Getting started

Docparser not working? A checklist for parsing rules and output

If Docparser is not working, check the source file, the rule, the output and the connection in order, then reconcile totals so wrong data does not slip through.

Editorial illustration for Docparser not working? A checklist for parsing rules and output

When Docparser is not working, the cause is nearly always one of four things: the document changed, a parsing rule no longer matches, the document is a scan rather than readable text, or the connection that carries the data onward has failed. Test with one known-good file first. If that parses, the problem is the new document, not the tool.

This is a general troubleshooting list for any document-parsing tool. It does not describe what Docparser's features or plans do. Check its own help pages for specifics.

Start with a control file

Choose a document that parsed correctly before and upload it again. If the result is still right, the tool is healthy. If it is now wrong, something in the rules or settings changed. This one test splits the problem in half.

Why did parsing break on a file that used to work?

  • The sender changed the layout, for example moving the total or adding a column
  • The new file is a photo or a scan, so there is no text to read
  • The file is password protected or corrupted
  • A page order change put the table on page two instead of page one
  • Dates or numbers now use a different format, for example 1,24,000.00 against 124,000.00

A numbered way to find the break

  1. Open the failed document and compare it with a good one side by side.
  2. Check each field the rule expects: is it in the same place and spelt the same?
  3. Look at the output and note which fields are empty or wrong.
  4. Adjust only that rule and reparse the one file.
  5. Reparse the control file to confirm you did not break the old layout.
  6. Only then run the batch again.

How do I know the numbers are right?

Never trust a clean-looking output. Add up the line items yourself and compare with the document's own total. Say a vendor invoice lists three lines: ₹12,500, ₹8,400 and ₹3,100. Their sum is ₹24,000. If the parsed total reads ₹21,500, a line was missed, and 24,000 - 21,500 = ₹2,500 does not match any single line. That means a digit was misread somewhere, so open the document and check each.

CheckHowPass if
Line totalAdd the parsed linesEquals the invoice subtotal
CountCount parsed rows against the documentSame number of lines
DatesRead three at randomThey match the document
DuplicatesSort by invoice numberNo repeats

When the data goes on to another system

If parsed data feeds a spreadsheet or accounting tool, check that connection separately: is the destination reachable, has a sign-in expired, did a column get renamed? A failure there looks like the parser stopped, when the parse itself worked.

For your own statements

If the documents are your personal bank or card statements rather than business invoices, you may not need a parsing tool at all. SpendixAI accepts statements as PDF, CSV or Excel and shows the rows for review before saving, and an optional feature can read an unfamiliar statement layout, which processes the uploaded document. It does not promise that every layout will parse perfectly, so keep the review step.

How do I set up a small test routine?

Keep three sample documents: one simple, one with many lines and one with an awkward layout. After any change to a rule, reparse all three and compare the totals with the originals. It takes five minutes and catches the common case where fixing one document breaks another. Write the expected totals on a note so the comparison is fast and does not rely on memory.

Before you write to support

  • The document that failed, with any personal details hidden
  • The control file that still works
  • What changed and when
  • The exact error message
SPENDIXAI JOURNAL

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

Explore your reports