A smaller export. A clearer starting point.

Agree the scope first

Choose a named finance owner, approved supplier set, historical period and as-of date. Confirm what each source date means and how expected credit values were approved. Include older notes or returns where later records depend on them. A limited review still needs complete relationships within that scope.

Use the eight-column CSV contract

The headers are kind, id, supplier, reference, expected_reference, currency, amount_minor and date. Use one CSV with expected, credits and ledger row kinds. expected_reference is populated for credit-note rows and points to the expected return reference. Ledger references point to credit-note references. Use UTF-8 and comma delimiters; quote values containing commas. Extra columns are rejected rather than silently ingested.

Keep the values unambiguous

Dates use YYYY-MM-DD. Monetary amounts are positive whole minor units: 12500 represents EUR 125.00, not EUR 12,500. The supported currencies are EUR, PLN, GBP, USD, CZK, SEK, RON and HUF under this two-decimal convention. The current CSV limit is 400 KB and 2,000 total source rows. Negative reversals, tax-basis differences and incompatible currency conventions need prior normalization.

Remove fields the review does not use

Exclude contact names, personal emails, addresses, bank details, tax identifiers, product descriptions, attachments and free-text comments. The matching engine does not need them. Replace supplier names and document identifiers with consistent codes if agreed. Keep the private lookup table on the client side so the finance team can reconnect findings to original evidence.

Check links before sharing

The same supplier must carry the same code in each row kind. Split credits must preserve the expected reference; partial applications must preserve the credit-note reference. The preparation tool keeps these links within one snapshot. Its codes are not stable across separate runs, so do not merge independently coded files or assume the codes refer to the same suppliers.

Understand what coding does not remove

Amounts, dates, currencies and transaction patterns stay in the export. Those fields can reveal business relationships or help identify a person. This is pseudonymisation, not anonymity or encryption. Treat the coded file as confidential and never send the private lookup table with it. An NDA alone does not establish an approved processing environment.

Use the tool, then review the result

Download the standalone preparation tool and use a synthetic example first. Open it on an approved company device and disconnect before choosing confidential data. Inspect the coded export locally without resaving it through spreadsheet auto-formatting. Keep the original source and mapping separately. Only an agreed, approved pilot should receive the minimized export.

Complete the transfer decision

Before a real-data transfer, confirm the operator, purpose, named access, storage location, processors, transfer method, retention period and backup deletion limits. Agree appropriate confidentiality and processing terms. The public enquiry is for a conversation, not financial attachments. These controls must be verified for the actual pilot, not inferred from a webpage.

CSV fields used by the current parser

Your action checklist

1. Use the downloaded technical CSV, not the manual Excel tracker.
2. Keep comma delimiters and exact headers.
3. Agree sign and net/gross normalization.
4. Include opening items and confirm completeness.
5. Do not send files through the public contact form.

Practical questions