Local Reconciliation Studio

Browser-local reconciliation

Use Local Reconciliation Studio

Both files are read and compared inside this tab. The transactions are not uploaded, and the reports are written here on your device.

Dataset A

CSV or Excel file

Dataset B

CSV or Excel file

Parsing conventions

Date interpretationNumber formatAmount toleranceMax date difference (days)

Map a date and an amount column in both datasets to enable reconciliation.

Card settlement against a sales ledger: when the processor reference differs from the invoice number, a zero-day window leaves two-day-late settlements unmatched until the maximum date difference is raised. When a unique reference does agree, the rows stay paired but the date gap is stated and the pair remains in Needs review. Repeated invoice numbers on one side are listed as duplicates rather than paired, because choosing either copy would not be defensible.

Explore developer and data tools tools

Find related browser-local tools for nearby tasks without starting another search.

Browse all developer and data tools tools

About Local Reconciliation Studio

Reconciliation is the routine of proving that two records of the same money agree. A bank statement against a sales ledger, a payment processor export against an accounting system, last month's file against this month's — in each case you are asking which lines correspond, which are missing from one side, and where the totals diverge. Done by eye in a spreadsheet it is slow and easy to get wrong; done by a tool that uploads your ledger it means handing customer names and amounts to somebody else's server. This studio does the comparison inside your browser tab: you choose two files, tell it which columns hold the reference, the date and the amount, state how your dates and numbers are written, and it produces a row-by-row report you can defend to a reviewer. Every row on both sides comes back with a classification and a reason in plain English, and the original values and source row numbers are preserved so any decision can be traced back to the file it came from.

Work through the input

Choose Dataset A and Dataset B. CSV is read with the same quote-aware parser used elsewhere on the site, and .xlsx workbooks are opened in the tab; if a workbook has several sheets you pick the one to use.
Map the columns for each file separately, because two systems rarely name them the same way. Reference or transaction ID is optional, date and amount are required, currency and description are optional and used for context and reporting.
State the conventions rather than letting the tool guess: whether dates read ISO, day-first or month-first, and whether numbers are written 1,234.56 or 1.234,56. Set an amount tolerance if small rounding differences are acceptable, and a maximum date difference if the two systems post on different days.
Read the warnings shown before you run. Unparseable rows, blank references and repeated references are counted up front so you fix a mapping mistake before it becomes a misleading report.
Reconcile, then work through Summary, Pairs and Needs review. Download the full report or either side's review rows as CSV when you are ready to inspect or share them.

When local reconciliation studio is useful

Bank statement against a sales ledger

Map the bank reference and the ledger invoice number, allow a two-day date window because settlement lags the invoice, and read the exceptions list to find the payments the ledger never recorded.

Processor payout against accounting entries

Set a small amount tolerance to absorb rounding on fees, then use the difference figure and the unmatched totals to explain why the two systems disagree by a specific amount.

Check the output

Matching is deliberately conservative. A row is paired only when the evidence is clear: a unique reference on both sides, or — where references are absent or differ — a single mutually unique candidate within your amount tolerance and date window. Where two or more rows could plausibly pair, the row is marked ambiguous and left for you, because a tool that guesses in a reconciliation is worse than one that asks.
Repeated references inside one file are reported as duplicates rather than matched. If the same invoice number appears twice on one side there is no non-arbitrary way to decide which copy belongs to the other side, so both are surfaced for a human decision.
Dates and amounts are never coerced. A value that does not fit the convention you selected is marked invalid with a reason instead of being reinterpreted, and two-digit years are rejected outright because the century would be a guess. Amounts are compared as whole minor units, so no floating-point drift creeps into a total.
The comparison is limited by what a browser tab can hold. Very large exports may exceed the CSV or workbook limits, in which case the tool says so plainly rather than failing silently; split the file or narrow the date range and run it again.
The reports assist a review. They are not an accounting approval, and a probable pair is a suggestion rather than a signed-off fact. Probable, ambiguous, duplicate, invalid and unpaired rows remain in Needs review for a person's judgement.

Questions about Local Reconciliation Studio

What does reconciliation actually decide?

For every row on both sides it decides one of: exact match, probable match, ambiguous, only in A, only in B, duplicate, or invalid — and records why. The summary shows each side's total, paired and unpaired value, and the difference between the files. Probable pairs remain in the needs-review report.

Why are some rows left ambiguous instead of matched?

Because more than one row could reasonably pair with them. Picking one would produce a report that looks complete but is not defensible, so those rows are listed with their candidate row numbers for you to resolve.

How are amounts compared?

As integer minor units derived from the number format you select, so 0.1 + 0.2 rounding problems cannot appear. Any tolerance you set is applied as an exact integer comparison too.

Do the files leave my device?

No. Both files are read, compared and turned into reports inside this browser tab, and there is no endpoint on this site that accepts a file.