Prepare a Bank Statement CSV for Reconciliation

Reconciliation becomes fragile when every matching rule knows the names used by a specific bank. A better pipeline normalizes the statement schema first, then lets reconciliation operate on stable transaction fields.

Normalize before matching

Start by mapping the bank's headers to the fields your reconciliation logic expects. For the source-backed example, that means TRAN_DATE → transaction_date, DR → debit_amount, CR → credit_amount, PARTICULARS → narration, and CHQNO → cheque_no. Keep balance or source-location fields if they help with control totals or traceability.

This removes bank-specific header aliases from the matching layer. A rule that compares transaction_date and debit_amount no longer needs branches for TRAN_DATE, bookingDate, or another bank's label.

Preserve evidence fields

Do not discard a payment reference, cheque number, counterparty name, IBAN, original currency, exchange rate, or category simply because it is not part of the minimum canonical schema. Those fields can improve matching and explain exceptions later.

Choose a small required core and retain optional evidence fields with clear names. Reordering them after the required fields makes downstream CSV review easier without changing their values.

Add value checks after schema mapping

Header normalization does not prove that dates parse correctly, that debits and credits use the expected sign convention, or that balances reconcile. Those are separate checks. If a bank provides one signed amount column, splitting it into debit and credit requires a transformation beyond this mapper.

Before importing, compare row counts, inspect opening/closing balances where available, and sample transactions with blank references, credits, debits, and foreign-currency data.

The browser-local normalizer can prepare the schema without uploading statement contents to Nablyx.

Prepare bank CSV for reconciliation

See the transformation

Awkward source headers become import-ready columns.

Rename the columns you need while keeping every value under the correct header.

cust_idcustomer_id

given_namefirst_name

mailemail

✓ Values stay with the right column