Supplier Returns & Credit Note Reconciliation

Reconcile approved monetary supplier returns, issued credit notes and their applications from three authorized structured exports. Separate missing or partial issuance from an unapplied credit balance. Every finding retains exact document IDs and amounts; it is a review candidate, not proof of money owed.

Quick start

Run the prefilled synthetic example. A EUR 1,000 approved return, EUR 800 credit and EUR 500 application produce a EUR 200 issuance difference and EUR 300 application balance. These are distinct stages, not a confirmed recovery total. Download the complete dataset report or FULL_RECORDS for individual return and credit review records.

Use the Input editor or Apify API with example-input.json. The Actor does not fetch URLs, read PDFs, connect to an ERP, send supplier emails, make payments or use an external AI model.

Input contract

scope is an exact pseudonymous accounting-book identifier. cutoffDate is an inclusive ISO date (YYYY-MM-DD). Explicit completeness flags for returns, credits and applications must be true only after the caller has verified the entire relevant scope. False flags prevent definitive missing-credit or unapplied-credit classifications.

Every row includes an exact supplierId and supported uppercase currency. IDs are nonempty strings up to 80 characters, unpadded and case-sensitive; leading zeros matter. IDs must be unique within each table and accounting scope; qualify document numbers yourself when different suppliers reuse a number. Amounts are positive decimal strings with at most two decimal places, at most 10^14 minor units. No floats, exponent notation, thousands separators, zero or negative amounts.

Table	Required fields
returns	returnId, supplierId, currency, approvedAmount, approvedAt, creditDueDate
credits	creditId, supplierId, currency, amount, issuedAt, applicationDueDate, allocations
applications	applicationId, supplierId, currency, creditId, amount, appliedAt

A credit's allocations contains explicit {returnId, amount} objects. Allocations must add up exactly to the credit amount; a shared credit is never counted in full under each return. Applications link to a credit, not to a guessed return. Supplier and currency must agree along every link. Missing, duplicate or conflicting links fail explicitly. Unknown fields are rejected.

USD, EUR, GBP, CAD, AUD, PLN and CZK use two decimal places. There is no currency conversion. Choose one consistent amount basis (such as net), resolve cancellations and corrections upstream, and supply contractually agreed due dates. Dates are not inferred from law or payment terms. Pre-approval credits and pre-issue applications are rejected. Future documents are excluded from the selected cut-off; a due date becomes overdue only on the following day. Overcredits and overapplications retain their negative signed differences for review.

Limits: 10,000 combined table rows, 20,000 allocations, 4 MB input and 8 MB per output artifact. Split oversized input into independent complete scopes, preserving all related documents. This Actor cannot authenticate approvals, deadlines, completeness, balances or the supplied identities.

Results and repeat checks

One dataset item is a complete report containing returns, credits, completeness, source counts and evidence. Monetary results are integer minor units; EUR 200 is 20000. humanReviewRequired is always true and confirmedRecovery always false. FULL_RECORDS and RECORDS are machine-readable review arrays; CHANGES explains differences from the previous complete supplied export; SUMMARY contains report metadata.

snapshot returns a full check. changes also performs and delivers a complete check but identifies differences from a previous successful run. Supply a persistent stateStoreId and the same scope. A first run establishes a baseline. Removed input rows mean absence from the new export, not supplier cancellation. Older cut-offs require a separate store. Serialize runs using one store and scope; conflicting state is rejected. Only verified complete deliveries advance the baseline. Resuming the same delivered run validates the saved report and avoids a second dataset delivery; changed or uncertain storage stops explicitly.

Pricing

$0.75 per completed report, platform usage included. One report covers the bounded input, not one fee per financial row. A new run is a new paid complete check, including unchanged data, no exceptions or incomplete-source findings. Invalid input and insufficient report budget do not deliver a paid report. Do not rerun unchanged exports unnecessarily. Long-term storage and subsequent external reads are subject to Apify platform rules. No third-party API subscription or AI bill is required.

Privacy and limitations

Submit only records you are authorized to process. Use pseudonymous book, supplier and document IDs; amounts and dates remain sensitive. No raw financial rows are logged or forwarded to third-party services, but Apify stores Actor inputs, datasets and key-value artifacts under your account's access and retention settings. This is not local-only or zero-retention processing. Review resource sharing and retention before uploading actual records. Do not put confidential data in public tasks or support issues.

The report is deterministic evidence review, not accounting, tax, legal advice, payment approval or an audit certification. Real discrepancies require review against underlying documents. For support, provide a synthetic reproduction and run ID without confidential data.