line of the SOW value. Ask firms for unbilled work-in-progress at quarter-ends. Long-lag services are the classic source of accrual misses - the run-rate method plus an annual true-up keeps them bounded.

Accrue cost-to-date from committed subcontract value times percent complete (from the project manager or pay application), less amounts already invoiced - and track retainage separately. The commitment ledger is the backbone; companies without committed-cost tracking are estimating blind.

Same concept, different evidence quality. GRNI is receipt-backed and mechanical (the system knows goods arrived uninvoiced); a general expense accrual covers obligations without receipt records - services, utilities, unbilled work - and relies on estimates. Auditors treat GRNI as system output and general accruals as judgments to test.

Yes, with the right scope: automation reliably assembles the receipt-backed and in-flight-invoice portions (the data already exists), and AI adds value flagging anomalies and predicting recurring-spend gaps. Human review still owns the judgment accruals. The win is mechanical assembly of 80% of the schedule, not a black-box number.

One line per accrued item: vendor, description, basis (PO/receipt, invoice-in-flight, run-rate, estimate), source reference, amount, GL coding, and preparer/reviewer sign-off - plus a summary tying to the JE total and a prior-month true-up section. The test: a reviewer can re-derive any line without asking you.

Because requests, POs, receipts, and in-flight invoices all live in Stampli's procure-to-pay workflow, the data an accrual needs - received-not-invoiced POs, captured-but-unposted invoices - is queryable at close instead of collected by chasing 30 budget owners. Stampli AI's immediate capture and coding also shrinks the accrual population itself: the more invoices are fully processed by cutoff, the less you estimate.