Getting started

Camelot not working: fixing PDF table extraction and statement imports

Camelot not working? Likely causes when tables will not extract from PDFs, how to check the file, and a safe way to review statement rows.

Editorial illustration for Camelot not working: fixing PDF table extraction and statement imports

Searches for Camelot usually mean a tool that pulls tables out of PDF files. If yours returns nothing, empty cells or jumbled columns, the cause is usually the PDF, not you: a scanned page that is only an image, a locked file, tables without clear lines, or text that wraps across rows. If you meant something else by the name, use its own help page. Here are the checks that apply to extracting tables from any PDF statement.

Is the PDF text or a picture?

Open the file and try to select one line of text with the cursor. If you can select it, the page holds real text and an extraction tool has something to read. If the cursor selects the whole page like a picture, the file is a scan, and a table tool will find nothing without text recognition first. Also check that the file opens without a password; if your bank locked it, you must unlock it using your own password in your own viewer.

Why do columns come out jumbled?

  • No ruled lines between columns, so the tool guesses where one column ends.
  • A long description wraps to a second line and looks like a new row.
  • Repeated headers on every page are read as data.
  • Opening and closing balance lines sit in the middle of the table.
  • Debit and credit amounts share one column with a sign or a suffix.

What can you try?

  1. Run the tool on a single page first and view the output as a grid.
  2. If the tool offers different modes for ruled and unruled tables, try the other one.
  3. Narrow the area to the table itself so headers and footers are excluded.
  4. Clean up afterwards in a sheet: merge wrapped lines, delete repeated headers, split amount columns.
  5. Compare the result with the statement before you trust it.

Worked example: does the count add up?

A statement has 3 pages with 38, 40 and 29 transaction lines, so 107 in all. The tool returns 112 rows. Five extra rows are probably repeated headers or balance lines, which you can find by sorting on the amount column. Then add the debits and credits in the sheet and compare with the statement's own totals. If both match, the data is sound.

CheckStatementExtracted
Transaction lines107112
Extra rows to delete-5 (headers and balances)
Debits total₹52,400₹52,400 after cleanup
Credits total₹61,000₹61,000 after cleanup

Is there a simpler route for expense tracking?

If your goal is expense tracking, not building a parser, SpendixAI lets you upload a bank or card statement as PDF, CSV or Excel, then shows the rows for review before saving. Optional features such as reading an unfamiliar statement layout do process the uploaded document. We cannot promise every layout will parse perfectly, so check the rows against your statement, and if one statement misbehaves, a CSV or Excel download from your bank, if offered, is usually cleaner than a PDF.

What if the statement is a scan?

A scanned statement needs text recognition before any table tool can read it, and the result needs careful checking, because small errors in digits matter a lot in money. Where possible, ask your bank for a CSV or Excel download, or a fresh PDF that holds real text, instead of working from a scan. If you must use a scan, compare every total against the printed totals.

Whichever route you take, keep the original statement unchanged and work on copies.

Next steps

  • Confirm the PDF contains real text.
  • Test one page, then all pages.
  • Reconcile row counts and totals.
  • Delete the working copies when you finish.
SPENDIXAI JOURNAL

Useful ideas start with a clearer picture of your own spending.

Explore your reports