PDFactory.org

Accounting module · live

A hundred receipts in.
One table of entries out.

Your scanned receipts, invoices and supporting documents, read and checked. The module finds each document — even when one page carries two — takes off the supplier, the date, the subtotal, the taxes and the total, and redoes the arithmetic before handing them back.

Drop in your receipts or invoices

PDF or images — 200 pages, 50 MB maximum

Drop in several: they make a single batch

Batch deleted 2 h after your last action

A scanned statement becomes a checked table of entries On the left, blurred lines with no structure. On the right, the same data set out in columns, the amounts right-aligned, one row flagged as out. Scanned statement Description Amount Net + tax ≠ gross — check this

What the module does

Feature Problem it solves What you get
Receipt and invoice entry Keying in supplier documents by hand, line after line. Supplier, date, number, subtotal, GST, QST and total taken off the document, as an Excel workbook, as CSV or as JSON.
Several documents per page A stack of receipts photocopied two at a time, the way everybody does it. The page is cut apart before it is read: two receipts side by side give two documents, not a blend.
Pages set back on their feet Sheets fed the wrong way up into the copier's tray. Orientation is settled page by page before reading, without anyone having to turn a hundred sheets round.
A double arithmetic check A misread figure that goes unnoticed until the filing. Subtotal + taxes = total, and each tax set against its own rate. Whatever fails to add up arrives flagged, with the discrepancy in figures.
Numeric tables cross-footed A photocopied budget, expense report or depreciation schedule you have to recompute by hand before you believe it. The grid is rebuilt, its rows added toward the totals column and its columns toward the totals rows — thirty-one cross-checks on one test budget annex — then handed back as CSV and as a sheet of the workbook. A misread figure is then refused twice, and the crossing of the two refusals names the cell.

What the module stands behind

Accuracy

Nothing is guessed

What is printed on the document is taken as it stands. What was not found stays empty rather than being worked out by subtraction: a figure invented in an accounting document is caught nowhere downstream.

Verification

A green that means something

A document is called verified only if a proof actually ran. When there is not enough to test anything at all, it is called incomplete — not sound. A green badge sitting on zero checks is worth less than no badge at all.

Correction

You decide, not the machine

When a total does not follow from its parts, the module states the discrepancy and the value the arithmetic calls for. It does not put it in place: silently replacing a figure that was read with one that was computed would make the document say what was expected of it.

What the module does not do yet

It reads receipts and invoices. A bank statement is recognised as such and its pages are read, but the structure of its table — the withdrawal column, the deposit column, the balance running from line to line — is not rebuilt yet, and the module says so rather than inventing columns. Same for a trial balance.

It does not write an FEC file nor a template specific to Sage, QuickBooks, Xero or Acomba. The CSV it produces carries named columns, which those programs import through a mapping; announcing a certified format without being able to test a single real import would be promising what we cannot verify.

It rebuilds a numeric table and cross-foots it, but only when the table announces its own totals — a "Total" column, a "Total" row. A grid of numbers that does not say how it adds up is left alone rather than guessed at: four arbitrary numbers have no reason to come out right, and flagging them would be inventing a fault.

On a table written in whole dollars, it does not flag the rounding difference : each figure there is within half a dollar of the true one, and a sum of four figures within two dollars. This was measured on a real budget, where five rows out of eleven differ from their total by exactly one dollar — the document is right, it is the rounding talking. Flagging them would make five false alarms on a sound page, and a false alarm teaches distrust of the signal, including when it tells the truth.

It takes as many files as you drop in, but not yet a ZIP archive. A two-hundred-page PDF, on the other hand, goes through in one.

Frequently asked questions

How do I pull the amounts off a scanned invoice or receipt?

You drop the document in, on its own or in a batch. The module takes off the supplier, the date, the number, the subtotal, the GST, the QST and the total, redoes the arithmetic, then hands it all back as an Excel workbook, as CSV for a spreadsheet or as JSON for the detail.

Does the module check the totals?

Yes, and twice rather than once: subtotal plus taxes must make the total, and each tax is set against its own rate — 5% for GST, 9.975% for QST. Whatever fails to add up arrives flagged with the discrepancy in figures, and is never corrected behind your back.

Can several files go into one batch?

Yes. Files stack up instead of replacing one another: five photos of receipts and a PDF of invoices make a single batch, and every page keeps the name of the file it came from. Up to 200 pages in all, 50 MB per file.

What about two receipts on the same page?

The page is cut apart before it is read: two receipts photocopied side by side give two documents, not a blend. A sheet fed into the scanner upside down is set back on its feet page by page, without anyone having to turn the stack over.

Does it read a numeric table, a budget, an expense report?

Yes, when the table announces its own totals. The grid is rebuilt, its rows added toward the totals column and its columns toward the totals rows; a misread figure is then refused twice, and the crossing of the two refusals names the cell.

Can I export the batch to Excel?

Yes. Alongside the CSV and the JSON, the batch comes out as an Excel workbook: one sheet for the entries, one per numeric table. The amounts in it are real numbers rather than text, so the column totals itself the moment the file opens, with no import assistant — and the workbook reads correctly whether your Excel is set to English or to French.

Does it read bank statements, or write an FEC file?

Not yet, neither one. A bank statement is recognised and its pages are read, but the structure of its table is not rebuilt, and the module says so rather than inventing columns. The CSV it produces carries named columns, which Sage, QuickBooks, Xero or Acomba import through a mapping.

Your accounting documents do not go to a third party

A folder of invoices tells you a company's turnover, its suppliers and its margins. It is processed in Frankfurt, on our own machines, and destroyed automatically. No third-party artificial intelligence service is called, and nothing you upload trains a model.

Our rules, in plain words

The other two modules

Live

The geomatics module turns PDF or image maps into GIS files, with georeferencing, detection by colour and correction by hand. The legal module makes a scanned case file searchable, picks out the dates, the amounts and the exhibits, and irreversibly redacts whatever you point it at. All three run on the same infrastructure and under the same rules.

See the geomatics module See the legal module