Turning Bank Statement PDFs into Spreadsheet Rows You Can Reconcile
How to stop retyping bank statement PDFs by hand, choose the right export shape, and reconcile the result against the printed closing balance before trusting any total.
physicsai <p>Most small-business bookkeeping does not fail on arithmetic. It fails on transcription. A bank hands you a PDF, someone reads it carefully, and two hundred rows get retyped into a spreadsheet by hand. A digit slips, a date shifts, and the closing balance no longer matches the figure the bank printed.</p>
<p>The fix is to remove the retyping step while keeping the checking step. Those are two different jobs, and treating them as one is why people either trust an extraction blindly or reject it out of habit.</p>
<h2>1. The input decides the ceiling</h2>
<p>A PDF exported from online bankinas a real text layer laid out as a table, so a converter reads structure rather than guessing. A scan of a scan has no text layer and depends entirely on optical recognition. A phone photo sits between the two: lay the page flat, keep the camera parallel to it, and keep all four corners in frame, because perspective and shadow cluster their errors near the row boundaries.</p>
<h2>2. Choose the export that matches the next step</h2>
<p>Excel when a person will open the file. CSV when it feeds an accounting import or a script, because nothing needs configuring. JSON when the rows merge with anher source and the numeric types have to survive.</p>
<p>The download shape matters just as much. Separate files keep each account distinct. A ZIP keeps a month together without merging. One merged export is fastest for a period summary, and it loses the account boundary unless you add a source-file column first.</p>
<h2>3. Keep the columns you will check against</h2>
<p>Transaction date, description, debit, credit, running or closing balance, plus currency and account identifier. If the statement prints a running balance, keep that column, because it turns verification from a judgement intrithmetic.</p>
<h2>4. Reconcile</h2>
<p>Do not sort and do not tidy. Take the last exported row and compare it against the printed closing balance. Matching to the cent means the table is very likely right. A mismatch means find the difference first, because every total downstream inherits that error.</p>
<p>The usual causes: a duplicated transaction where the statement wraps across a page boundary, a multi-line description split into two rows, a currency conversion read from the wrong column, and a bank fee sitting in a separate summary block.</p>
<h2>5. Categorise with restraint</h2>
<p the largest, least ambiguous lines first, derive the rule, apply it to the rest, and leave the lines you cannot classify visible. An unclassified row is information, not noise.</p>
<p>For a single-file check before committing a month, this browser-based converter takes a statement PDF or photo and returns Excel, CSV or JSON: <a href="https://bank-statement-converter.online/">https://bank-statement-converter.online/</a></p>
<h2>6. Know when not to automate</h2>
<p>Old archival formats, layouts that depart from the norm, and pages shot at a severe angle. If the balance does not reconcile afttwo careful passes, manual entry is the honest answer, and recording which files needed it tells you which sources are worth re-downloading.</p>
<p>The habit beats the tool. A recurring slot, every account processed in one sitting, each balance reconciled before moving on.</p>er >Worko aotg h
