How to tie bank statements so the other expert can rerun it

Two words that get used loosely

A statement ties when the rows as read, added to the printed opening balance, reach the printed closing balance to the cent. A run of statements for one account chains when each month's closing balance is the next month's opening balance, with no gap and no overlap. Tying is a property of one statement. Chaining is a property of the sequence. An exhibit built on statements that do neither is a schedule of numbers with nothing underneath it, and the first thing an opposing expert does is check whether it ties.

Why it matters more when the exhibit is filed

A summary of voluminous records is admissible when the underlying records are what the summary says they are, and when the other side can examine them. A tie is the cheapest proof that the rows are the bank's rows and not a typist's. A chain is the cheapest proof that the months are all there. When both hold, the argument moves from whether the figures are right to what they mean, which is where the examiner is paid to be. When either fails and the exhibit does not say so, the exhibit is the argument.

The procedure

1. Get complete runs, per account, from the bank where possible. A statement produced by a party is read the same way, but it is declared as party-sourced and the authenticity notes apply to it. A statement from the bank is taken as the bank's record on the reviewer's declaration.
2. Keep every file exactly as received. Do not print to PDF, do not re-save, do not merge. The register records each file's SHA-256 as received, and a file that has been touched will not match the one the other side gets from the same source.
3. Tie each statement. Every row is read against the printed opening and closing balance. A statement that ties is marked as proven. A statement that does not tie is marked as read but not proven, with the difference shown, and it is never silently corrected.
4. Chain each account. Consecutive statements must chain to the penny. A break is shown with its amount, and the note names the usual causes: a missing statement, an adjustment posted between the two, or a misread balance.
5. When a row was misread, fix it in the open. A hand edit made in review is recorded as read versus as exported, on the row, in the grid, and in a section of its own at the end of the exhibit. An exhibit with no edits says so. An edit that is not recorded is the edit the other side will find.
6. Print the parameters. The key date, the keyword list, the round-amount threshold and the transfer window are part of the method, and they print inside the exhibit so a rerun uses the same ones.
7. Hand over the register, not just the PDF. The register lists the files, their order, their declared sources and their hashes. With it and the method version, the other expert can rebuild the matter and compare the findings hash on the cover. Without it, they can only argue.

Scanned pages

A scan is read by OCR, and a different OCR build can read a picture differently. So when OCR was used, the page is marked as OCR-read, the OCR engine is named in the engine line on the cover, and it is part of the findings hash. A scanned statement that ties is proven like any other. One that does not is marked as read but not proven, and the row-level notes say where the read failed. Nothing is inferred from a scan that the scan does not show.

What a rerun proves

The same files, in the same order, with the same declared sources, the same parameters and the same edits, give the same rows, the same proofs, the same schedules and the same findings hash, on any computer that runs the method version printed on the cover. That proves the figures follow from the files. It does not prove that a statement is genuine, that a transfer had a purpose, or that a note means anything for the matter. Those are the examiner's calls, and the basis of preparation says so.

Try the procedure before a real matter

The sample exhibit on this site is built from a fictional matter. Its files are downloadable as a zip with a README that gives the order and the sources. Drop them into a free matter, under any matter name and any reviewer name, and the findings hash on screen must read the one on the sample's cover. A matter you have paid for can be rerun free for ninety days, so a statement that turns up late goes in and the exhibit is rebuilt with its hash on record.