Convert Bank Statement PDF to Excel Without Retyping

Check that your statement has a text layer

The extractor reads the text stored inside the PDF, so it only works when there is text to read. Open the statement in any PDF viewer and drag across a line of transactions. If the words highlight and you can paste them into a note, the PDF has a text layer and you can go ahead.

If nothing highlights, the page is a picture, which is what you get from scanning or photographing a paper statement. Docubox's OCR to text can read printed English from a picture like that, but it returns plain text without rows or columns, so you would still be building the spreadsheet yourself.

Convert the statement

1. Upload the PDF. Open PDF tables to XLSX and add your statement. The upload area says up front that the file is processed on Docubox's servers and that you sign in with a free account to download the result.
2. Choose the pages. The Pages box starts as all. If the transactions begin on page 2 and the last page is only small print, type a range such as 2-5 and the rest is skipped.
3. Choose XLSX or CSV. Output format starts on XLSX, which gives you a workbook. CSV gives one plain text file holding every table.
4. Select Convert to XLSX or Convert to CSV. The button follows the format you picked. Docubox reads the statement one page at a time, and the progress bar counts the pages.
5. Download after signing in. Use Google or a code sent to your email (free, no password), then compare the file with the PDF before you rely on any figure.

For a longer statement, cut out the pages you need first with Split PDF, which accepts longer files, and convert the smaller PDF.

Read the confidence notes

Docubox does not present the result as exact. On each page it looks first for tables drawn with ruling lines, then for columns of text that line up in the rest of the page, so unruled transactions below an account-summary box come out too. When part of a page looks like a table but cannot be read reliably, the result page names the page to check. Each table it keeps is rated high, medium or low, from how full it is, how consistent its rows are and whether it has ruling lines.

• In the workbook, the first sheet is called Review. It lists every table with its page, its size in rows and columns, its rating and a note saying whether it came from ruling lines or from text alignment.
• In the CSV, a comment row above each table gives its page and rating, for example # page 2 table 1 (confidence high).
• On the result page, every table rated below high gets a warning telling you to check its rows and columns.

A statement without drawn lines gives tables found from text alignment, which are rated medium at best, and low if a column boundary slices through words or the rows are patchy and uneven. Either rating means reading every row against the PDF. Look in particular for a description that runs onto a second line: it can come out as a row of its own with empty date and amount cells, and many wrapped lines can get the table dropped.

Amounts arrive as text: turn them into numbers in Excel

In the workbook, every cell is stored as text, amounts included. That keeps each value as the PDF's text has it, and a cell that starts with = is never run as a formula. The catch is that Excel does not treat those cells as numbers until you convert them; Microsoft notes that numbers stored as text can give unexpected results and that Excel usually marks them with a green triangle.

1. Use the error button. In Excel for Windows, select the amount cells, select the error indicator that appears at the corner of the selection (or press Alt+Shift+F10), then choose Convert to Number, as described in Convert numbers stored as text to numbers.
2. Or use VALUE. In an empty column, type =VALUE( with the first amount cell between the brackets, fill the formula down, then copy the new column and paste it over the old one with Paste Special > Values.
3. Fix what VALUE cannot read. VALUE converts text in a number format Excel recognizes, and Microsoft's own example turns the text $1,000 into the number 1000 (VALUE function). Anything else returns the #VALUE! error. If your bank prints letters such as CR or DR beside amounts, note which rows are credits, select only the amount column, remove the letters with Find and Replace and convert again.

Statements that span many pages

Docubox reads each page on its own, so a list of transactions that runs over four pages comes out as four tables. In the workbook they are four worksheets named p1-t1, p2-t1 and so on; in the CSV they are four blocks, one after another. If the bank repeats the column headings on every page, each table starts with its own heading row.

To get a single list, copy the rows from each worksheet under the first one and delete the repeated headings. Or choose CSV, which holds every table in one file, and delete the comment rows and blank lines before you sort. Text away from the tables, such as the bank's address, is left out, but a heading just above a text-aligned table can become its first row.

Bank statements are sensitive

A statement carries your account details and every payment you made. Docubox processes the PDF on its servers: it uploads to private storage, and Docubox's systems can read it while the tables are extracted. Before you upload, see whether your online banking offers a CSV or spreadsheet download of the same transactions. That is the bank's own data, with nothing to extract, and it never has to pass through Docubox.

Saving keeps only the converted file, never the PDF you uploaded. If you also need a redacted copy of the statement to send to someone, extract the tables first and redact afterwards. Redact PDF exports an image-only PDF with no text layer, so the extractor finds nothing in a redacted file.

If no tables are found

When nothing qualifies, the conversion stops with the message “No tables with a text layer were detected. Scanned pages need OCR first.” Possible causes:

• The PDF is a scan or a photo. There is no text to read; go back to the first section.
• The pages you chose have no table. Check the Pages box, or set it back to all.
• The layout is too loose. A text-aligned candidate that is less than half full, or whose columns would slice through more than a few of its words, is dropped rather than guessed at, so very irregular spacing can give nothing.

A run that ends with this message does not count against your conversion allowance. A PDF that needs a password to open stops with a different message; one that only restricts printing, copying or editing works.