CSV Reconciliation Tool

This CSV reconciliation tool compares two sets of records supplied by you: business records such as orders, and receipts or settlement records. It proposes groups using exact references or an explicitly mapped batch identifier, then checks their amounts and dates. A suggestion is not a completed reconciliation. You must inspect the members and confirm the group before it enters the confirmed output. The tool does not connect to a bank, post accounting entries, cancel an invoice or notify a customer.

A common task is matching several orders to one settlement. For example, orders O1 and O2 contain 40 and 60 in the same currency and batch P1. Settlement S1 contains 100 and batch P1. Grouping on that batch produces a proposed 100-to-100 match with a difference of zero. A second settlement of 100 without that reference is not automatically treated as another match. Equal amounts alone cannot prove that records describe the same transaction.

Import the two roles independently. Confirm each text file encoding, delimiter and header, or select and verify an XLSX sheet and data area. Pasted tables and the current workbench table are also available. Values remain text until an explicitly selected amount field is interpreted, preserving long identifiers and leading zeros. Each side may map a unique record ID; without one, the task uses internal source identities, never an amount-and-date combination as a substitute identifier.

Map amount and date fields and choose currency either from a field or from an explicit whole-file constant such as USD. Currency is never inferred from the interface language. Map reference numbers for exact matching, batch or collection identifiers for group matching, and optionally customer fields on both sides to restrict automatic candidates to the same customer. A customer restriction applies only when both sides have that field mapped. All such keys compare exact text without automatic fuzzy matching.

Confirm decimal and grouping separators and the notation used for negative values. Then state whether each side keeps its source sign or reverses every amount sign. Refunds remain signed values; the tool never takes absolute values to make them appear to match. The maximum decimal-place setting rejects excess precision instead of silently rounding it. Review raw and normalized amounts together, especially when one source represents incoming payments as negative ledger entries.

The arithmetic is business total minus explicitly mapped fees minus net settlement, producing the actual difference. A mapped fee retains its source sign and is subtracted; it does not inherit the amount sign reversal. If a refund or fee reversal has a different convention, prepare or explicitly interpret that source before confirming. Fees can be mapped on either side, but mapping the same economic fee twice would deduct it twice. Review the member details and your source definitions.

A business total of 100 and net settlement of 98 produce a difference of 2 when no fee field is mapped. If you explicitly map a fee of 2, the difference becomes zero. The tool never invents that fee merely because the numbers happen to differ. Decimal arithmetic also makes 0.10 plus 0.20 equal 0.30. An absolute tolerance changes whether a difference is acceptable, not what the difference is: a tolerated 0.01 is still displayed and exported as 0.01.

Exact-reference mode forms one-to-one candidate pairs using a shared reference and currency. Several eligible pairs that compete for the same source record are marked ambiguous and require your choice. Explicit-batch mode aggregates the complete mapped batch on each side, supporting one-to-many and many-to-one relationships without an unrestricted search for subsets that happen to sum to the same amount. A blank reference or batch does not become a shared group for unrelated records.

Manual selection lets you choose up to two hundred whole valid records from the two sides. Search and pagination make a bounded group review practical. The current totals update as you select records, and checking the group calculates its fee, difference, date range and occupancy before confirmation. Manual selection is your declared grouping evidence; it does not mean the tool independently established a commercial relationship. Different currencies cannot be placed in the same group.

The maximum date distance is optional. It compares the earliest and latest member across both sides, using calendar days for pure dates or elapsed hours for instants. Choose the same time meaning on both sides and declare the source formats. An unzoned timestamp needs an explicit source zone; browser location does not decide it. A group outside the date limit is reported separately from an amount mismatch, and neither can be confirmed merely by clicking the confirm button.

Each candidate shows its evidence, currency, business amount, explicit fee, net settlement, actual difference, date distance and member count. Expand the member lists to inspect raw and normalized values and file, sheet and logical record references. The confirmation button states how many records it will occupy. Once confirmed, each member belongs to exactly one confirmed group. A competing candidate updates to occupied and cannot be confirmed until the original group is undone.

Undo confirmation releases every member of that group. The remaining candidates can then be reviewed again. Re-running checks the complete set of confirmations for duplicate occupancy rather than trusting a disabled interface button. Changes to source data or interpretation rules invalidate previous approvals. Saved rules omit confirmations, and loading them requires a fresh run. A new manual candidate can coexist with earlier confirmed groups, but it is subject to the same occupancy check.

This version does not allocate part of a record across several groups. A single payment of 100 cannot be divided into 40 and 60 while both pieces are claimed as separate completed matches. If your task needs such allocation, retain it as unresolved and use an appropriate allocation workflow. Combining several complete payment records into one explicitly supported group is different from splitting one record into artificial pieces.

The totals view keeps each side and currency separate. It shows valid input amount, confirmed amount and unconfirmed amount so they can be reconciled. Invalid amounts, dates, identities or currencies stay in input-error reports and are excluded from valid totals. The error-record count shown for a side is a whole-side scope, not a figure to add repeatedly across currencies. An issue count can exceed the number of error records because one record may have several problems.

Download confirmed groups and their member details as separate selected scopes, or inspect unconfirmed records, candidate ambiguity, differences and input errors. A confirmed group is not reduced to an unexplained pass label: its totals and members remain available. Preview search does not silently filter exported scope. CSV and TSV are generated with declared encoding and read back to verify structure and values. Spreadsheet formula protection is separate from the explicitly acknowledged raw-text option.

The processing worker runs locally, supports cancellation and stops on input, candidate, member or time limits. It does not present a limited scan as complete reconciliation. Files and confirmations stay in session memory; refreshing or closing clears them. Website resources still generate ordinary page requests, but source transaction content is not submitted to a reconciliation API. Keep your source files and reviewed exports if you need durable evidence after the browser session ends.